"TSMC making Taiwan its global R&D hub and manufacturing hub has not changed," Wang said. "TSMC investing in Taiwan, investing in Kaohsiung, also has not changed."

TSMC started building its Kaohsiung plant last year. The plant will produce 28-nanometer semiconductors. Although the company initially planned to include advanced 7-nanometer chips in the expansion, it postponed these plans.

TSMC has ramped up its investments abroad recently to cater to the semiconductor shortage that happened during the pandemic. The company has announced new chip factories in the United States and Japan. However, it is still committed to its operations in Taiwan. Plans to develop more fabs come amid ever-increasing demand for semiconductors, which are used in a range of products - from smartphones to cars. TSMC's clients include major tech companies like Apple, Nvidia, and Qualcomm.

The chip-making company had previously announced that it expects lower demand due to a slowing global economy and would decrease its capital expenditure this year to $32-$36 billion, from $36.3 billion last year.

Source: Reuters

TSMC says that it is still committed to Taiwan's investment plans, amid slowdown reports

The FAQ page on the game's official website has the numbers for four levels of PC gaming. As you can see, the hardware requirements get much higher if you want to play the game at more than 1080p resolution:

Minimum

• Frames per second - 30
• Resolution - 1080p
• OS - Windows 10
• CPU - AMD FX-9590, Intel Core i7-77000HQ
• RAM - 10GB
• Storage - 70GB
• GPU - AMD Radeon RX 480, NVIDIA GeForce GTX 1060

Recommended

• Frames per second - 60
• Resolution - 1080p
• OS - Windows 10
• CPU - AMD Rysen 5 5600X, Intel Core i9-9900K
• RAM - 16GB
• Storage - 70GB
• GPU - AMD Radeon RX 6600 XT, NVIDIA GeForce RTX 2070 Super

High

• Frames per second - 60
• Resolution - 1440p
• OS - Windows 10
• CPU - AMD Ryzen 7 7700X, Intel Core i5-12600KF
• RAM - 16GB
• Storage - 70GB
• GPU - AMD Radeon RX 6750 XT, NVIDIA GeForce RTX 3070

Ultra

• Frames per second - 60
• Resolution - 4K
• OS - Windows 10
• CPU - AMD Ryzen 9 7900 X, Intel Core i7-13700K
• RAM - 16GB
• Storage - 70GB
• GPU - AMD Radeon RX 6950 XT, NVIDIA GeForce RTX 3090

The FAQ page also has some other info about the game. The PC version only supports DirectX 12, but if you have an AMD Radeon graphics card, it will support both AMD FidelityFX Super Resolution (FSR2) and AMD FidelityFX Contrast Adaptive Sharpening.

While it won't support online cross-play between platforms, it will support cross-gen for PlayStation 4 and 5 and Xbox One, Xbox Series X, and Xbox Series S, with more info on that to be revealed later. There will be co-op support in Dead Island 2 for up to four players, and again more info on that will be revealed at a later date. You can pre-order the game on Amazon right now.

As an Amazon Associate when you purchase through links on our site, we earn from qualifying purchases.

Dead Island 2 has four levels of PC hardware requirements including an "Ultra" level

As announced by Steve Jobs Archive earlier this month, the free memoir on the life of Apple co-founder Steve Jobs is now up for grabs. The digital memoir, which collates Steve's past speeches, emails, and interviews, is edited by the Archive's executive director Leslie Berlin.

The memoir also features some rarely seen stuff, including internal meeting transcripts and childhood photos of the Apple co-founder. It is divided into three parts and offers a timeline view of major events that happened throughout his life. For instance, Steve's perspective on his childhood, getting kicked out of Apple, and his experience with NeXT and Pixar.

Steve Jobs Archive has made the ebook available via multiple channels, including Apple Books for iPhone, iPad, or Mac, and on Libby app for participating libraries. Users can also read the ebook directly on the Archive's website or download the EPUB file that's around 32 MB in size and use their preferred e-reader like Kindle or Nook.

Steve Jobs 'Make Something Wonderful' free ebook is now available to download

Apple Mac shipments declined by 40% in Q1

Apple's Senior Vice President and Chief Financial Officer, Luca Maestri, had revealed in an earnings call in February, that the company expects revenue from Mac to drop by a double-digits percentange in Q1, 2023. The company's CEO, Tim Cook had said that a year-on-year comparison to previous successful launches made comparisons of sales figures difficult.

The Cupertino company halted the production of its M2 chipsets in January. This was seen as a shock, as the Silicon Valley giant had never before stopped manufacturing a current-gen chipset. According to a report by Bloomberg, PC market shipments saw a slump of 29% to 56.9 million units. Lenovo and Dell registered drops of over 30%, ASUS' numbers dropped by 30.3%, while HP saw a 24.2% decline. Apple however took the hardest hit with a decline of 40.5%, its worst drop since Q3 of 2000. This may explain why the company had to stop the production of the M2 chips.

It is worth taking into account the part that the COVID 19 pandemic played in the sales of computers. A lot of people had to work from home, and even study remotely during the pandemic-era. This caused a sharp increase in the sales of mobile phones, and computers, specifically laptops such as MacBooks. Once the pandemic ended and people returned to their offices and schools, there was a steep decline in the demand for such devices. This is likely what Cook was referring to about comparing the current economic slowdown against the sales numbers of the previous years.

Anurag Rana, a Bloomberg Intelligence analyst, says that a rebound is possible for the PC market in the 2nd half of 2023. Apple is expected to announce its M3 MacBooks and iMac at the WWDC 23 event in June, so it could benefit from the demand.

Personally, I think that high prices are to blame too. When Apple launched its M2 MacBooks last year, many people were surprised by the prices of the devices. They were quite high, in fact the new machines were more expensive than the previous gen models, while just offering a marginal upgrade over them. I use a MacBook Air M1, and while I love the laptop, I find it absurd that the company is still sticking to 8 GB of RAM, and 256 GB of storage for the basic model of its new MacBooks. The tech community was riddled with complaints about how the SSDs on the M2 Macs had slower read and write speeds than the previous gen models. The M1 laptops were not only faster, but offered a similar battery life at a cheaper price tag.

It's not like people have to buy new computers every year. Apple's machines last for a long time, so it wouldn't make sense to expect consumers to upgrade their systems frequently, like they may do with iPhones. Apple rarely offers discounts for its devices, at least through its official channels, i.e. Apple Stores.

If you want a good deal for your Mac, look for an officially refurbished one, as it comes with a regular warranty. The condition of the device is as good as a new one, the only difference is that it usually ships in a different box. The only problem is that Apple does not sell refurbished devices in a lot of Countries. Apple's Education discounts can be a good option for students or teachers.

Apple hit hardest by decline in personal computer shipments in Q1

This first update is pretty minor, at least compared to the many, many changes and improvements that Microsoft has put into its own Bing Chat since it launched in February. In fact, there are only two updates listed. One was hinted at several days ago by Google CEO Sundar Pichai. He stated that Bard would switch from the smaller LaMDA language model to the larger PaLM model. The Bard update page states the chatbot has now been updated to provide "better capabilities for math and logic." That's likely due to the switchover to the PaLM model.

The other update listed is that Bard now includes "additional suggested Search topics when people click 'Google it.'". The page states, "People will be able to explore a broader range of interests with more related topics." Hopefully we will get more detailed updates on Bard's additions and improvements via this page over the coming weeks and months.

Google Bard now has an "experiment updates" page to promote its latest improvements

As per the sales figure available for the German retailer Mindfactory, AMD Ryzen 7 7800X3D is the best-selling CPU this week, leaving behind the previous-gen Ryzen 7 5800X3D.

Few days ago, AMD Ryzen 7 7800X3D was released with great reviews. This is because the CPU comes with a 3D V-Cache technology, which adds an extra layer of 64 MB cache on top an existing 32 MB cache. Taking the total L3 CPU cache to 96 MB.

This relatively huge cache increase gives the processor a massive boost in cache intensive tasks, such as gaming. In-fact, so impressive is the boost, that AMD Ryzen 7 7800X3D is possibly the best gaming CPU ever made. It has almost same or better performance than the higher priced AMD Ryzen 9 7950X3D and Intel Core i9-13900K processors. This while consuming half the power while being cheaper too.

Everyone was waiting for Ryzen 7 7800X3D to release from some time now. The reviews weren’t disappointing either. So when the processor got released, everyone expected it to sell a lot. What we didn’t expect is it to sell so much.

Best Seller In Germany

Whenever we want to see how a CPU or a GPU is doing in terms of sale figures, one needs to look at the German retailer Mindfactory. Mindfactory is a big computer hardware seller in the European country. What it, however, is best known for, is that it shares it’s weekly sales report with its associates. This data then gets available to the public.

In a by tweet by @TechEpiphany, who regularly shares this data about CPU, GPU and others, has shared the sale figures for the top CPUs sold at Mindfactory in the previous week. The results are somewhat surprising.

In the week when AMD Ryzen 7 7800X3D was just launched, it has become the best-selling CPU in Mindfactory in Germany.

In that time period, 1860 units of AMD Ryzen 7 7800X3D CPU were sold. The next in the list was the previous generation X3D processor, the AMD Ryzen 7 5800X3D, with 1070 units sold. With the AMD Ryzen 7 5700X is the third in the list with 380 units sold.

It’s important to mention that Intel is way behind in the list. At tenth position, Intel’s bestseller is the Core i7-13700K, with just 170 units sold. While it’s true that i7-13700K is a great processor which provides a great value and can stand on its own against the more powerful and expensive processors. The problem is the numbers, where 1000+ units sold for AMD’s top processor and where 100+ units sold for Intel’s one.

Not to forget, TechEpiphany’s tweet also reveals that overall, AMD is having 83% market-share and 85% revenue compared to Intel’s 17% and 15% respectively during the same period.

Conclusion

Mindfactory is a big German retailer. But it doesn’t represent the whole of Europe or the world. For example, the AMD Ryzen 7 5800X3D is the top-selling processor at Mindfactory from almost a year. The Ryzen 5800X3D, which no matter how great, is not cheap.

So it’s quite possible that people from poorer countries might not buy the $449 CPUs and might prefer the sub-$200 or sub-$300 ones instead. So Intel might be doing nicely in other countries. But this revelation gives us a fair idea of where things stand.

Another thing which is important is that AMD Ryzen 7 5800X3D remains one of the top seller too. This because it has become cheaper than the original MSRP and even a good quality previous-gen AMD AM4 motherboard is cheaper and easier to find than a good quality current-gen AMD AM5 motherboard which AMD Ryzen 7 7800X3D requires.

Not to forget, DDR5 RAM is still slightly more expensive than DDR4 RAM. However, that difference is massively reduced in recent times.

Intel on the other hand can do better. We still believe Intel will need to bring a stacked 3D cache version of its own if it wants to have a good fight against AMD in the top CPU sales everywhere.

AMD Ryzen 7800X3D Becomes The Best Selling CPU

Late on Friday, the My Cloud status page posted an update, stating that while its online storage service may not be working, users who have files on a local storage device can now access them with a feature called Local Access. It states:

The Local Access feature allows you to directly access your personal files from a Windows or macOS computer that is connected to the same network as your device. To enable Local Access, use your favorite browser and connect to your device's Dashboard. Then enable the Local Access feature and create a new Local Access account.

You can go to this support page for more information on Local Access.

Meanwhile, we are still waiting for more information about the Western Digital attack on its network, including whether any personal information was stolen. The company's online store has been down since the incident. Hopefully, Western Digital will offer an update on the company's status soon.

Western Digital My Cloud is still down, but now allows for Local Access for user files

Google cuts off third-party smart displays as Assistant support dwindles

Nowadays, we have countless options for buying a technological device, like a PC, Mac, tablet, etc. There are different things to consider, such as operating systems, and sometimes this might be an issue. However, most of these issues can be fixed either with third-party apps or different methods.

How to transfer files from Android to PC or Mac: Full guide

You have many different options for transferring files from Android to PC or Mac. Below you will find the most-used ways, and you can choose whichever suits you best. Let's cut to the chase and give you all the necessary information!

USB Cable

Of course, we will start with the most traditional way possible. Buying an Android device, whether a smartphone or a tablet, will come with a USB cable. The most important point of companies putting these cables in the boxes because they help these devices charge. However, you could also plug the USB cable into your PC or Mac to transfer files from your Android device.

1. Connect your Android device to your PC or Mac with the USB cable.
2. Tap Allow access on your Android device. (If you are using a newer Android version, select "File transfer" on the USB Preferences menu)
3. Click My Computer on your Windows PC and find your device. On Mac, your device will be visible on the desktop.
4. Click and drag any files you want to transfer from Android to PC or Mac.

Bluetooth

Bluetooth is mostly used for transfers between two smartphones, tablets, etc. It is not the most common way when it comes to Windows or Mac computers. However, it is still a choice that you might want to consider.

1. Enable Bluetooth on your PC, Mac, and Android device.
2. Go to the Bluetooth menu on your computer.
3. Click search for available devices and pair them with your Android device. (If you are on a Mac, enable Bluetooth Sharing)
4. On your Android device, go to the file or image you want to transfer, tap Share, and select Bluetooth.
5. Right-click the Bluetooth icon on your taskbar and select "Receive a file" on Windows. You don't have to do anything on your Mac, as the file will appear in your Downloads folder.

Google Drive

Google Drive is one of the most-used cloud-based storage solutions. There are other cloud storage solutions, too. For example, Dropbox is another cloud-based solution that you could try. What is appealing about cloud-based storage solutions is that you can access them anywhere, time, as long as you are connected to the internet.

Dropbox and Google Drive are alternatives to each other; you can also add Microsoft OneDrive to the list. You can upload your file to the cloud on your Android device and log into your account from a PC or Mac to download it. Today, we will cover Google Drive. Here is how to do it:

1. Go to the Google Drive app on your Android device.
2. Click the "+" button,  hit "Upload," and select all the files you want to transfer. This will upload your files to the cloud.
3. On your PC or Mac, go to the official Google Drive website.
4. Log in to your Google account.
5. Find the files you have uploaded from your Android device.
6. Click the three dots at the top right of your file.
7. Select Download.

Snapdrop

Snapdrop lets two different devices connected to the same Wi-Fi share files, which is what we are looking for in this article! Let's cut to the chase and look at how to use it:

1. Download the official Snapdrop app on your Android device.
2. Go to the official Snapdrop website on your PC or Mac.
3. Connect your Android device and PC or Mac to the same Wi-Fi network.
4. Open the app on your Android device, and select your PC or Mac on the screen.
5. Choose the files you want to transfer.

Related: Can't use Google Nearby Share? Use open source Snapdrop instead

FTP Server

Lastly, you can also set up an FTP server to transfer files from Android to PC or Mac. FTP stands for File Transfer Protocol; if you are a newbie, the other four steps are better and easier for you.

1. Download Solid Explorer on your Android device.
2. Tap the three-line menu icon at the top left. (Also known as the hamburger menu)
3. Scroll down and select FTP Server.
4. Launch File Explorer on your PC and type the server address you see on your Android device. It starts with "ftp://192" and goes on. You will see all your Android files there.
5. If you are using a Mac, go to Finder and click Go at the top menu bar. Select Connect to Server and type the same server address as seen in the previous step.

How to transfer files from Android to PC or Mac: 5 easy ways

Last-gen ultralight laptops are nearly as fast as new models—and much cheaper

Both IDC and Canalys agree that Lenovo shipped the most PCs in the last quarter, followed by HP, Dell, Apple, and ASUS. Apple PCs suffered the biggest shipment drops during Q1 2023, with IDC stating its numbers were down 40.5 percent from the same period a year ago, and Canalys claiming Apple's PC shipments were down 45.5 percent.

Despite the bad news, both firms claim that the PC business could see the start of a recovery in the second half of 2023 and a larger recovery predicted for 2024. IDC says that by 2024, there will be a lot of older PCs that need upgrading. It added:

If the economy is trending upwards by then, we expect significant market upside as consumers look to refresh, schools seek to replace worn down Chromebooks, and businesses move to Windows 11. If recession in key markets drags on into next year, recovery could be a slog.

Canalys stated:

The PC market has strong fundamentals to drive long-term growth, with shipment volumes higher than in the pre-pandemic era. A much larger installed base post-COVID-19, the transition to Windows 11, and both refresh and new demand from digital education will all be key drivers as the global economy enters a period of recovery in 2024 and beyond.

That firm added that for Q1 2023, notebook PC shipments were down 32 percent to 41.8 million units, while desktop PCs declined slightly less, down 28 percent to 12.1 million units.

Analysts claim global PC shipments went down as much as 33 percent in Q1 2023

While it is coming back as a live event, there will be some changes. While previous in-person events have been free to attend, the QuakeCon site now says it will only be for paid and pre-registered attendees. There will also be no general admission to the event the days of QuakeCon. Also, this year the event will concentrate on the BYOC hall and there will be no exhibit hall. There will still be lots of events in the BYOC, including TableTop Village, community events, and a QuakeCon party. It will also host the world finals of the Quake Champions tournament. The event will also be live-streamed on Twitch.

Registration begins tomorrow, April 11, for the Elite BYOC Ticket drawing. You can enter your name at that time to be selected into the drawing for a chance at one of these limited $400 tickets. They include some extra benefits like priority seat selection and check-in, extra space for your gaming PC setup, and a mug that will get unlimited refills of Pecos Pete beverages. Two other packages, the $75 standard ticket, and the $200 Premium Ticket, will go on sale normally on April 21.

It's likely that id Software and Bethesda Softworks will demo some upcoming games. We have our fingers crossed that will include showing off the next game or games from id Software's franchises like Doom, Quake, and Wolfenstein.

QuakeCon returns as an in-person event in Grapevine, Texas August 10-13

Inside Silicon Valley’s AI sector, fierce divisions are growing over the impact of a new wave of artificial intelligence: While some argue it’s imperative to race ahead, others say the technology presents an existential risk.

Those tensions took center stage late last month, when Elon Musk, along with other tech executives and academics, signed an open letter calling for a six-month pause on developing “human-competitive” AI, citing “profound risks to society and humanity.” Self-described decision theorist Eliezer Yudkowsky, co-founder of the nonprofit Machine Intelligence Research Institute (MIRI), went further: AI development needs to be shut down worldwide, he wrote in a Time magazine op-ed, calling for American airstrikes on foreign data centers if necessary.

The policy world didn’t seem to know how seriously to heed these warnings. Asked if AI is dangerous, President Biden said Tuesday, “It remains to be seen. Could be.”

The dystopian visions are familiar to many inside Silicon Valley’s insular AI sector, where a small group of strange but influential subcultures have clashed in recent months. One sect is certain AI could kill us all. Another says this technology will empower humanity to flourish if deployed correctly. Others suggest the six-month pause proposed by Musk, who will reportedly launch his own AI lab, was designed to help him catch up.

The subgroups can be fairly fluid, even when they appear contradictory and insiders sometimes disagree on basic definitions.

But these once-fringe worldviews could shape pivotal debates on AI. Here is a quick guide to decoding the ideologies (and financial incentives) behind the factions:

AI SAFETY

The argument: The phrase “AI safety” used to refer to practical problems, like making sure self-driving cars don’t crash. In recent years, the term — sometimes used interchangeably with “AI alignment” — has also been adopted to describe a new field of research to ensure AI systems obey their programmer’s intentions and prevent the kind of power-seeking AI that might harm humans just to avoid being turned off.

Many have ties to communities like effective altruism, a philosophical movement to maximize doing good in the world. EA, as it’s known, began by prioritizing causes like global poverty but has pivoted to concerns about the risk from advanced AI. Online forums, like Lesswrong.com or AI Alignment Forum, host heated debates on these issues.

Some adherents also subscribe to a philosophy called longtermism that looks at maximizing good over millions of years. They cite a thought experiment from Nick Bostrom’s book “Superintelligence,” which imagines a safe superhuman AI could enable humanity to colonize the stars and create trillions of future people. Building safe artificial intelligence is crucial to secure those eventual lives.

Who is behind it: In recent years, EA-affiliated donors like Open Philanthropy, a foundation started by Facebook co-founder Dustin Moskovitz and former hedge funder Holden Karnofsky, have helped seed a number of centers, research labs and community-building efforts focused on AI safety and AI alignment. FTX Future Fund, started by crypto executive Sam Bankman-Fried, was another major player until the firm went bankrupt after Bankman-Fried and other executives were indicted on charges of fraud.

How much influence do they have?: Some work at top AI labs like OpenAI, DeepMind and Anthropic, where this worldview has led to some useful ways of making AI safer for users. A tightknit network of organizations produces research and studies that can be shared more widely, including this 2022 survey that found that 10 percent of machine learning researchers say AI could end humanity.

AI Impacts, which conducted the study, has received support from four different EA-affiliated organizations, including the Future of Life Institute, which hosted Musk’s open letter and received its biggest donation from Musk. Center for Humane Technology co-founder Tristan Harris, who once campaigned about the dangers of social media and has now turned his focus to AI, cited the study prominently.

(Illustration by Elena Lacey/The Washington Post; Photos by The Washington Post; Getty Images; Twitter)

AGI BELIEVERS

The argument: It’s not that this group doesn’t care about safety. They’re just extremely excited about building software that reaches artificial general intelligence, or AGI, a term for AI that is as smart and as capable as a human. Some are hopeful tools like GPT-4, which OpenAI says has developed skills like writing and responding in foreign languages without being instructed to do so, means they are on the path to AGI. Experts explain that GPT-4 developed these capabilities by ingesting massive amounts of data, and most say these tools do not have a humanlike understanding of the meaning behind the text.

Who is behind it?: Two leading AI labs cited building AGI in their mission statements: OpenAI, founded in 2015, and DeepMind, a research lab founded in 2010 and acquired by Google in 2014. Still, the concept might have stayed on the margins if not for the same wealthy tech investors interested in the outer limits of AI. According to Cade Metz’s book, “Genius Makers,” Peter Thiel donated $1.6 million to Yudkowsky’s AI nonprofit and Yudkowsky introduced Thiel to DeepMind. Musk invested in DeepMind and introduced the company to Google co-founder Larry Page. Musk brought the concept of AGI to OpenAI’s other co-founders, like CEO Sam Altman.

How much influence do they have?: OpenAI’s dominance in the market has flung open the Overton window. The leaders of the most valuable companies in the world, including Microsoft CEO Satya Nadella and Google CEO Sundar Pichai, now get asked about and discuss AGI in interviews. Bill Gates blogs about it. “Because the upside of AGI is so great, we do not believe it is possible or desirable for society to stop its development forever,” Altman wrote in February.

(Illustration by Elena Lacey/The Washington Post; Photos by The Washington Post; Getty Images; Twitter)

AI DOOMERS

The argument: Though doomers share a number of beliefs — and frequent the same online forums — as people in the AI safety world, this crowd has concluded that if a sufficiently powerful AI is plugged in, it will wipe out human life.

Who is behind it?: Yudkowsky has been the leading voice warning about this doomsday scenario. He is also the author of a popular fan fiction series, “Harry Potter and the Methods of Rationality,” an entry point for many young people into these online spheres and ideas around AI.

His nonprofit, MIRI, received a boost of $1.6 million in donations in its early years from tech investor Thiel, who has since distanced himself from the group’s views. The EA-aligned Open Philanthropy donated about $14.8 million across five grants from 2016 to 2020. More recently, MIRI received funds from crypto’s nouveau riche, including ethereum co-founder Vitalik Buterin.

How much influence do they have?: While Yudkowsky’s theories are credited by some inside this world as prescient, his writings have also been critiqued as not applicable to modern machine learning. Still, his views on AI have influenced more high-profile voices on these topics, such as noted computer scientist Stuart Russell, who signed the open letter.

In recent months, Altman and others have raised Yudkowsky’s profile. Altman recently tweeted that “it is possible at some point [Yudkowsky] will deserve the nobel peace prize” for accelerating AGI, later also tweeting a picture of the two of them at a party hosted by OpenAI.

(Illustration by Elena Lacey/The Washington Post; Photos by The Washington Post; Getty Images; Twitter)

AI ETHICISTS

The argument: For years, ethicists have warned about problems with larger AI models, including outputs that are biased against race and gender, an explosion of synthetic media that may damage the information ecosystem, and the impact of AI that sounds deceptively human. Many argue that the apocalypse narrative overstates AI’s capabilities, helping companies market the technology as part of a sci-fi fantasy.

Some in this camp argue that the technology is not inevitable and could be created without harming vulnerable communities. Critiques that fixate on technological capabilities can ignore the decisions made by people, allowing companies to eschew accountability for bad medical advice or privacy violations from their models.

Who is behind it?: The co-authors of a farsighted research paper warning about the harms of large language models, including Timnit Gebru, former co-lead of Google’s Ethical AI team and founder of the Distributed AI Research Institute, are often cited as leading voices. Crucial research demonstrating the failures of this type of AI, as well as ways to mitigate the problems, “are often made by scholars of color — many of them Black women,” and underfunded junior scholars, researchers Abeba Birhane and Deborah Raji wrote in an op-ed for Wired in December.

How much influence do they have?: In the midst of the AI boom, tech firms like Microsoft, Twitch and Twitter have been laying off their AI ethics teams. But policymakers and the public have been listening.

Former White House policy adviser Suresh Venkatasubramanian, who helped develop the blueprint for an AI Bill of Rights, told VentureBeat that recent exaggerated claims about ChatGPT’s capabilities were part of an “organized campaign of fear-mongering” around generative AI that detracted from stopped work on real AI issues. Gebru has spoken before the European Parliament about the need for a slow AI movement, ebbing the pace of the industry so society’s safety comes first.

Source

The friction between open source and API-based distribution is one of the most interesting battles looming in the generative AI ecosystem. In the text-to-image domain, the release of Stable Diffusion clearly signaled that open source was a viable distribution mechanism for foundational models. However, the same cannot be said in the large language model (LLM) space, in which the biggest breakthroughs are coming from models like GPT-4, Claude, and Cohere, which are only available via APIs. The open source alternatives to these models haven’t shown the same level of performance, specifically in their ability to follow human instructions. However, an unexpected research breakthrough and a leaked release are starting to change that.

A few weeks ago, Meta AI announced Llama, an LLM designed to advance research in the space. Llama was released in different versions, including 7B, 13B, 33B, and 65B parameters, and despite being notoriously smaller than alternative models, was able to match the performance of GPT-3 across many tasks. Llama was not initially open-sourced, but a week after its release, the model was leaked on 4chan, sparking thousands of downloads.

What could have been seen as an unfortunate incident has become one of the most interesting sources of innovation in the LLM space in the last few weeks. Since the leak of Llama, we have seen an explosion of innovation in LLM agents built on it. Just to cite a few examples:Stanford University released Alpaca, an instruction following model based on LLama 7B model.

• Researchers from  UC Berkeley, CMU, Stanford, and UC San Diego open sourced Vicuna, a fine-tuned version of LLama that matches GPT-4 performance.
• Berkeley AI Research Institute(BAIR) released Koala, a version of LLama fine-tuned using internet dialogs.
• Nebuly open sourced ChatLLama, a framework for creating conversational assistants using your own data.
• FreedomGPT is an open source conversational agent based on Alpaca which is based on LLama.
• The Colossal-AI project from UC Berkeley released ColossalChat, a ChatGPT type model with a complete RLHF pipeline based on LLama.

Several other projects are worth mentioning in this list, and I am sure more will be released soon. One thing is certain: the accidental leak of Llama might have turned out to be one of the biggest sparks of innovation in the open source LLM space.

🔎 ML Research

OpenAI Safety

OpenAI published a detailed blog post outlining some of the principles used to ensure safety in their models. The post emphasize in areas such as privacy, factual accuracy and harmful content prevention which are essential for the wide adoption of foundation models —> Read more.

BloombergGPT

Bloomberg published a paper introducing BloombergGPT, a 50 billion LLM fine tuned in financial data. The model is based on BLOOM and fine tuned on a 363 billion token dataset —> Read more.

Segment Anything

Meta AI  published a paper outlining the Segment Anything Model(SAM), a large scale model for image segmentation. The model was open sourced together with Segment Anything 1-Billion mask dataset (SA-1B), the largest computer vision segmentation ever released —> Read more.

Koala

Berkeley AI Research(BAIR) released a paper detailing Koala, a dialogue model fine tuned for academic research. The model is based on Meta AI’s Llama and matches the performance of ChatGPT —> Read more.

BayesOpt for Hyperparameter Optimization

Google Research published a paper that models hyperparameter optimization as a Bayesian optimization problem. The paper proposes Hyper BayesOpt, a hyperparameter optimization algorithm that removes the need quantifying model parameters for Gaussian processes in BayesOpt —> Read more.

🤖 Cool AI Tech Releases

Vicuna

Vicuna is an open source Chatbot based on Meta AI Llama which matches ChatGPT quality —> Read more.

 ColossalChat

The team from the Colossal-AI project open sourced ColossalChat, an open source clone of ChatGPT with RLHF capabilities —> Read more.

🛠 Real World ML

Generative AI at LinkedIn

Linkedin discusses some of the lessons learned and best practices for building generative AI application —> Read more.

Lyft Recommendations

Lyft discusses the ML models and architecture used in their recommendation systems —> Read more.

📡AI Radar

•     AI legends Andrew Ng and Yann LeCun recorded a session expressing their oppposition to the AI moratorium proposal.
•     Quantexa raised $129 million for its AI-based financial fraud prevention platform.
•     Adthos launched its platform for creating audio ads using generative AI.
•     Meta discussed their initiatives to use generative AI to create ads.
•     Robotics company Covariant raised another $75 million.
•     AI-search company Glean incorporated generative AI capabilities into its search platform.
•     Some leaked documents revealed OpenAI’s competitor Anthropic’s intentions to raise about $5 billion in the next two years.

Source

Everyone seems to be worried about the potential impact of artificial intelligence (AI) these days. Even technology leaders including Elon Musk and the Apple co-founder Steve Wozniak have signed a public petition urging OpenAI, the makers of the conversational chatbot ChatGPT, to suspend development for six months so it can be “rigorously audited and overseen by independent outside experts”.

Their concerns about the impact AI may have on humanity in the future are justified – we are talking some serious Terminator stuff, without a Schwarzenegger to save us. But that’s the future. Unfortunately, there’s AI that’s being used right now which is already starting to have a big impact – even financially destroy – businesses and individuals. So much so that the US Federal Trade Commission (FTC) felt the need to issue a warning about an AI scam which, according to this NPR report “sounds like a plot from a science fiction story”.

But this is not science fiction. Using deepfake AI technology, scammers last year stole approximately $11m from unsuspecting consumers by fabricating the voices of loved ones, doctors and attorneys requesting money from their relatives and friends.

“All [the scammer] needs is a short audio clip of your family member’s voice – which he could get from content posted online – and a voice-cloning program,” the FTC says. “When the scammer calls you, he’ll sound just like your loved one.”

And these incidents aren’t limited to just consumers. Businesses of all sizes are quickly falling victim to this new type of fraud.

That’s what happened to a bank manager in Hong Kong, who received deep-faked calls from a bank director requesting a transfer that were so good that he eventually transferred $35m, and never saw it again. A similar incident occurred at a UK-based energy firm where an unwitting employee transferred approximately $250,000 to criminals after being deep-faked into thinking that the recipient was the CEO of the firm’s parent. The FBI is now warning businesses that criminals are using deepfakes to create “employees” online for remote-work positions in order to gain access to corporate information.

Deepfake video technology has been growing in use over the past few years, mostly targeting celebrities and politicians like Mark Zuckerberg, Tom Cruise, Barack Obama and Donald Trump. And I’m sure that this election year will be filled with a growing number of very real-looking fake videos that will attempt to influence voters.

But it’s the potential impact on the many unsuspecting small business owners I know that worries me the most. Many of us have appeared on publicly accessed videos, be it on YouTube, Facebook or LinkedIn. But even those that haven’t appeared on videos can have their voices “stolen” by fraudsters copying outgoing voicemail messages or even by making pretend calls to engage a target in a conversation with the only objective of recording their voice.

This is worse than malware or ransomware. If used effectively it can turn into significant, immediate losses. So what do you do? You implement controls. And you enforce them.

This means that any financial manager in your business should not be allowed to undertake any financial transaction such as a transfer of cash based on an incoming phone call. Everyone requires a call back, even the CEO of the company, to verify the source.

And just as importantly, no transaction over a certain predetermined amount must be authorized without the prior written approval of multiple executives in the company. Of course there must also be written documentation – a signed request or contract – that underlies the transaction request.

These types of controls are easier to implement in a larger company that has more structure. But accountants at smaller businesses often find themselves victim of management override which can best be explained by “I don’t care what the rules are, this is my business, so transfer the cash now, dammit!” If you’re a business owner reading this then please: establish rules and follow them. It’s for your own good.

So, yes, AI technology like ChatGPT presents some terrifying future risks for humanity. But that’s the future. Deepfake technology that imitates executives and spoofs employees is here right now and will only increase in frequency.

Source

Though the task is daunting—a recent test with the tax code "completely baffled" ChatGPT—leaders of the Johns Hopkins University-based effort not only believe it's doable, they worry a corporate-funded effort will beat them to the punch and use their own AI to find even more tax loopholes.

"The rules of the game are if you're operating within the law, how you bend the law to save on taxes is OK," said Benjamin Van Durme, a Johns Hopkins computer scientist specializing in AI who's leading the effort. "It's very difficult to predict all possible ways a law will work in the real world. The point of using AI is it could guess ahead before these laws get locked down."

What's understood as "tax law" is a massive, unwieldly array of laws enacted by Congress, Treasury regulations, Internal Revenue Service rulings, and court decisions. Even if all of it had been enacted with the best intentions, workarounds exist for taxpayers clever enough to find them.

"The problem is when a clever taxpayer takes one ruling and combines it with another to produce an idea that saves a lot of money," said Andrew Blair-Stanek, a tax attorney turned law professor who is part of the AI team. "It's the pairing of all these tax laws by clever lawyers."

The IRS estimates an annual tax gap of about $500 billion, and some part of that results from these clever pairings. Hoping to narrow the gap with better technology, Blair-Stanek enrolled at Johns Hopkins to get a doctoral degree in computer science in Van Durme's lab. They recruited another student, Nils Holzenberger, to help create an AI system they're calling Shelter Check.

The idea is to create software that Congress, the IRS, or the courts, could use to easily scan proposed tax legislation or rulings for loopholes they might unintentionally create.

"That's why we call it Shelter Check—it's like a spell checker, but for tax shelters," Van Durme said. "We want to build a system that could read proposed changes in the law and inform Congress and the IRS about ramifications for the tax code, or warn people writing new policies about unintended side effects."

Such sophisticated analysis will require a machine that can not only read, parse and understand complex tax law, but blend all of it like an accountant or a tax attorney.

The project is in the early stages, with the team building on the lab's extensive work in machine learning to explore language processing options and test them. As yet, they haven't found a way to conquer the legalese let alone replicate the deductive reasoning of a real human.

Despite recent advances in language-based artificial intelligence, the latest in tech is no match for tax code. First, while the material is text-based, the legal language it's written in is complex if not downright arcane. Then, the thousands of pages of documents are peppered with tables that are vital to interpreting tax outcomes.

The team is experimenting now with language prediction technology like ChatGPT and GPT-3. Tax code stumped both.

"GPT-3 was completely baffled by the tax code," said Blair-Stanek, a law professor at the University of Maryland. "Literally flipping a coin on these would get 50% of the questions we were asking right and GPT-3 only got about 70%. And these were just basic questions about the tax code like so-and-so is a dependent, makes $100,000 a year, does this tax section apply? It couldn't handle it."

Still, the team feels that a solution is possible. For example, initial experiments with the recently announced GPT-4 suggest that progress is being made but that there is still work to be done.

"We have yet to find a way to easily produce the judgements of a human tax professional with high accuracy, but we have been making progress," Van Durme said. "But, if we're making progress, certainly a much larger, better funded corporate effort could be making it faster."

The team worries that corporations, which have the most to gain from tax shelters, are also working on this type of artificial intelligence—and might already have it. Team members believe their AI could also be adapted for broader law uses in medicine and business.

"I'm planning to spend the rest of my career trying to make it work," Blair-Stanek said.

Source

The feature, first released in 2020, claims a 100x zoom rate, and Samsung used sparkling clear images of the Moon in its marketing.

Ibreakphotos took his own pictures of the Moon—blurry and without detail—and watched as his phone added craters and other details.

The phone's artificial intelligence software was using data from its "training" on many other pictures of the Moon to add detail where there was none.

"The Moon pictures from Samsung are fake," he wrote, leading many to wonder whether the shots people take are really theirs anymore—or if they can even be described as photographs.

Samsung has defended the technology, saying it does not "overlay" images, and pointed out that users can switch off the function.

The firm is not alone in the race to pack its smartphone cameras with AI—Google's Pixel devices and Apple's iPhone have been marketing such features since 2016.

The AI can do all the things photographers used to labor over—tweaking the lighting, blurring backgrounds, sharpening eyes—without the user ever knowing.

But it can also transform backgrounds or simply wipe away people from the image entirely.

And the debate over AI is not limited to hobbyists on message boards—professional bodies are raising the alarm too.

Sidestepping the tech

The industry is awash with AI, from cameras to software like Photoshop, said Michael Pritchard of the Royal Photographic Society of Britain.

"This automation is increasingly blurring boundaries between a photograph and a piece of artwork," he told AFP.

The nature of AI is different to previous innovations, he said, because the technology can learn and bring new elements beyond those recorded by film or sensor.

This brings opportunities but also "fundamental challenges around redefining what photography is, and how 'real' a photograph is", Pritchard said.

Nick Dunmur of the Britain-based Association of Photographers said professionals most often use "RAW" files on their digital cameras, which capture images with as little processing as possible.

But sidestepping the tech is less easy for a casual smartphone shooter.

Ibreakphotos, who posted his finding on Reddit, pointed out that technical jargon around AI is not always easy to understand—perhaps deliberately so.

"I wouldn't say that I am happy with the use of AI in cameras, but I am OK with it as long as it is communicated clearly what each processing pipeline actually does," he told AFP, asking not to use his real name.

Not 'human-authored'

What professional photographers are most concerned about, though, is the rise of AI tools that generate completely new images.

In the past year, DALL-E 2, Midjourney and Stable Diffusion have exploded in popularity thanks to their ability to create images in hundreds of styles with just a short text prompt.

"This is not human-authored work," Dunmur said, "and in many cases is based on the use of training datasets of unlicensed work."

These issues have already led to court cases in the United States and Europe.

According to Pritchard, the tools risk disrupting the work of anyone "from photographers, to models, to retouchers and art directors".

But Jos Avery, an American amateur photographer who recently tricked thousands on Instagram by filling his feed with stunning portraits he had created with Midjourney, disagreed.

He said the lines drawn between "our work" and "the tool's work" were arbitrary, pointing out that his Midjourney images often took many hours to create.

But there is broad agreement on one fundamental aspect of the debate—the risk for photography is not existential.

"AI will not be the death of photography," Avery said.

Pritchard agreed, noting that photography had endured from the daguerreotype to the digital era, and photographers had always risen to technical challenges.

That process would continue even in a world awash with AI-generated images, he said.

"The photographer will bring a deeper understanding to the resulting image even if they haven't directly photographed it," he said.

© 2023 AFP

Source

In 1956, during a year-long trip to London and in his early 20s, the mathematician and theoretical biologist Jack D Cowan visited Wilfred Taylor and his strange new "learning machine". On his arrival he was baffled by the "huge bank of apparatus" that confronted him. Cowan could only stand and watch "the machine doing its thing". The thing it appeared to be doing was performing an "associative memory scheme" – it seemed to be able to learn how to find connections and retrieve data.

It may have looked like clunky blocks of circuitry, soldered together by hand in a mass of wires and boxes, but what Cowan was witnessing was an early analogue form of a neural network – a precursor to the most advanced artificial intelligence of today, including the much discussed ChatGPT with its ability to generate written content in response to almost any command. ChatGPT's underlying technology is a neural network. (Read more about the AI emotions dreamed up by ChatGPT)

As Cowan and Taylor stood and watched the machine work, they had no idea exactly how it was managing to perform this task. The answer to Taylor's mystery machine brain can be found somewhere in its "analogue neurons", in the associations made by its machine memory and, most importantly, in the fact that its automated functioning couldn't really be fully explained. It would take decades for these systems to find their purpose and for that power to be unlocked.

The term neural network incorporates a wide range of systems, yet centrally, according to IBM, these "neural networks – also known as artificial neural networks (ANNs) or simulated neural networks (SNNs) – are a subset of machine learning and are at the heart of deep learning algorithms". Crucially, the term itself and their form and structure are "inspired by the human brain, mimicking the way that biological neurons signal to one another".

There may have been some residual doubt of their value in its initial stages, but as the years have passed AI fashions have swung firmly towards neural networks. They are now often understood to be the future of AI. They have big implications for us and for what it means to be human. We have heard echoes of these concerns recently with calls to pause new AI developments for a six-month period to ensure confidence in their implications.

It would certainly be a mistake to dismiss the neural network as being solely about glossy, eye-catching new gadgets. They are already well established in our lives. Some are powerful in their practicality. As far back as 1989, a team at AT&T Bell Laboratories used back-propagation techniques to train a system to recognise handwritten postal codes. The recent announcement by Microsoft that Bing searches will be powered by AI, making it your "copilot for the web", illustrates how the things we discover and how we understand them will increasingly be a product of this type of automation.

Drawing on vast data to find patterns, AI can similarly be trained to do things like image recognition at speed – resulting in them being incorporated into facial recognition, for instance. This ability to identify patterns has led to many other applications, such as predicting stock markets.

Neural networks are changing how we interpret and communicate too. Developed by the Google Brain Team, Google Translate is another prominent application of a neural network.

You wouldn't want to play Chess or Shogi with one either. Their grasp of rules and their recall of strategies and all recorded moves means that they are exceptionally good at games (although ChatGPT seems to struggle with Wordle. The systems that are troubling human Go players (Go is a notoriously tricky strategy board game) and Chess grandmasters, are made from neural networks.

But their reach goes far beyond these instances and continues to expand. A search of patents restricted only to mentions of the exact phrase "neural networks" produced 135,828 results at the time of writing. With this rapid and ongoing expansion, the chances of us being able to fully explain the influence of AI may become ever thinner. These are the questions I have been examining in my research and my new book on algorithmic thinking.

Mysterious layers of 'unknowability'

Looking back at the history of neural networks tells us something important about the automated decisions that define our present or those that will have a possibly more profound impact in the future. Their presence also tells us that we are likely to understand the decisions and impacts of AI even less over time. These systems are not simply black boxes, they are not just hidden bits of a system that can't be seen or understood.

There is a good chance that the greater the impact that artificial intelligence comes to have in our lives the less we will understand how or why

It is something different, something rooted in the aims and design of these systems themselves. There is a long-held pursuit of the unexplainable. The more opaque, the more authentic and advanced the system is thought to be. It is not just about the systems becoming more complex or the control of intellectual property limiting access (although these are part of it). It is instead to say that the ethos driving them has a particular and embedded interest in "unknowability". The mystery is even coded into the very form and discourse of the neural network. They come with deeply piled layers – hence the phrase deep learning – and within those depths are the even more mysterious sounding "hidden layers". The mysteries of these systems are deep below the surface.

There is a good chance that the greater the impact that artificial intelligence comes to have in our lives the less we will understand how or why.

Today there is a strong push for AI that is explainable. We want to know how it works and how it arrives at decisions and outcomes. The European Union is so concerned by the potentially "unacceptable risks" and even "dangerous" applications that it is currently advancing a new AI Act intended to set a global standard for "the development of secure, trustworthy and ethical artificial intelligence".

Those new laws will be based on a need for explainability, demanding that "for high-risk AI systems, the requirements of high-quality data, documentation and traceability, transparency, human oversight, accuracy and robustness, are strictly necessary to mitigate the risks to fundamental rights and safety posed by AI". This is not just about things like self-driving cars (although systems that ensure safety fall into the EU's category of high-risk AI), it is also a worry that systems will emerge in the future that will have implications for human rights.

This is part of wider calls for transparency in AI so that its activities can be checked, audited and assessed. Another example would be the Royal Society's policy briefing on explainable AI in which they point out that "policy debates across the world increasingly see calls for some form of AI explainability, as part of efforts to embed ethical principles into the design and deployment of AI-enabled systems".

But the story of neural networks tells us that we are likely to get further away from that objective in the future, rather than closer to it.

Inspired by the human brain

These neural networks may be complex systems, yet they have some core principles. Inspired by the human brain, they seek to copy or simulate forms of biological and human thinking. In terms of structure and design they are, as IBM also explains, comprised of "node layers, containing an input layer, one or more hidden layers, and an output layer". Within this, "each node, or artificial neuron, connects to another". Because they require inputs and information to create outputs, they "rely on training data to learn and improve their accuracy over time". These technical details matter but so too does the wish to model these systems on the complexities of the human brain.

Grasping the ambition behind these systems is vital in understanding what these technical details have come to mean in practice. In a 1993 interview, the neural network scientist Teuvo Kohonen concluded that a "self-organising" system "is my dream", operating "something like what our nervous system is doing instinctively". As an example, Kohonen pictured how a system that monitored and managed itself "could be used as a monitoring panel for any machine... in every airplane, jet plane, or every nuclear power station, or every car". This, he thought, would mean that in the future "you could see immediately what condition the system is in".

The overarching objective was to have a system capable of adapting to its surroundings. It would be instant and autonomous, operating in the style of the nervous system. That was the dream, to have systems that could handle themselves without the need for much human intervention. The complexities and unknowns of the brain, the nervous system and the real world would soon come to inform the development and design of neural networks.

Mimicking the brain – layer after layer

You may already have noticed that when discussing neural networks the image of the brain and the complexity this evokes are never far away. The human brain acted as a sort of template for these systems. In the early stages, in particular, the brain – still one of the great unknowns – became a model for how the neural network might function.

So, these experimental new systems were modelled on something whose functioning was itself largely unknown. The neurocomputing engineer Carver Mead has spoken revealingly of the conception of a "cognitive iceberg" that he had found particularly appealing. It is only the tip of the iceberg of consciousness of which we are aware and which is visible. The scale and form of the rest remains unknown below the surface.

In 1998, James Anderson, who had been working for some time on neural networks, noted that when it came to research on the brain "our major discovery seems to be an awareness that we really don't know what is going on".

A detailed account in the Financial Times in 2018, technology journalist Richard Waters noted how neural networks "are modelled on a theory about how the human brain operates, passing data through layers of artificial neurons until an identifiable pattern emerges". This creates a knock-on problem, Waters proposed, as "unlike the logic circuits employed in a traditional software program, there is no way of tracking this process to identify exactly why a computer comes up with a particular answer". Waters' conclusion is that these outcomes cannot be unpicked.

The application of this type of model of the brain, taking the data through many layers, means that the answer cannot readily be retraced. The multiple layering is a good part of the reason for this.

'Adaptation is the whole game'

Scientists like Mead and Kohonen wanted to create a system that could genuinely adapt to the world in which it found itself. It would respond to its conditions. Mead was clear that the value in neural networks was that they could facilitate this type of adaptation. At the time, and reflecting on this ambition, Mead added that producing adaptation "is the whole game". This adaptation is needed, he thought, "because of the nature of the real world", which he concluded is "too variable to do anything absolute".

As the layers of neural networks have piled higher their complexity has grown and has led to the growth of 'hidden layers' within these depths

This problem needed to be reckoned with especially as, he thought, this was something "the nervous system figured out a long time ago". Not only were these innovators working with an image of the brain and its unknowns, they were combining this with a vision of the "real world" and the uncertainties, unknowns and variability that this brings. The systems, Mead thought, needed to be able to respond and adapt to circumstances without instruction.

Around the same time in the 1990s, Stephen Grossberg – an expert in cognitive systems working across maths, psychology and bioemedical engineering – also argued that adaptation was going to be the important step in the longer term. Grossberg, as he worked away on neural network modelling, thought to himself that it is all "about how biological measurement and control systems are designed to adapt quickly and stably in real time to a rapidly fluctuating world". As we saw earlier with Kohonen's "dream" of a "self-organising" system, a notion of the "real world" becomes the context in which response and adaptation are being coded into these systems. How that real world is understood and imagined undoubtedly shapes how these systems are designed to adapt.

Hidden layers

As the layers multiplied, deep learning plumbed new depths. The neural network is trained using training data that, computer science writer Larry Hardesty explained, "is fed to the bottom layer – the input layer – and it passes through the succeeding layers, getting multiplied and added together in complex ways, until it finally arrives, radically transformed, at the output layer". The more layers, the greater the transformation and the greater the distance from input to output. The development of Graphics Processing Units (GPUs), in gaming for instance, Hardesty added, "enabled the one-layer networks of the 1960s and the two to three-layer networks of the 1980s to blossom into the 10, 15, or even 50-layer networks of today".

Neural networks are getting deeper. Indeed, it's this adding of layers, according to Hardesty, that is "what the 'deep' in 'deep learning' refers to". This matters, he proposes, because "currently, deep learning is responsible for the best-performing systems in almost every area of artificial intelligence research".

But the mystery gets deeper still. As the layers of neural networks have piled higher their complexity has grown. It has also led to the growth in what are referred to as "hidden layers" within these depths. The discussion of the optimum number of hidden layers in a neural network is ongoing. The media theorist Beatrice Fazi has written that "because of how a deep neural network operates, relying on hidden neural layers sandwiched between the first layer of neurons (the input layer) and the last layer (the output layer), deep-learning techniques are often opaque or illegible even to the programmers that originally set them up".

Although we understand how signals pass along neurons, we still don't have a clear idea of how the human brain works (Credit: Getty Images)

As the layers increase (including those hidden layers) they become even less explainable – even, as it turns out, again, to those creating them. Making a similar point, the prominent and interdisciplinary new media thinker Katherine Hayles also noted that there are limits to "how much we can know about the system, a result relevant to the 'hidden layer' in neural net and deep learning algorithms".

Pursuing the unexplainable

Taken together, these long developments are part of what the sociologist of technology Taina Bucher has called the "problematic of the unknown". Expanding his influential research on scientific knowledge into the field of AI, Harry Collins has pointed out that the objective with neural nets is that they may be produced by a human, initially at least, but "once written the program lives its own life, as it were; without huge effort, exactly how the program is working can remain mysterious". This has echoes of those long-held dreams of a self-organising system.

I'd add to this that the unknown and maybe even the unknowable have been pursued as a fundamental part of these systems from their earliest stages. There is a good chance that the greater the impact that artificial intelligence comes to have in our lives the less we will understand how or why.

But that doesn't sit well with many today. We want to know how AI works and how it arrives at the decisions and outcomes that impact us. As developments in AI continue to shape our knowledge and understanding of the world, what we discover, how we are treated, how we learn, consume and interact, this impulse to understand will grow. When it comes to explainable and transparent AI, the story of neural networks tells us that we are likely to get further away from that objective in the future, rather than closer to it.

* David Beer is professor of sociology at the University of York and is the author of The Tensions of Algorithmic Thinking: Automation, Intelligence and the Politics of Knowing

Source

ChatGPT bans

We'll start this week's digest off with news about major AI endeavors and the bumps in the road that they have been facing in the past few days. It seems that Germany may be joining Italy in attempting to ban ChatGPT from being used in the country. The reasoning appears to be the same, being that the service does not restrict children under the age of 13 from signing up and that there may be data collection violations impacting consumer privacy.

In the same vein, an Australian mayor is considering a defamation lawsuit against OpenAI because ChatGPT falsely claimed that he was jailed for taking part in a bribery scandal involving a company called Note Printing Australia in the early 2000s. And while not as large of a scope as the two cases just discussed, ChatGPT was also tricked into "generating" valid activation keys for Windows 95. Of course, you could argue that if you are giving specialized prompts to ChatGPT because you know the formula for generating the keys already, perhaps this is not as big of a deal, but let us know your thoughts in the comments section here.

Following the publication of a letter in which 1,100+ influential personalities called on AI companies to stop work on large language models more powerful than GPT-4 for at least six months, Microsoft co-founder Bill Gates has stated that this is a pointless initiative. Everyone should instead work to identify tricky areas and iron out problems with the technology. Interestingly, Microsoft also emphasized the importance of its principles for Responsible AI, which is ironic considering that it laid off its Ethical AI team just a few weeks ago.

Despite all of these ongoing debates on the subject, Microsoft is continuing its mission to integrate and improve AI technologies in its core consumer products. OneNote is getting Copilot integration soon and Microsoft has also issued multiple updates to Bing Chat to ensure fewer refusals and longer message lengths. It is also testing Bing Image Creator in the Sidebar of its Edge browser. In fact, Bing Chat has already been integrated into SwiftKey Beta.

Free USBs for Windows Insiders and other updates

You may remember that when Microsoft introduced its new Windows 11 Canary Channel last month, it offered Insiders free USBs so that they could clean install Windows 11 on their machines if they so wished. While it initially stated that it would take six to eight weeks to ship these USBs, it appears that some Insiders have begun to receive their free hardware. The USB features 64GB of storage and instructions for how to clean install Windows 11.

Speaking of the Windows Insider Program, there were a flurry of Insider builds this week. Canary build 25336 contained only a few minor improvements, Dev build 23430 brought improvements to Do Not Disturb along with lots of bug fixes and known issues, while Beta build 22624.1546 (KB5025310) added Snipping Tool support for the Prt Sc key, among lots of other improvements. A Windows Server vNext release was flighted with a known issue too, but interestingly, it does not share its build number with the Canary Channel, and is instead tagged as build 25335.

Although Microsoft documents most features and enhancements present in its Insider releases, eagle-eyed enthusiasts manage to discover undocumented changes all the time. This time around, we have seen movement on the Cloud PC integration front as well as a potentially controversial restriction to limit the number of recent tabs you see via the Alt + Tab shortcut.

Over on the marketshare side, Windows 11's position in the Steam hardware survey fell by a massive 7.92 percentage points and now stands at 22.41%. That said, it is important to understand that these results are based on optional and random surveys so may not necessarily paint an accurate picture of reality. Either way, Microsoft is trying to convince more people to upgrade to Windows 11 too as Windows 10 version 21H2 reaches its end of support date. However, if you're an AMD customer, note that there also seems to be a Windows 11 TPM Attestation bug on some supported Ryzen processors.

And if you like tinkering around with Windows, you should know that third-party Windows 11 debloating app BloatyNosy has one of its modules removed from the Microsoft Store version because it was causing conflicts with the OS. On the other hand, DoNotSpy11 is now supported on Windows 11 version 22H2 "Moment 2". Similarly, you can now enable a "Windows 12" theme using Stardock's WindowsBlinds or run the now-defunct Windows 10X on the Surface Duo with a third-party mod.

Enhancements for IT admins and upgrades for Microsoft services

Microsoft had tons of updates for its app and services for IT admins, organizations, and consumers. Starting with the first category, Windows Autopatch has new alerts and features that you can try out right now. In addition, Microsoft is changing the release cadence of Configuration Manager so that it "better aligns" with the release schedule of Windows. When it comes to Windows 365, the Redmond tech giant has rolled out a new edition for Frontline Workers and explained why it doesn't release concrete performance numbers for its Cloud PCs.

Microsoft had lots of Edge news updates sprinkled throughout the week too. The company has officially announced a preview of Workspaces for its browser and also released a utility called Browser Essentials to help you closely monitor Edge's performance and security. The browser is set to receive a major feature update with the ability to save and edit web images soon too. Until then, Microsoft has enabled some security improvements in Edge 112 and also made it easier to block autoplaying videos. Edge 113 is now available in the Dev Channel as well with lots of fixes, improvements, and policies.

Coming over to Microsoft 365 apps and services, Outlook for Windows preview has added support for Gmail accounts, with more to follow. Furthermore, following lots of user feedback, the Quick Access Toolbar default has also been returned to Office apps. Meanwhile, Teams customers will be happy to know that they will soon be able to get rid of unread notifications quickly and that Education customers have also received the upgraded Teams app with faster performance. There's also a new My Day feature and a green screen capability to leverage but the latter is just for Intel PCs.

Git gud

First-party Xbox game Quantum Break has been removed from digital storefronts due to some of its licenses expiring. However, Microsoft and Remedy have emphasized that negotiations are ongoing and that the title will return soon. Meanwhile, another first-party title Minecraft has received new icons and logos (a movie based on the IP has been pushed to April 2025) while an update to Halo: The Master Chief Collection has enabled multiplayer matchmaking on the Steam Deck. That said, fans of the latter will be sad to know that Halo chief Joseph Staten has recently left Microsoft. On the other hand, Age of Empires II enthusiasts will be pleased to hear that the Return of Rome expansion is out in May with Age of Empires I content as well.

Over on the Activision Blizzard acquisition front, Sony has criticized the UK CMA's provisional ruling in favor of Microsoft, claiming that it's full of errors. It noted that the regulator's findings are "surprising, unprecedented, and irrational" when it comes to assessing the economic impact if Microsoft were to make Activision's Call of Duty franchise an Xbox exclusive.

On the hardware side of the fence, Microsoft has announced the Xbox Wireless Controller – Remix Special Edition. The unique selling point for this item is that it features recovered plastics, with one-third coming from recycled controllers and reclaimed materials. But if you're more utility-oriented when it comes to your purchases, you might have new options for Xbox Series X|S storage expansions cards soon. In addition, if you're an Xbox Insider, check out the changelog for some new builds here too. However, if you like to tinker around with emulators on the Xbox's retail mode, you'll be sad to know that this is now disabled.

Lastly, when it comes to deals and promotions, Borderlands and Diablo II headline this week's Deals with Gold while Elder Scrolls Online and more are on offer via Xbox Free Play Days. Microsoft is also offering up to 90% off on select Xbox games during the ongoing Spring Sale. But if none of that tickles your fancy, check out this Weekend's PC Game Deals handpicked by our News Editor Pulasthi Ariyasinghe himself.

Dev Channel

• Microsoft is bringing macOS integration to Universal Print
• PowerToys 0.69 is now out with the new Registry Preview utility and a bunch of fixes
• Following a leak, the Thunderbolt 4-based Surface Dock has been announced with a price tag of $299
• Microsoft fixes audio issues on Surface Laptop 3
• OBS now blocks rogue Windows DLL files, gets improved AV1 support on AMD and Nvidia
• Microsoft and AWS could be investigated by UK authorities for anti-competitive behavior
• Ventoy has received high-capacity NTFS support

Under the spotlight

Neowin News Reporter John Callaham celebrated Microsoft's 48th birthday this week by taking a trip down memory lane and detailing some trivia items about the company's history.

With April Fools happening at the start of this month, John also took a look back at some of Microsoft's memorable stunts on this silly day, read more about them here.

John was pretty much on a spree with original content pieces and published his thoughts on an early preview that he got to witness for Minecraft Legends too.

Finally, forum member Adam Bottjen detailed how you can use a "hidden" version of the Start menu in his latest Tech Tip Tuesday piece.

Logging off

Our most interesting news item of the week relates to confirmation from Google CEO Sundar Pichai that the company is gearing up to compete with Microsoft in the field of AI-based search by integrating its large language models in Google Search. The executive also noted that Google's generative AI product Bard will switch over to the PaLM language model, which is supposed to be better in this domain than Bard's current language model.

Microsoft Weekly: ChatGPT bans, free USBs for Windows Insiders, and IT admin upgrades

OpenAI and Microsoft rely too much on artificial intelligence tools, and you can see why by looking at the most insane things ever done on ChatGPT, thanks to GPT-4. There is still a long road ahead, but after seeing its capabilities in the below examples, you will agree that the future looks bright for AI tools.

Food Recipes 

In the recent versions, ChatGPT could only understand texts and respond accordingly. OpenAI added an image-to-text feature with the latest model, and it looks like it works better than expected. You can show it several images and get accurate answers almost every time.

Twitter user, @sudu_cb, gave the photo of his fridge to GPT-4, and the chatbot came up with several food recipes in under 60 seconds. It means you can plan your meal in under 60 seconds and eliminate the hassle of looking for what to eat. It also gives you recipes, saving you more hassle from searching for "how tos" online.

Virtual Volunteer (Be My Eyes)

Be My Eyes is a Danish mobile software that attempts to make it easier for blind and visually impaired persons to recognize objects and deal with day-to-day challenges. Sighted volunteers are given images and assist visually impaired people in a live chat.

OpenAI and Be My Eyes partnered before the release of GPT-4. The Danish mobile app launched its new GPT-4-powered features on the day it was announced. Thanks to its image-to-text capability, GPT-4's virtual assistant will provide visual assistance in real-time within the app.

Hand drawn sketch to a working website

During the GPT-4 announcement, Greg Brockman showed several features that blew everyone's mind. However, one of them was extraordinarily outstanding. Brockman showed GPT-4 a hand-drawn sketch of a website, and miraculously, the chatbot turned it into a working website.

If you are building websites for a living or have plans on building your own website, you must use this feature to save time.

Dating advice

Dating apps have always been very popular since they first appeared on the market. However, people always find it hard to find the right match for themselves. According to a tweet by Jake Kozloski, co-founder of Keeper, GPT-4 will help the application match people by comparing their profile data and preferences. If the match is worth pursuing, it automates the followup.

For some people, it might sound " very automated," and you have to find the "right person" by following your heart. It still sounds like you could give it a shot, though.

Coding an entire game

ChatGPT has greatly helped developers and engineers regarding simple coding tips, showing errors, and such. Besides that, the chatbot can also code an entire game on its own Just give it the prompt and wait for it to work its magic!

Ammaar Reshi created a Snake game in less than 20 minutes without any Java knowledge. The Twitter user got the help of GPT-4 and Replit to complete the task.

Most INSANE things ever done on ChatGPT (GPT-4)

Since April 2nd, the outage has prevented users from accessing files stored on their WD NAS devices, as it required access to the company's cloud services.

The complete list of services that were down throughout this week includes My Cloud, My Cloud Home, My Cloud Home Duo, My Cloud OS 5, SanDisk ibi, and SanDisk Ixpand Wireless Charger, together with linked mobile, desktop, and web apps.

While the company said on Sunday, April 2nd, that it was working to restore connectivity and promised to provide updates the next day, a Monday statement thanked customers for their patience but failed to add any details.

Four days later, Western Digital finally shared a temporary solution that helps My Cloud Home, My Cloud Home Duo, and SanDisk ibi users access their files locally.

"The Local Access feature allows you to directly access your personal files from a Windows or MacOS computer that is connected to the same network as your device," Western Digital said.

"To enable Local Access, use your favorite browser and connect to your device's Dashboard. Then enable the Local Access feature and create a new Local Access account."

This knowledge base article provides detailed information on how to toggle on Local Access. Local access is already enabled for My Cloud OS5 (My Cloud PR series and EX series) products.

My Cloud outage follows recent security breach

This massive and ongoing My Cloud outage follows the disclosure of a network breach on Monday, April 3rd, discovered by Western Digital more than a week earlier, on March 26th.

An incident investigation involving external security and forensic experts is still in the early stages, and the company said it's also coordinating efforts with law enforcement authorities.

Based on evidence discovered so far, Western Digital believes that the attackers gained access to some of its systems and obtained data from its network.

While the company is yet to link this week's outage with the recently disclosed security breach, it did confirm on Monday that it took "systems and services offline" to secure "business operations" and said it's working to "restore impacted infrastructure and services."

BleepingComputer reached out to Western Digital several times since the outage started to ask if there's a link between the breach and the My Cloud service interruption, but we only received a reply on Friday when we were told that our message wasn't delivered.

"We are experiencing a network service outage and your message has not been delivered. When services are restored, your message will be delivered to the intended recipient. Thank you for your patience," the email reply read.

My Book Live devices wiped clean in 2021 attacks

This isn't the first time Western Digital customers lost their data, with attackers scanning for Internet-exposed and out-of-support My Book Live and My Book Live Duo NAS devices worldwide in June 2021 to factory reset and remotely wipe them clean to destroy all data.

The threat actors are believed to have targeted an unauthenticated factory reset vulnerability tracked as CVE-2018-18472.

"In some cases, the attackers have triggered a factory reset that appears to erase all data on the device," the company told BleepingComputer at the time.

While customers expressed concerns that Western Digital's servers were hacked to push out remote factory reset commands to all affected devices, the company denied the rumors and said that its network had not been breached.

"Our investigation of this incident has not uncovered any evidence that Western Digital cloud services, firmware update servers, or customer credentials were compromised," Western Digital said.

Western Digital struggles to fix massive My Cloud outage, offers workaround

• Microsoft’s partnership with OpenAI could mean billions of dollars a year in new revenue as workloads pile up in Azure.

• The investment, which most recently values OpenAI at a reported $29 billion, is complicated given the company’s “capped-profit” model

• Microsoft is integrating the technology into its Bing search engine, sales and marketing software, GitHub coding tools, Microsoft 365 productivity bundle and Azure cloud.

When Microsoft first invested $1 billion in OpenAI in 2019, the deal received no more attention than your average corporate venture round. The startup market was blazing hot, and artificial intelligence was one of many areas attracting mega-valuations, alongside electric vehicles, advanced logistics and aerospace.

Three years later, the market looks very different.

Startup funding has cratered following the collapse of public market multiples for high-growth, money-losing tech companies. The exception is artificial intelligence, specifically generative AI, which refers to technologies focused on producing automated text, visual and audio responses.

No private company is hotter than OpenAI. In November, the San Francisco-based startup introduced ChatGPT, a chatbot that went viral thanks to its ability to craft human-like replies to users’ queries about nearly any topic.

Microsoft’s once under-the-radar investment is now a major topic of discussion, both in venture circles and among public shareholders, who are trying to figure out what it means to the potential value of their stock. Microsoft’s cumulative investment in OpenAI has reportedly swelled to $13 billion and the startup’s valuation has hit roughly $29 billion.

That’s because Microsoft isn’t just opening up its fat wallet for OpenAI. It’s also the arms dealer, as the exclusive provider of computing power for OpenAI’s research, products and programming interfaces for developers. Startups and multinational companies, including Microsoft, are rushing to integrate their products with OpenAI, which means massive workloads running on Microsoft’s cloud servers.

Microsoft is integrating the technology into its Bing search engine, sales and marketing software, GitHub coding tools, Microsoft 365 productivity bundle and Azure cloud. Michael Turrin, an analyst at Wells Fargo, says it could all add up to over $30 billion in new annual revenue for Microsoft, with roughly half coming from Azure.

What does that mean for Microsoft’s investment and broader arrangement?

“It’s so good that I have investors asking me how they pulled it off, or why OpenAI would even do this,” Turrin said in an interview.

However, the financial implications are anything but straightforward.

Bragging rights

OpenAI was founded in 2015 as a nonprofit. The structure changed in 2019, when two top executives published a blog post announcing the formation of a “capped-profit” entity called OpenAI LP. The current setup restricts the startup’s first investors from making more than 100 times their money, with lower returns for later investors, such as Microsoft.

After Microsoft’s investment is paid back, it will receive a percentage of OpenAI LP’s profits up to the agreed-upon cap, with the rest flowing to the nonprofit body, an OpenAI spokesperson said. A Microsoft spokesperson declined to comment.

Greg Brockman, an OpenAI co-founder and one of the blog post’s authors, wrote in a 2019 Reddit comment that, for investors, the system “feels commensurate with what they could make investing in a pretty successful startup (but less than what they’d get investing in the most successful startups of all time!).”

It’s an unfamiliar model in Silicon Valley, where maximizing returns has long been the priority of the venture community. Nor does it make much sense to Elon Musk, who was one of OpenAI’s founders and early backers. Several times this year, Musk has tweeted his concerns about OpenAI’s unconventional structure and its implications for AI, particularly given Microsoft’s level of ownership.

″OpenAI was created as an open source (which is why I named it ‘Open’ AI), non-profit company to serve as a counterweight to Google, but now it has become a closed source, maximum-profit company effectively controlled by Microsoft,” Musk tweeted in February. “Not what I intended at all.”

Brockman said on Reddit that if OpenAI succeeds, it could “create orders of magnitude more value than any company has to date.” As a major OpenAI investor, Microsoft would benefit.

Aside from its investment, leaning on OpenAI has the potential to help Microsoft dramatically reverse its fortunes in AI, where it’s stumbled publicly and didn’t build a meaningful business on its own. Microsoft pulled the Clippy assistant from Word, Cortana from the Windows taskbar and its Tay chatbot from Twitter.

Unlike areas such as advertising or security, Microsoft hasn’t disclosed the scale of its AI business, though CEO Satya Nadella said in October that revenue from its Azure Machine Learning service had doubled for four consecutive quarters.

If nothing else, the work with OpenAI has given Nadella bragging rights. Here’s what he said at Microsoft’s annual shareholder meeting in December, a month after ChatGPT was launched:

“When I think about Azure, one of the things that we have done, in fact, in the context of even ChatGPT, which today is one of the more popular AI applications out there, guess what? It’s all trained on the Azure supercomputer.”

In February, Microsoft held a press event at its headquarters in Redmond, Washington, to announce new AI-powered updates to its Bing search engine and Edge browser. Altman was one of the featured speakers.

It’s been a bumpy ride since then, as the Bing chatbot has held some highly publicized and creepy conversations with users, and it also served up some incorrect answers at the launch. Somewhat fortunately for Microsoft, Google’s rollout of its rival Bard AI service was underwhelming, leading employees to describe it as “rushed” and “botched.”

Despite the early hiccups, the enthusiasm for new technologies based on large language models, or LLMs, is palpable across the tech industry.

At the core of OpenAI’s bot is an LLM called GPT-4 that’s learned to compose natural-sounding text after being trained on extensive online information sources. Microsoft has an exclusive license on GPT-4 and all other OpenAI models, the OpenAI spokesperson said.

There are plenty other LLMs available.

Last month, Google said it had given some developers early access to an LLM called PaLM.

Startups AI21 Labs, Aleph Alpha and Cohere offer their own LLMs, as does Google-backed Anthropic, which has picked Google as its “preferred” cloud provider. Like Altman and Musk, Anthropic cofounder Dario Amodei, who was previously vice president of research at OpenAI, has expressed concerns about the unbridled power of AI.

In 2021, Anthropic registered in Delaware as a public-benefit corporation, signifying an intention to have a positive impact on society even as it pursues profits.

“We were and are focused on developing innovative structures to provide incentives for safe development and deployment of AI systems and will have more to share on this in the future,” an Anthropic spokesperson told CNBC in an email.

Across the industry, one thing is clear: it’s early days.

Quinn Slack, CEO of code-search startup Sourcegraph, said he hasn’t seen proof that the OpenAI partnership has given Microsoft a notable advantage, even though he called OpenAI the top LLM provider.

“I don’t think people should look at Microsoft and say they’ve totally locked up OpenAI and OpenAI is doing their bidding,” Slack said. “I truly believe people there are motivated to build amazing technology and make it as widely used as possible. They view Microsoft as a great customer but not someone that’s controlling. That’s good, and I hope it stays that way.”

OpenAI has plenty of skeptics. Late last month the nonprofit Center for Artificial Intelligence and Digital Policy called on the Federal Trade Commission to stop OpenAI from releasing new commercial releases of GPT-4, describing the technology as “biased, deceptive, and a risk to privacy and public safety.”

When considering potential exits for OpenAI, Microsoft — which does not hold an OpenAI board seat — would be the natural acquirer given its close entanglement. But that sort of deal would likely attract regulatory scrutiny, because of concerns about AI and about Microsoft stifling competition. By remaining an investor and not becoming OpenAI’s owner, Microsoft could avoid Hart-Scott-Rodino reviews from U.S. competition regulators.

“I’ve gone through it. It’s painful,” said David Zilberman, a partner at Norwest Venture Partners.

Based on its existing valuation, the more probable path for OpenAI is an eventual IPO, said Scott Raney, a managing director at Redpoint Ventures.

According to PitchBook data, OpenAI is on pace to generate $200 million in revenue this year, up 150% from 2022, and then $1 billion in 2024, which would imply 400% growth.

“When you raise at a $30 billion valuation, it’s kind of like, there’s no turning back at that point,” Raney said. You’re saying, “Our plan is to be a big independent standalone company.”

OpenAI’s spokesperson said there are no plans to go public or get acquired.

Source

Dogfooding is a term used to describe a process of internally testing pre-release software, with developers using their own software in practice to discover errors and fix them prior to rolling it out to a wider audience, however, the fact that the term has been used in this context suggests that this was meant to be sent to internal Google accounts only.

Users who have received the cash got a message from Google confirming the error that had deposited the funds into their accounts, and also that the issue had been resolved and, where possible, the credit had been taken back out of the account.

This isn't the first time when a bug or issue with software has caused money to suddenly become available to members of the public, with malware also being a cause in the past.

Source: TechSpot

Google Pay gives users free money thanks to a bug

CNBC reports that Samsung has released its preliminary financial guidance for the first quarter of 2023. Among those numbers, it revealed that its profits went down to 600 billion won ($455.5 million), compared to 14.12 trillion won in profit from a year ago. That's the lowest profit number Samsung has announced in 14 years.

Those low numbers were due in part to its poor performance in its semiconductor division. Analysts believe that when Samsung reveals its final Q1 2023 numbers, its chip business will have a loss of 2.1 trillion won ($1.6 billion). Samsung will reveal its final Q1 2023 numbers later in April.

Samsung did say today that memory chip demand has fallen due to the overall weaker economic situation. The companies that buy Samsung's chips have cut their orders as they try to use up their current amount of inventory.

As a result, Samsung stated, "We are lowering the production of memory chips by a meaningful level, especially that of products with supply secured." There's no word on how much Samsung will cut its chip production. However, that decision is likely to raise memory chip prices as a whole over the coming months.

Source: Samsung via CNBC

Samsung's semiconductor division is cutting production of its chips as demand plummets

Microsoft basically has a set of Responsible AI principles that govern how AI products should be built. The foundations include ensuring fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability. The company notes that:

Implementing a Responsible AI strategy is a challenge many organizations struggle with. As a result, Microsoft has standardized the Responsible AI practices and made them available for other companies or machine learning professionals to adopt in designing, building, testing or deployment of their AI systems. For instance, customers or developers can leverage the responsible AI impact assessment template to help identify the AI systems’ intended use; data integrity, any adverse impact to people or organizations; and how it addresses goals of each of the six core responsible AI principles: Fairness, Inclusiveness, Safety & Reliability, Accountability and Transparency. In addition, this fosters a practice for AI developers to take accountability and be able to provide transparency to end-users on what the AI system does; how it should be used; its limitations/restriction and known issues. This helps machine learning teams evaluate their development lifecycle approach to validate that they are not overlooking factors that could cause their AI solution not to behave as intended.

Although Microsoft says that it has governance teams that ensure that its own AI products follow these principles too, recent news out of Redmond indicated that the firm has laid off its entire Ethics and Society team. This team was reportedly built to ensure that Microsoft's Responsible AI principles are closely tied to product design. At that time, however, Microsoft maintained that it still maintains an active Office of Responsible AI department that creates rules and principles to govern its AI initiatives and continues to make investments in the area.

The company didn't clearly say why the team was laid off either, but a former employee was quoted as saying:

People would look at the principles coming out of the office of responsible AI and say, "I don’t know how this applies". Our job was to show them and to create rules in areas where there were none.

Regardless, Microsoft has emphasized that it is investing in tools and products that ensure that Responsible AI principles are followed. A notable example is the Responsible AI dashboard it launched back in 2021. Of course, only time will tell if these initiatives are enough to curb the potential harmful effects of AI products. Recently, overall notable tech personalities like Steve Wozniak, Evan Sharp, and Elon Musk petitioned AI labs to pause development on products for at least six months and not rush unprepared into it.

Microsoft emphasizes Responsible AI principles... after laying off Ethical AI team