An estimated 3 billion batteries are thrown away in the US each year. Most of those batteries end up in landfills, where they leak their toxic materials into the environment. Now Mario Caironi at the Italian Institute of Technology and colleagues have created a safe and green alternative, demonstrating the first edible, rechargeable battery [1]. The team says that the battery, which is made of food-derived materials, could power tools used to diagnose and treat gastrointestinal tract diseases and to monitor food quality.

Creating edible electronics—fully digestible devices made of biodegradable components—is a growing enterprise. While ingestible devices, such as pills containing cameras or digital systems, are already used in healthcare, those devices currently contain inedible components that need to be retrieved after use.

One of the most significant challenges in creating a fully edible electronic system is designing a power source that is safe to eat and that has the needed power-generation properties. Currently, most ingestible technologies fail this requirement. “There is a limited selection of food-grade chemicals with the right electric properties to build a battery,” Caironi says.

One edible material that does have the right electrical properties is riboflavin (RF), a vitamin found in almonds. Another is quercetin (Q), an antioxidant plant pigment found in berries, capers, and citrus fruits. Caironi and his team used both these materials in their battery. They also used activated charcoal (AC), a digestible material that is typically made from bamboo or from coconut shell; nori algae, which is found in most sushi; beeswax; and gold foil, the same kind used to decorate baked goods.

The team’s battery has a sandwich-like structure consisting of two gold-laminated cellulose films separated by a layer of seaweed that has been soaked in salt water. The gold-cellulose films are each partially coated with AC into which either RF or Q has been mixed. This structure is wrapped in beeswax, with uncoated ends of the gold-cellulose films poking out on either side. The RF–AC and Q–AC layers of the battery act as the anode and cathode; the salt water as the electrolyte; and the gold as the current collector. The seaweed electrically isolates the anode and cathode from each other while the beeswax acts as the battery packaging.

The team tested the battery’s properties by connecting the battery to different heating elements—a standard testing method for evaluating the performance of batteries. The test showed that the RF–AC and Q–AC layers generated a stable current. The team also checked the recharging capacity of the battery, cycling it through tens of charging–discharging cycles, finding that it retained its needed performance for more than 50 cycles. They found that at 0.65 V, a voltage that can power an edible device without damaging the body, the battery sustained a current of 48 µA for 12 minutes. Finally, the team showed that two batteries connected in series were powerful enough to light a low-power LED or run a small sensor.

The demonstration shows that rechargeable batteries can be made solely from edible materials, says Conor Boland, a physicist at the University of Sussex, UK, who develops battery technologies. He notes that most of the research on developing new power sources focuses on making them smaller rather than safer or greener. “When producing new technologies, we need to think more broadly about how [those technologies] impact the planet,” which is what Caironi’s team does here, he says. Christopher Bettinger, a biomedical engineer at Carnegie Mellon University in Pennsylvania, agrees. He notes that most of the currently available edible devices contain at least some materials that have potential health risks to patients. “This work describes another step in developing application-specific materials for use in ingestible devices that could improve the safety and utility of this class of medical devices,” he says.

Source

Early robot vacuum cleaners were quite simple: they randomly bumped around a room, changing course whenever they encountered an object so that eventually their random pattern of vacuuming would cover the whole floor. Robot vacuum cleaners today are packed full of sensors and computers and are considerably cleverer.

LiDAR, also used by autonomous vehicles, is often used to detect walls and obstacles by bouncing invisible beams of light off them. Some robots use VSLAM (visual simultaneous localisation and mapping) – a way of processing images from a camera to understand the surroundings of the robot and its location.

The robots might also use ultrasonic ToF (time of flight) sensors, which work in a similar way to the echolocation of a bat, to avoid obstacles or detect if the floor is hard, soft or missing. They have accelerometers to identify motion and sensors to detect if their wheels are tangled up. Finally, suction sensors detect if their airways are blocked and the dust bucket is full.

All this information is used by AI software to determine the location of the robot and help it build a map of your rooms and furniture as it cleans. The robots are designed to clean every square inch of the floor and so they attempt to plan the most efficient route while remembering the way back to the charging station.

However, they cannot always see everything (and we may move things as they clean) so the coverage is not always perfect. Navigation becomes more difficult if their sensors become partially blocked – a robot under your sofa may be largely blind and can become quite lost!

The cleverest robot vacuum cleaners show you their maps of your home and allow you to tell them where they should concentrate and where they should avoid. The stairs are still beyond them… but Dyson filed a patent for a stair-climbing cleaning robot in 2020, so watch this space!

Source

Researchers found that by listening to the signals of eight LEO satellites for about 10 minutes, their algorithm could achieve unprecedented accuracy in locating a stationary receiver on the ground and was able to converge on it with an error of only about 5.8 meters.

The research, led by Zak Kassas, a professor of electrical and computer engineering at The Ohio State University and director of the Department of Transportation Center for Automated Vehicles Research with Multimodal AssurEd Navigation (CARMEN), was presented last week at the IEEE/ION Position Location and Navigation Symposium (PLANS) 2023 conference in Monterey, California. Along with Ohio State Ph.D. students Sharbel Kozhaya and Haitham Kanj, the paper, which demonstrated the first ever exploitation of unknown OneWeb LEO satellite signals, won the conference's Best Student Paper award.

The researchers did not need assistance from the satellite operators to use the signals, and they emphasized that they had no access to the actual data being sent through the satellites—only to publicly available information related to the satellites' downlink transmission frequency and a rough estimate of the satellites' location.

From transportation to communication systems to the power grid and emergency services, nearly every aspect of modern society relies on positioning, navigation and timing data from global navigation satellite systems (GNSS), or GPS, that orbit the Earth. Despite this, because GPS system signals are weak and susceptible to interference, they can often become unreliable in certain places such as indoor environments or in deep urban canyons. In addition, GNSS signals are spoofable, which poses serious security risks in safety-critical applications, such as aviation.

In the long term, such complications could lead to a number of navigational and cybersecurity issues, especially as virtually all of our current systems rely heavily on GPS, Kassas said. Technologies on the rise, such as autonomous vehicles, he noted, are beginning to amplify the limitations of our current GNSS systems.

"It's becoming more pressing to find civilian and military alternatives to GPS, whether as a backup or in the case when GPS isn't there whatsoever," said Kassas.

This study builds on previous research by Kassas' lab that solely used six SpaceX satellite signals to pinpoint a location within 10 meters of accuracy, which was recently reduced to 6.5 meters.

"The Starlink study scratched the surface of what is possible," said Kassas.

His work suggests utilizing signals from LEO satellites as an alternative for humans' positioning, navigation and timing needs, as they reside about 20 times closer to Earth compared to GNSS satellites, which reside in medium Earth orbit—a little more than 20,000 kilometers above the planet. According to Kassas, the technology could potentially usher in a new era of positioning, navigation and timing.

"We are witnessing a space renaissance. Tens of thousands of LEO satellites will be launched into space over the next decade, leading to what is referred to as mega-constellations," he said. "Signals transmitted by these satellites will revolutionize numerous technologies and benefit scientific inquiry in fields such as remote sensing."

What also makes the study so different from all other attempts at creating an alternative to GPS is, unlike previous studies, this algorithm doesn't reverse engineer the signal, said Kassas.

"Our algorithm is agnostic to the LEO constellation," said Kassas. "Our receiver can listen to virtually any satellite signal, trains on the data it's receiving on-the-fly, then deciphers certain features of the signal in a way where we can reconstruct what they are transmitting into location data." To demonstrate the team's new approach, the team applied the algorithm to four different LEO satellite constellations: Starlink, OneWeb, Orbcomm and Iridium. The algorithm cracked all these signals, with virtually no prior knowledge about what is being transmitted.

Additionally, their algorithm is so sophisticated that the researchers were also able to estimate where the satellites are in space. In order to use the satellite to position ourselves, we need to know where the satellite is located. "That's a very challenging problem because LEO satellites don't normally broadcast their location, and our publicly available estimates of where they are is off by a few kilometers," said Kassas.

During a stationary experiment to test how the signals worked as an accurate positioning system, researchers set a ground receiver's initial position estimate to the roof of an engineering parking structure at the University of California, Irvine, a spot more than 2,000 miles away from its actual position: the roof of Ohio State's Electroscience Laboratory (ESL) in Columbus, Ohio. Using the satellite constellations to guess where exactly in the country the receiver actually was, the algorithm was only off by about 5 meters.

In another experiment, the researchers tested how the algorithm would fare on a moving vehicle, and mounted the receiver onto the top of a car. First, they used today's navigation technology, which relies on a GPS receiver coupled with an inertial navigation system (INS). They navigated for about 100 meters before cutting GPS off, after which they drove for nearly a kilometer. They found that by relying on today's GPS-INS system, they were said to be located about 500 meters away from their true location, but with their algorithm, they were found about 4.4 meters away. "Our result showed that our system is getting close to what you can do with GPS today," said Kassas.

Although a patent has been filed on the algorithm, the team does plan to continue evolving all of the algorithm's technical abilities, said Kassas.

"GPS is a very mature system that we trust with our lives," he said. "To be able to trust new types of signals with our lives, there will have to be more studies on their accuracy, integrity and continuity."

Source

"The development of artificial intelligence could spell the end of the human race," Stephen Hawking warned.

"[AI] scares the hell out of me. It's capable of vastly more than almost anyone knows, and the rate of improvement is exponential," said OpenAI cofounder Elon Musk.

AI technologies present "profound risks to society and humanity," according to a letter signed earlier this year by more than 1,000 technology leaders urging a moratorium on AI research until more is understood about potential risks.

"We need to be very careful," said Yoshua Bengio, a professor and AI researcher at the University of Montreal.

While not disregarding the promise of tremendous good that AI will bring to a broad range of sectors in industry, economics, education, science, agriculture, medicine and research, media reports are increasingly sounding an alarm over the unintended consequences of this burgeoning disruptive technology.

One area of concern is emergent behavior, defined as a series of unanticipated, unprogrammed interactions within a system stemming from simpler programmed behaviors by individual parts.

Researchers say evidence of such behavior is seen in models that learn languages on their own, when systems trained to play chess and Go generate original strategies to advance, or when robots exhibit variability in motion patterns that were not originally programmed.

"Despite trying to expect surprises, I'm surprised at the things these models can do," said Google computer scientist Ethan Dyer, responding to an AI experiment in which a computer unexpectedly deduced on its own the title of a movie based on a string of emojis.

But Dyer himself may be surprised to learn that a research team at Stanford University is throwing cold water on reports of emergent behavior.

Ryan Schaeffer, Brando Miranda and Sanmi Koyejo said in a paper posted last week that evidence for emergent behaviors is based on statistics that likely were misinterpreted.

"Our message is that previously claimed emergent abilities … might likely be a mirage induced by researcher analyses," they said.

In their paper posted on the arXiv preprint server, the researchers explained that the abilities of large language models are measured by determining the percentage of its correct predictions.

Statistical analyses may be represented in numerous ways. The researchers contend that when results are reported in non-linear, or discontinuous, metrics, they appear to show sharp, unpredictable changes that are erroneously interpreted as indicators of emergent behavior.

However, an alternate means of measuring the identical data using linear metrics shows "smooth, continuous" changes that, contrary to the former measure, reveal predictable—non-emergent—behavior.

The Stanford team added that failure to use large enough samples also contributes to faulty conclusions.

"Existing claims of emergent abilities are creations of the researcher's analyses, not fundamental changes in model behavior on specific tasks," the team said.

They added that while methodology in past research likely yielded misleading conclusions, "nothing in this paper should be interpreted as claiming that large language models cannot display emergent abilities," suggesting proper methodology may well reveal such capacities.

"The main takeaway," the researchers said, "is for a fixed task and a fixed model family, the researcher can choose a metric to create an emergent ability or choose a metric to ablate an emergent ability."

Or as one notable commenter stated, "The output of the algorithm is only as good as the parameters which its creators set, meaning there is room for potential bias within the AI itself."

And who was that notable commentator? Microsoft Bing's ChatGPT.

Source

iMac was the first Apple product to have the "i" prefix in its name which was eventually used for an entire lineup of Apple products, including iTunes, iPod, iBook, iPhone, iPad, iCloud, and more. The meaning of "i" is widely considered as the 'internet' for which the machine was primarily built. But Apple co-founder Steve Jobs said during the announcement keynote that the prefix has other meanings such as individual, instruct, inform, and inspire.

It was later known that the name iMac was suggested to Jobs by Ken Segall who worked for the LA-based ad agency TBWA Chiat/Day. Initially, Jobs didn't like the name and wanted to go with MacMan instead. But it was later accepted after Segall pitched it to him twice.

The original iMac was also referred to as iMac G3 because the device packed a 233MHz PowerPC G3 processor inside along with an ATI Rage IIc graphics card. It came with a 4GB IDE hard drive, 32MB SDRAM, and a built-in 15-inch 1024 x 768 pixels CRT display. For connectivity, the iMac was loaded with two USB ports, an infrared port, a 24x CD-ROM drive, a built-in 33 Kbps modem, dual stereo headphone jacks, a 100Mbps Ethernet port, etc.

It also had built-in stereo speakers with Sound Retrieval System (SRS) sound and shipped with Mac OS 8.1 out-of-the-box. The original iMac was priced at $1,299 at the time and was launched a few months later in August 1998. Apple added more color options when it refreshed the iMac in 1999, including Grape, Lime, Tangerine, and Strawberry.

The original iMac was discontinued a few years later in 2003 when it was replaced by the redesigned iMac G4. Fast forward to now, the iMac is powered by Apple's in-house M-series chips and it has gotten a lot slimmer. However, the machine continues to flaunt a colorful lineup just like its forefathers.

Source

AI pioneer Geoffrey Hinton, a 75-year-old computer scientist known as “the Godfather of AI,” has made waves this week after resigning from Google to warn that AI could soon surpass humans in intelligence and learn how to destroy humanity on its own.

But Hinton’s warnings, while dire, are missing the point, says Meredith Whittaker, a prominent AI researcher who was pushed out of Google in 2019 in part for organizing employees against the company’s deal with the Pentagon to build machine vision technology for military drones. Now the president of the Signal Foundation, Whittaker tells Fast Company why Hinton’s alarmism is a distraction from more pressing threats, and how workers can stand up against technology’s harms from within. (Hinton did not respond to Fast Company’s request for comment.)

This interview has been condensed and edited for clarity.

Fast Company: Let’s start with your reaction to Geoffrey Hinton’s big media tour around leaving Google to warn about AI. What are you making of it so far?

Meredith Whittaker: It’s disappointing to see this autumn-years redemption tour from someone who didn’t really show up when people like Timnit [Gebru] and Meg [Mitchell] and others were taking real risks at a much earlier stage of their careers to try and stop some of the most dangerous impulses of the corporations that control the technologies we’re calling artificial intelligence.

So, there’s a bit of have-your-cake-and-eat-it-too: You get the glow of your penitence, but I didn’t see any solidarity or any action when there were people really trying to organize and do something about the harms that are happening now.

FC: You started organizing within Google in 2017 to oppose Project Maven, a contract the company signed to build machine vision technology for U.S. military drones. Did you anticipate being forced out for speaking up?

MW: I didn’t plan it with that in mind. But after our letter opposing Project Maven blew up, I remember realizing at that moment, based on my understanding of history, that I would be pushed out eventually. You don’t mess with the money like that and not get pushed out.

FC: Did you know Geoffrey Hinton when you were at Google?

MW: Not well, but we were in the same conferences, in the same room sometimes.

FC: And did he express any kind of support when you were organizing?

MW: I didn’t see him show up for any of the rallies or the actions or bottom-line any work. And then when it was getting dicey, when Google started taking the gloves off and hiring a union-busting firm, I didn’t see him come out and support.

But the effectiveness of raising concerns hinges on the ability of people who raise them to be safe. So, if you don’t come out when me and Meg are being fired, if you don’t come out when others are being retaliated against, when research is being suppressed, then you are tacitly endorsing an environment that punishes people for raising concerns. [Editor’s note: While Whittaker says she was forced out of her job by Google, the company claims she opted to leave.]

FC: There’s also a pattern that you’ve been pointing out: A lot of the people being punished for speaking out have been women.

MW: Women and particularly women who aren’t white. And that isn’t just in Google, but in the space where people are discussing artificial intelligence, the folks who have come out earliest and with the most grounded and materially-specific concerns have generally been women, particularly Black women and women of color. I think it’s just notable who seems easy to ignore, and whose concerns are elevated as like: “Well, finally we’re hearing it from the father, so it must be true.”

FC: On CNN recently, Hinton downplayed the concerns of Timnit Gebru—who Google fired in 2020 for refusing to withdraw a paper about AI’s harms on marginalized people—saying her ideas were not as “existentially serious” as his own. What do you make of that?

MW: I think it’s stunning that someone would say that the harms [from AI] that are happening now—which are felt most acutely by people who have been historically minoritized: Black people, women, disabled people, precarious workers, et cetera—that those harms aren’t existential.

What I hear in that is, “Those aren’t existential to me. I have millions of dollars, I am invested in many, many AI startups, and none of this affects my existence. But what could affect my existence is if a sci-fi fantasy came to life and AI were actually super intelligent, and suddenly men like me would not be the most powerful entities in the world, and that would affect my business.”

FC: So, we shouldn’t be worried that AI will come to life and wipe out humanity?

MW: I don’t think there’s any evidence that large machine learning models—that rely on huge amounts of surveillance data and the concentrated computational infrastructure that only a handful of corporations control—have the spark of consciousness.

We can still unplug the servers, the data centers can flood as the climate encroaches, we can run out of the water to cool the data centers, the surveillance pipelines can melt as the climate becomes more erratic and less hospitable.

I think we need to dig into what is happening here, which is that, when faced with a system that presents itself as a listening, eager interlocutor that’s hearing us and responding to us, that we seem to fall into a kind of trance in relation to these systems, and almost counterfactually engage in some kind of wish fulfillment: thinking that they’re human, and there’s someone there listening to us. It’s like when you’re a kid, and you’re telling ghost stories, something with a lot of emotional weight, and suddenly everybody is terrified and reacting to it. And it becomes hard to disbelieve.

FC: What you said just now—the idea that we fall into a kind of trance—what I’m hearing you say is that’s distracting us from actual threats like climate change or harms to marginalized people.

MW: Yeah, I think it’s distracting us from what’s real on the ground and much harder to solve than war-game hypotheticals about a thing that is largely kind of made up. And particularly, it’s distracting us from the fact that these are technologies controlled by a handful of corporations who will ultimately make the decisions about what technologies are made, what they do, and who they serve. And if we follow these corporations’ interests, we have a pretty good sense of who will use it, how it will be used, and where we can resist to prevent the actual harms that are occurring today and likely to occur.

FC: Geoffrey Hinton also said on CNN, “I think it’s easier to voice your concerns if you leave the company first.”

MW: If personal ease is your metric, then you’re not wrong. But I don’t think it’s more effective. This is one of the reasons that myself and so many others turn to labor organizing. Because there’s a lot of power and being able to withhold your labor collectively, and joining together as the people that ultimately make these companies function or not, and say, “We’re not going to do this.” Without people doing it, it doesn’t happen.

Source

Seeing how the adoption of such technologies can take so much time, the Alliance for Open Media (AOMedia) has already begun to add support for the succeeding AV2 codec. The support is still experimental though and quite preliminary. It looks to be in relation to the AVIF (AV1 Image File Format) for now. The pull request says (via Phoronix😞

Add AV2 support

Experimental. The AV2 specification is not finalized.

Handle av02 image item type and av2C configuration property.
Handle av02 track sample format.
Add avifavmtest.
Disable AV1 tests when all AV1 codecs are disabled.
Add presubmit ci-unix-static-av2.yml.

libavif can only generate AV2 files if it is built with AVIF_CODEC_AVM set to ON. The avifEncoder API can only output AV2 files if codecChoice is explicitly set to AVIF_CODEC_CHOICE_AVM to avoid mistakes. avifenc can only output AV2 files if --codec=avm is specified to avoid mistakes. aom and avm conflict and cannot be used together in the same libavif binary. Other AV1 codecs can be enabled alongside aom. libavif will decode any AV1 or AV2 file by default as long as the relevant codecs are enabled. The main behavior change for libavif with only AV1 codecs enabled is that before this change, items or tracks of type 'av02' were ignored. After this change, conformant AV1-AVIF files with extra 'av02' items will fail to decode unless avm is enabled. This change allows decoding files with mixed AV1 and AV2 items or tracks (untested because no encoding tool for that). This decision can be changed in the future.

In related news, Microsoft's Edge is also getting support for the image format. Competing browsers like Chrome and Firefox have had AVIF support for a while.

As AV1 is still slowly taking off, experimental support for next gen AV2 lands

In an interview, Sam Altman told CNBC that customers “clearly want us not to train on their data, so we’ve changed our plans: We will not do that.” Unfortunately for those using ChatGPT directly, this is not the case by default. The collection of data is such an issue Samsung has banned employees from using chatbots like ChatGPT over security leaks.

As an entirely new category of software, companies like OpenAI as well as wider society are still getting to grips with the best practices. Earlier today, Neowin reported on the fact that the Competition and Markets Authority was going to start investigating how these generative AI products could affect competition and consumers.

Another way in which these bots have had to retrospectively be improved upon is in relation to guard rails. Since their launch, they’ve been adapted a little bit to ensure they don’t say offensive things. When users try to get the bots to say something offensive, the bots recall pre-written scripts letting the user know they can’t help with that request.

Source: CNBC

OpenAI no longer uses API customer data to train its LLMs

LinkedIn started in Hoffman's living room back in the day but now it has a presence in over 200 countries globally with 930 million users onboard. However, this wasn't his first experience with online social platforms. In 1994, Hoffman worked on Apple's eWorld which was an online service that combined email, news, and software downloads. He also launched an online dating site called SocialNet.com in 1997.

LinkedIn's founding team included business and tech people from PayPal and SocialNet. By 2004, it had amassed 1.2 million users and the number of connected contacts had reached 50 million. Its Series B funding pitch to Greylock reveals that Linkedin had acquired a 73% market share at the time competing against other professional networks such as Ryze, OpenBC, Spoke, and ZeroDegrees.

A major milestone in LinkedIn's history came in 2016 when Microsoft acquired it for $26.2 billion. However, the Sunnyvale-headquartered company has also made numerous acquisitions of its own since 2010, including Mspoke, IndexRank, SlideShare, Lynda.com, Pulse, and most recently EduBrite.

While anyone can signup and use the platform for free, its paid tier LinkedIn Premium unlocks additional features for the users. In fact, paid products have been on the company's radar since its initial days. The company launched LinkedIn Jobs as its first premium service in 2005 which allowed hiring managers and recruiters to list positions at $95 per job.

As of June 2020, LinkedIn is headed by the current CEO Ryan Roslansky who joined the company back in 2009. He is known for playing a crucial role in the $1.5 billion acquisition of Lynda.com. More recently, the company jumped on the AI bandwagon by adding tools based on GPT-4 that can write the About section in users' profiles.

Another LinkedIn AI tool launched recently can draft messages to the hiring teams on behalf of job seekers. To boost their credibility, Linkedin also allows users to verify their work credentials. However, LinkedIn is no stranger to the wave of mass tech layoffs that started last year.

LinkedIn has turned 20 years old today: Here's a brief history of the social network

"I used to be an adventurer like you, then I took an arrow in the knee." If that sentence is forever seared into your brain through sheer repetition, you may have played Skyrim a little too often. Games can only contain so much recorded dialogue and so many dialogue options, so even in huge games like Skyrim, you get a few repeats. Well, that might not be the case for much longer if ChatGPT enthusiasts have anything to say about it. 

One modder has created a version of Skyrim where the non-playable characters (NPCs) use dialogue generated by the artificial intelligence chatbot ChatGPT. As well as allowing players the chance to talk to the in-game characters – who can respond to you via generated voices – the mod gives the characters a memory.

"I have a basic memory system set up to allow NPCs to remember past conversations with the player," mod creator Art from the machine explained on Reddit. "In-game events such as the time of day and the NPC's location are also passed to ChatGPT to give context."

A video demonstrating the mod shows the player talking to characters, some of whom are able to "view" and respond to items held by the player.

"So were you ever told campfire stories," one character by a campfire is asked. After responding "let me think" and leaving a reasonable pause, the NPC goes into a fairly uninspiring anecdote.

"Yes, my family and I used to tell campfire stories during our hunting trips. We would share tales of great Nordic Heroes and legendary creatures such as trolls," the NPC says in the video. "I remember once my father told us a story about a warrior who battled a powerful Dragon and emerged Victorious. It was a thrilling and inspirational tale that instilled courage in me and my siblings."

In another interaction, a shop owner responds to a sword shown to him by the player. It's clunky – though of course, it is just a mod designed for a bit of fun, not a polished game release. Actual games that use chatbot AIs are in the works, however.

The Skyrim mod, like other projects, has been met with enthusiasm by some – largely large language model enthusiasts – who are impressed with the concept. Others, meanwhile, are concerned for the future of gaming itself.

"If you’re happy with word soup dialogue written by a machine that was trained on stuff that was already pretty generic in the first place, no amount of me saying 'we need to value human art as the only true human experience' will convince you that if this is the future of video games that you want," games writer Luke Plunkett wrote for games website Kotaku, "you’re going to get everything you deserve".

Source

(Source: AdoredTV)

(Source: RedGamingTech)

Massive AMD Zen 5 IPC gain of 25%+ reported as leak alleges increased L1 cache and revamped L3 cache for Ryzen 8000 CPUs

The report from Bloomberg, using unnamed sources, claims Microsoft is giving financial support for AMD to enter the AI chip market. In turn, AMD is reportedly assisting Microsoft in the making of the in-house Athena AI processor, which reportedly already has several hundred people working on the chip. Neither AMD nor Microsoft have confirmed these reports.

If accurate, it would seem like AMD is trying to go after the AI market share that's currently dominated by rival NVIDIA, with processors like the H100 inside the servers that power Microsoft's Bing Chat and other AI services. Microsoft would also seem to want to make its own AI chips, so it wouldn't be dependent on NVIDIA's solutions.

Of course, Microsoft and AMD are old buddies when it comes to hardware partnerships. Custom AMD chips have been inside the Xbox One, the Xbox Series S, and the Xbox Series X consoles.

New report claims Microsoft and AMD are teaming up to make AI processors

Apple Q2 earnings are down, as everything but iPhones is harder to sell

Important new technology industries that originated in the US are now overseas. 

As a result of the manufacturing migration, the US trade balance for technology products shifted, after the year 2000, from a US surplus to a growing deficit. US dependence on foreign sources for strategic products has raised concerns both based on national security risk and a massive trade deficit. 

In response, the US government has offered more financial support for technological industries of strategic importance, for example, the CHIPS and Science Act of 2022 with US$52.7 billion pledged to support the domestic semiconductor manufacturing industry. 

It is too early to judge its effectiveness but there have been earlier programs to support strategic new industries that succeeded and failed. Their lessons should be learned. 

The biggest success was the long-term support starting in the 1970s of the semiconductor and laser industries. Then, the US Defense Department’s Defense Advanced Research Projects Agency (DARPA) was responsible, in partnership with industrial organizations, to set the foundation for the modern electronic world in chips, lasers and communications technologies. 

But there are noteworthy failures of smaller, more recent efforts to support new industries, renewable energy among them. Invented in the US, solar panels were produced by many startups in the early 2000s.

One of these new companies, Solyndra, based its product on a new thin film technology that proved to be impractical on a commercial scale. It received a federal loan of $525 million to fund the manufacture of solar panels but failed to do so and closed in 2011.

Installation of Solyndra’s cylindrical solar module design before the company’s collapse. Photo: Solyndra

There is little solar panel production in the US now. China produces most of them – an industry in the $30 billion range annually, much of it exported to the US – using an effective adaptation of silicon technology completely different from Solyndra. As a result, costs have dropped sharply and no American company can compete with overseas vendors. 

Another failure was an effort to build batteries for electric vehicles in the US. A123 was spun out of the Massachusetts Institute of Technology (MIT) in the early 2000s with a novel lithium-ion technology, entering a market for rechargeable consumer electronic batteries dominated by Asian manufacturers such as LG.

A123, on the other hand, focused on large systems, specifically on batteries for electric vehicles (EVs), which were anticipated eventually to become a big market, presuming the availability of practical and cost-competitive rechargeable batteries. 

The company entered supply agreements with companies committed to EVs including General Motors (GM), but remained focused on the materials side of battery technology while leaving system integration technology to its partners. This was a major strategic error because it impacted the EV performance of its customers.

Notably absent from its customer list was Tesla, which worked with the battery technology of established Asian battery manufacturers. Its proprietary system technology allowed it to manufacture and successfully sell early EVs. Meanwhile, an important A123 customer, Fisker, went bankrupt and the company failed to build profitable products. 

To be sure, A123 was initially a visible success. The company had an initial public offering (IPO) and raised nearly a billion dollars, including $250 million from the US government under the Federal Recovery Act of 2009. 

But the company kept losing money on the product it sold despite its investment in quality manufacturing facilities. A123 was unable to raise enough money to operate in the face of continuing losses and filed for bankruptcy in 2012. It was eventually acquired by a major Chinese automotive parts manufacturer. 

NHR 9200 battery test systems at A123 Systems R&D Laboratory. Image: Facebook

These all involved government funding supplementing private funds and are clear-cut examples of industrial policy aimed at fostering new industries.

The semiconductor industry’s development was a huge success while the other two were failures. The semiconductor program attracted the most talented people in the world. 

Its success was owed to sustaining competitive activities in industry, academia, and research laboratories with smart US federal funding, mostly managed through DARPA, to solve key problems and make the results broadly available to stimulate manufacturing. 

Capital needs were met by venture capital, public markets, and corporate funds attracted by the great opportunities emerging from the technology. All the elements were there to transform great ideas into great new products. 

The two failed programs were much more limited. The numerous startups were mostly developing clever new ways to build solar panels and worked with limited funds. What eventually succeeded in building China’s low-cost, high-performance solar panel industry was a complete rethinking of the elements and methods to reduce costs by clever vertical integration.

This required billions of dollars of investment to design and build such factories. The funding was there to do it in China. This scale of industry building was never in the program here, hence the difference. Having decided to build the solar panel industry, China’s investment program was comprehensive and ingenious. 

By contrast, the investment in Solyndra was an original approach that ended up failing with limited funding. 

China’s solar panel industry has grown by leaps and bounds. Photo: Supplied

The same criticism applies to the battery program. A123 was practically alone in trying to build a domestic US battery supplier for EVs. What strategies or tactics it applied to succeed were open to question, but there were no effective competitors to provide alternatives. Nor was there an equivalent to DARPA to steer technology development. 

My conclusion: Repeating the success of the semiconductor industry’s development with government support requires two things.

First, a compelling vision supported by the highest level of government; and, second, the ability to attract and fund the highest levels of technology and business talent combining industry, academia, and government. And don’t forget the role of DARPA in steering technology development.

Henry Kressel is a technologist, inventor, author and long-term private equity investor. 

Source

We recently covered a study by Secure Data Recovery, an HDD, SSD, and RAID data recovery company, of 2,007 defective hard disk drives it received. It found the average time before failure among those drives to be 2 years and 10 months. That seemed like a short life span, but considering the limited sample size and analysis in Secure Data Recovery's report, there was room for skepticism. Today, Backblaze, a backup and cloud storage company with a reputation for detailed HDD and SSD failure analysis, followed up Secure Data Recovery's report with its own research using a much larger data set. Among the 17,155 failed HDDs Backblaze examined, the average age at which the drives failed was 2 years and 6 months.

2 years, 6 months

Backblaze arrived at this age by examining all of its failed drives and their respective power-on hours. The company recorded each drive's failure date, model, serial number, capacity, failure, and SMART raw value. The 17,155 drives examined include 72 different models and does not include failed boot drives, drives that had no SMART raw attribute data, or drives with out-of-bounds data.

If Backblaze only looked at drives that it didn't use in its data centers anymore, there would be 3,379 drives across 35 models, and the average age of failure would be a bit longer at 2 years and 7 months.

Backblaze said its results thus far "are consistent" with Secure Data Recovery's March findings. This is despite Backblaze currently using HDDs that are older than 2 years and 7 months.

"When we first saw the Secure Data Recovery average failed age, we thought that 2 years and 10 months was too low. We were surprised by what our data told us, but a little math never hurt anyone," Backblaze's blog says.

"Given we are always adding additional failed drives to our dataset, and retiring drive models along the way, we will continue to track the average failed age of our drive models and report back if we find anything interesting."

When looking at the average age of drive failure by model, Backblaze reduced the model count to 30, eliminating drives with fewer than 50 failures.

One standout in the table above is the 12TB Seagate ST12000NM0007, which saw 2,023 failures, occuring when each drive was, on average, 1 year and 6 months old. The only model with more failures was the 4TB Seagate ST400DM000. Backblaze saw 5,249 failures, and the drives were each an average of 3 years and 3 months old upon failure.

When breaking things down by drive capacity, at first glance, it seems lower-capacity drives may last longer before failing. However, Backblaze's blog points out that Backblaze doesn't have any 1TB, 1.5TB, 2TB, 3TB, or 5TB HDDs in operation in its sample group, making the data set for the smaller capacity drives complete. In contrast, most of the larger-capacity HDDs are still being used.

"In other words, as these larger drives continue to fail over the coming months and years, they could increase or decrease the average failure age of that drive model," Klein explained.

Failure rates since April 2013

As usual, Backblaze provided a quarterly update of the HDDs in its active arsenal. Today, it added data from Q1 2023.

The following table shows annualized failure rates (AFRs) for 236,893 HDDs across 30 models and through nearly 10 years. The chart doesn't include drives "only used for testing purposes" or of which Backblaze had fewer than 60 units.

Overall, the average AFR was 1.4 percent. The drive with the lowest AFR (0.28 percent) and at least 2.2 million drive days is Western Digital Corporation (WDC)'s 16TB WUH721816ALE6L4 (Backblaze has 14,098 units).

The model with the highest AFR (2.57 percent) and at least 2.2 million drive days is Seagate's 4TB ST4000DM000 (Backblaze had 18,070 units). It's worth noting, though, that the model has the largest number of drive days out of any model on Backblaze's table. In February, Backblaze said Seagate HDDs generally show higher failure rates in Backblaze's environment and are also "less expensive," so "their failure rates are typically not high enough to make them less cost-effective over their lifetime. You could make a good case that for us, many Seagate drive models are just as cost-effective as more expensive drives."

Backblaze's full data set is on its Hard Drive Test Data page.

Source

Today, US Vice President Kamala Harris will be meeting with the CEOs of four of the leading companies in AI development: Microsoft's Satya Nadella, Google's Sundar Pichai, OpenAI's Sam Altman, and Anthropic's Dario Amodei. The White House says the participants will discuss the growing use of AI and "the importance of driving responsible, trustworthy, and ethical innovation with safeguards that mitigate risks and potential harms to individuals and our society."

The Biden Administration also announced that those four companies, along with Hugging Face, NVIDIA, and Stability AI, have all agreed to have their AI systems tested at the annual DEFCON hacking conference. DEFCON 31 is scheduled to be held in Las Vegas from August 10-13. The White House press release stated:

This independent exercise will provide critical information to researchers and the public about the impacts of these models, and will enable AI companies and developers take steps to fix issues found in those models. Testing of AI models independent of government or the companies that have developed them is an important component in their effective evaluation.

 

In addition, the US Office of Management and Budget will create a draft document for a policy on how the US government should use AI, and will offer it for public comment. Also, the National Science Foundation will add $140 million for funding of seven new National AI Research Institutes, which will bring the number of those groups to 25.

Source

According to the announcement, the government is requesting commitments from top AI companies to take part in a "public evaluation" of their AI systems at DEFCON 31, demanding that the Office of Management and Budget (OMB) draft policy guidance for federal employees, and allocating $140 million to the establishment of seven new AI R&D centers as part of the National Science Foundation.

"At a time of rapid innovation, it is essential that we make clear the values we must advance, and the common sense we must protect. With [Thursday's announcement] and the blueprint for an AI bill of rights, we've given company and policymakers and the individuals building these technologies, some clear ways that they can mitigate the risks [to consumers]," the administration official said.

More than a dozen top AI firms, including Anthropic, Google, Hugging Face, Microsoft, NVIDIA, OpenAI, and Stability AI, also confirmed their "independent commitment" to submit their AI systems for review by the general public at DEFCON 31 (August 10–13).

US Vice President Kamala Harris Cheney Orr | Afp | Getty Images

Countries take precautions against AI and cyber attacks

Recently, France and Singapore joined forces to run a research and development laboratory in which they will develop AI capabilities that can be used in cybersecurity solutions. The laboratory will be in Singapore. Researchers from Singapore and France will collaborate to carry out research that might one day be useful to both nations.

With 32,000 employees and 1,000 labs, CNRS is the second-largest scientific research institution in the world. As part of the SAFARI (Singapore and France Advanced Research Initiative) collaboration, they will collaborate closely with experts from Singapore.

The Biden administration also announced a new cyber plan, assigning more responsibilities to tech firms to prevent cyberattacks and such. Biden urged American computer firms to take extra precautions against cyberattacks and hackers and to be accountable for maintaining the security of their networks. rather than taking immediate legal action, the plan established goals.

Source

Microsoft stated:

For example, if you’re researching the latest restaurant for dinner in Bing chat, it will leverage OpenTable to help you find and book a reservation. Or, with Wolfram|Alpha, you can create powerful visualizations and get answers to complex science, math and human-curated data-based questions directly from Bing chat.

The company said it will be working with OpenAI, the team who developed GPT-4 which is the engine at the heart of Bing Chat, to help add plug-in features. OpenAI has already added that same plugin support to ChatGPT.

While there are few details about this kind of support, Microsoft says more will be revealed soon at its Build 2023 developer conference. It's being held in Seattle from May 23-24, with most of its sessions scheduled to be streamed live. We expect even more info about Bing Chat to be unveiled at the conference.

Microsoft confirms Bing Chat third-party plug-ins are coming, with more info at Build 2023

While Microsoft seemed to have eliminated the Bing Chat waitlist a couple of months ago, the company has now officially stated that there is indeed no more waitlist to try out the chatbot. All you need is a Microsoft Account and you are all set to use Bing Chat on mobile or via Edge on the desktop.

Microsoft claims that there are now over 100 million daily active users for Bing Chat, with over half a billion total chats since its launch. It also says that daily installs of the Bing mobile app have gone up four times more than normal since Bing Chat was first introduced.

One of the most requested features by Bing Chat users is a way to save previous chats. Today, Microsoft said that ability will be put in "shortly", allowing users to stop a chat in mid-session and then return again where the chat was paused. It added:

And when you want to dig into something deeper and open a Bing chat result, your chat will move to your Edge sidebar, so you can keep your chat on hand while you browse. Over time, we’re exploring making your chats more personalized by bringing context from a previous chat into new conversations.

Microsoft will also be adding ways to quickly export chat conversations and also share them on social media. It stated:

For times when you want to easily share your conversation with others in social media or continue iterating on a newly discovered idea, you can export it directly – the format stays the same to make an easy transition to continue in collaborative tools like Microsoft Word.

In March, Microsoft launched Bing Image Creator, a way to use text prompts to create art with generative AI. Since then the company claims that over 200 million images have been generated with the AI program. Today, the company said that Bing Image Creator is expanding its reach.

Previously it only supported English, but now the AI art maker can now accept text prompts from all of Bing's over 100 supported languages. The company added it is working to add visual searches to Bing Chat, so you can upload images in the chat to search for content related to it in Bing. It will also add visual elements for searches in chat, including things like charts and graphs.

Microsoft also revealed some changes coming to its Edge browser in relationship to Bing:

Edge mobile will also soon include page context, so you can ask questions in Bing chat related to the mobile page you're viewing. The compose feature in sidebar can also now tailor drafts based on feedback you give like tone, length, phrasing and more.

Bing Chat will also offer improvements for summaries of large articles and documents within Edge, and will introduce what Microsoft calls Edge actions:

Available in the coming weeks, people will soon be able to lean on AI to complete even more tasks with fewer steps. For example, if you want to watch a particular movie, actions in Edge will find and show you options in chat in the sidebar and then play the movie you want from where it’s available.

Microsoft adds that it will continue to monitor the use of Bing Chat to make sure that its safeguards are working for the ethical use of its AI tools. It stated:

Together with our partners at OpenAI, we’ve continued to implement safeguards to defend against harmful content based on what we’re learning and seeing in preview. Our teams continue to work to address issues such as misinformation and disinformation, content blocking, data safety and preventing the promotion of harmful or discriminatory content in line with our AI principles.

The new info on Bing Chat updates comes on the same day that the CEOs of Microsoft, Google, OpenAI, and Anthropic will be meeting with US Vice President Kamala Harris in Washington DC to discuss plans for keeping the use of AI ethical and responsible.

Microsoft Bing Chat is in open preview, and will add chat history, visual search and more

The Xbox Insider Alpha Ring update has the build number 2306.230501-2200. Here is the changelog:

Fixes Implemented

Thanks to the hard work of Xbox engineers, we are happy to announce the following fixes have been implemented for this build:

Games

• Fixes to address an issue where some titles would fail to launch unexpectedly.

My Games & Apps

• Additional fixes to address an issue where installing or updating titles could remain in the Queue ‘finishing things up’ longer than expected.
• Fixed an issue where titles could fail to install unexpectedly.

System

• Fixes to address an issue where the console could shut down unexpectedly when using Report a Problem.
• Various updates to properly reflect local languages across the console.
  • Note: Users participating in Preview may see “odd” text across the console, for more information go here.

The Xbox Insider Alpha Skip-Ahead ring update has the build number 2308.230430-2200. Here is the changelog:

Fixes Implemented

Thanks to the hard work of Xbox engineers, we are happy to announce the following fixes have been implemented for this build:

My Games & Apps

• Fixed an issue where Remote Tools for Microsoft Edge would repeatedly prompt for an update.
• Additional fixes to address an issue where installing or updating titles could remain in the Queue ‘finishing things up’ longer than expected.

System

• Various updates to properly reflect local languages across the console.
  • Note: Users participating in Preview may see “odd” text across the console, for more information go here.

Members of the Alpha and Alpha Skip-Ahead rings are also currently testing the new Xbox Home Experience as well.

Microsoft Xbox Insider Alpha and Alpha Skip-Ahead rings get new bug fixing updates

Quick Games to Play and Longest Games collections added to your Home screen:

• When browsing your Home Game Pass screen, you can scroll down to view various collection channels, including these new collections that list average gameplay times to fit your gaming preferences.

Accessibility filters for All PC Games:

• Filtering the All PC Games list just got better by adding a wide variety of Accessibility filters to help find the right game for your needs.

Bug Fixes

• Fixed a bug where no games appeared in the All PC Games section.
• Fixed a bug where some text was cut off when directed to install the Riot or Ubisoft game launcher.
• Fixed a bug where some cloud games were not showing in the Cloud Gaming section.
• Fixed a bug where installation errors did not provide additional details and troubleshooting suggestions.

You can check out the full blog post here. Just a reminder: the Xbox app for Windows lets users download and play PC games with the PC Game Pass, along with the ability to play Xbox console games in the cloud via Xbox Game Pass Ultimate. It also lets users chat with their Xbox friends on PC, console, and mobile platforms.

Microsoft Xbox app for Windows update has Quick Games to Play and Longest Games collections

Starlink nixes plan to impose 1TB data cap and per-gigabyte overage fees

< Watch the video here. >

One day recently, on a table in Jean Oh’s lab in the Squirrel Hill neighborhood of Pittsburgh, a robot arm was busy at a canvas. Slowly, as if the air were viscous, it dipped a brush into a pool of light gray paint on a palette, swung around and stroked the canvas, leaving an inch-long mark amid a cluster of other brushstrokes. Then it pulled back and paused, as if to assess its work.

The strokes, mostly different shades of gray, suggested something abstract — an anthill, maybe. Dr. Oh, the head of the roBot Intelligence Group at Carnegie Mellon University, dressed in a sweatshirt bearing the words “There Are Artists Among Us,” looked on with approval. Her doctoral student, Peter Schaldenbrand, stood alongside.

Dr. Oh’s work, which includes robot vision and topics in autonomous aviation, often touches on what is known as the sim-to-real gap: how machines trained in a simulated environment can act in the real world. In recent years, Mr. Schaldenbrand has led an effort to bridge the sim-to-real gap between sophisticated image-generation programs like Stable Diffusion and physical works of art like drawings and paintings. This has mainly been manifest in the project known as FRIDA, the latest iteration of which was rhythmically whirring away in a corner of the lab. (FRIDA is an acronym for Framework and Robotics Initiative for Developing Arts, although the researchers chose the acronym, inspired by Frida Kahlo, before deciding what it stood for.)

The process of moving from language prompts to pixelated images to brushstrokes can be complicated, as the robot must account for “the noise of the real world,” Dr. Oh said. But she, Mr. Schaldenbrand and Jim McCann, a roboticist at Carnegie Mellon who also helped develop FRIDA, believe that the research is worth pursuing for two reasons: It could improve the interface between humans and machines, and it could, through art, help connect people to one another.

“These models are trained based on everybody’s data,” Dr. McCann said, referring to the large language models that power tools like ChatGPT and DALL-E. “And so I still think we’re figuring out how projects like this, that use such models, can deliver value back to people.”

Jean Oh, who leads the roBot Intelligence Group at Carnegie Mellon University; Jim McCann, a roboticist who helped develop FRIDA; Peter Schaldenbrand, P.D. candidate from the School of Computer Science.Credit...Kristian Thacker for The New York Times

The sim-to-real gap provides a surprisingly tricky problem for roboticists and computer engineers. Some artificial intelligence systems can list the steps involved in walking (tighten your quadriceps and flex your tibialas posterior, tilt your weight back and tense your gluteus maximus) and can make a simulated body walk in a virtual world. So it’s tempting to think that these systems could easily make a physical body walk in the real world.

Not so. In the 1980s, the computer scientist Hans Moravec noted that A.I. was good at engaging in complicated reasoning and parsing vast amounts of data but that it was bad at simple physical activities, like picking up a bottle of water. This is known as Moravec’s paradox. (The physical superiority of humans might be explained by our body’s long evolutionary history; the tasks that are simple for us are supported by millions of years of Darwinian experimentation.)

Painting, which often mixes high-concept ideas and basic physical actions, throws the paradox into relief: How do we manage to capture the absurdity of human consciousness with the motions of an arm?

A.I. image-generating tools like Midjourney, DALL-E and Stable Diffusion are trained by feeding neural networks massive databases of images and corresponding text descriptions. The programmed goal is to model the relationships between the meanings of words and the features of images, and to then use these relationships in a “diffusion model” to create original images that retain the meaning of particular descriptions. (The prompt “A family picnicking in the park” will generate a new image every time it is used; each one will be understandable as a family picnicking in the park.)

< Watch the video here. >

Dr. Oh emphasized that humans were still essentially involved in FRIDA’s painting, prompting the machine, mixing the paints, setting up the canvas and limit the number of total brushstrokes in each piece.CreditCredit...Kristian Thacker for The New York Times

But such images exist only in the sim-world of computers, composed of pixels of varying hue and intensity. Leave the simulation and the image stays behind.

To solve this problem, Dr. Oh and her colleagues took FRIDA’s physicality into account. Taped to a wall in their lab is a piece of paper with 130 different brushstrokes in black: curlicues and lines, some long and straight, some little more than dots. The marks represent the range of the robot’s motion, and they were programmed into its diffusion model.

“We take pictures of the brushstrokes, model that interaction, and then get a really accurate simulation of brushstrokes grounded in what the robot can actually do,” Mr. Schaldenbrand said. When prompted, the model would create an image of a frog ballerina in pixels, but only in configurations that were possible for the robot to paint using those 130 brushstrokes.

The researchers also developed a way for the robot to occasionally step back from its painting, to gauge how close it was to the goal it had generated in pixels, and to then revise that pixelated goal. A wayward mark could become the motion of the frog ballerina leaping, or the raised eyebrow of someone in her audience. So, every few dozen brushstrokes, FRIDA pulled away from the canvas, took a photo of its work thus far, paused, and then went back to work.

“It’s how maybe human artists do this,” Dr. Oh said. “Add some brushstrokes, and then go back and look at the full canvas and replan. We wanted to mimic that process.” A process of artistic self-discovery, in a way, albeit a mechanized, algorithmic and statistical one.

The results of these methods are on display in the lab. Portraits of professors, historical figures, landscapes, cityscapes, that frog ballerina, all in a distinctive, abstract style — even a self-portrait of the first FRIDA robot. The consistency of the paintings suggests a unified artistic vision, for which Mr. Schaldenbrand, Dr. McCann and Dr. Oh decline to claim credit. They attribute each of the works to FRIDA.

But could FRIDA have an oeuvre without a will, a heart or fingernails? Can a robot be an artist?

Amy LaViers, a computer scientist and dancer who runs the Robotics, Automation and Dance Lab, an independent nonprofit, said that such questions wouldn’t seem so crazy, or scary, if people were open to dissolving the hard distinction between the artist and the medium. Everything — whether it’s watercolor or A.I. image generators or a desire for expressivity — is wrapped up in the art. Even something as simple as paint can seem to have a mind of its own, and a painter has to react to the way it glides on the canvas. Dr. LaViers suggested viewing FRIDA as a “robotic paintbrush,” rather than a painting robot.

“There are things you can do with artificial bodies that humans can’t do,” she said. “It broadens the palette of human expression.”

Dr. Oh emphasized that humans were still essentially involved in FRIDA’s painting. They prompt the machine and mix the paints, set up the canvas and limit the number of total brushstrokes in each piece. The data sets that FRIDA and other image generators are trained on contain paintings and photographs created by other people. But, Dr. Oh added, the goal was never to make something to compete with human artists. “We want to promote human creativity,” she said. “We want people to express their thoughts in different ways.”

In the lab, Mr. Schaldenbrand watched as a painting slowly emerged from FRIDA’s deliberate gray brushstrokes: a foggy road, the shapes of cars, taillights. “This is hard to explain,” he said. “I don’t want to give some false notion that there’s a consciousness going on here. But it’s kind of fun sometimes to pretend.”

Source

In October 2021, Google promised to stop placing ads alongside content that denied the existence and causes of climate change, so that purveyors of the false claims could no longer make money on its platforms, including YouTube.

And yet if you recently clicked on a YouTube video titled “who is Leonardo DiCaprio,” you might have found a ramble of claims that climate change is a hoax and the world is cooling after a Paramount+ ad for the film “80 for Brady,” starring Lily Tomlin, Jane Fonda, Sally Field and Rita Moreno.

Before another video that purported to detail “how climate activists distort the evidence,” some users saw an ad for Alaska Airlines.

These are not aberrations, according to a coalition of environmental organizations and the Center for Countering Digital Hate. In a report released on Tuesday, researchers from the organizations accused YouTube of continuing to profit from videos that portrayed the changing climate as a hoax or exaggeration.

They found 100 videos, viewed at least 18 million times in total, that violated Google’s own policy. They found videos accompanied by ads for other major brands like Adobe, Costco, Calvin Klein, and Politico. Even an ad for Google’s search engine popped up before a video that claimed there was no scientific consensus about the changing climate.

“It really begs the question about what Google’s current level of enforcement is,” Callum Hood, the head of research at the Center for Countering Digital Hate, said in an interview.

It is difficult to assess the full extent of misinformation on YouTube, researchers said, because watching videos is time-intensive work and they have limited data access, leaving them to rely on laboriously searching the platform with keywords. “I think it’s fair to say it’s probably the tip of the iceberg,” Mr. Hood added, referring to what they had found.

Ms. Fonda, who runs a political action committee devoted to fighting climate change, said in a statement that it was “abhorrent that YouTube would violate its own policy” by running climate hoax videos with ads, giving the content further validity while “the earth is burning.”

“I am appalled that an ad for one of my movies appears on one of those videos, and hope YouTube stops this practice immediately,” Ms. Fonda said.

Ads for Grubhub, the food-delivery service, appeared before climate-denial videos numerous times, The New York Times found. A Grubhub spokeswoman said that the company was working with YouTube and other partners to “prevent Grubhub ads from appearing alongside content that promotes misinformation.”

Michael Aciman, a YouTube spokesman, said in a statement that the company allowed “policy debate or discussions of climate-related initiatives, but when content crosses the line to climate change denial, we remove ads from serving on those videos.”

He added that YouTube removed ads from several videos that the researchers flagged, including the one that promoted “80 for Brady.”

As misinformation has grown into a greater scourge online, YouTube has tried to balance its desire to be an open platform for diverse views with an interest in serving users proven facts on important topics. In recent years, the platform clamped down on the lie that the 2020 presidential election was stolen and false claims about vaccines.

In 2021, when the company changed its rules on climate change, it said that advertisers and publishing partners had grown increasingly uncomfortable appearing alongside inaccurate climate content.

Google’s policy applies to content that refers to climate change as a hoax or a scam, denies the long-term trend that the climate is warming, or denies that greenhouse gas emissions or human activity are contributing to climate change.

Under some of the climate videos that researchers found — some with ads and some without — YouTube had a “context” box with authoritative information, signaling that it knew the videos featured false or at least disputed claims. “Climate change refers to long-term shifts in temperatures and weather patterns, mainly caused by human activities, especially the burning of fossil fuels,” YouTube wrote, while linking to a United Nations site on the topic.

The research by the Center for Countering Digital Hate and Climate Action Against Disinformation, an international coalition of more than 50 environmental advocacy groups, suggested that YouTube had overlooked or ignored violative content. They identified another 100 videos that did not explicitly violate Google’s policies, but met a broader definition of climate disinformation that should also be covered.

“This demonstrates that YouTube is currently profiting from a much broader range of climate disinformation than is covered by its narrowly drawn policies,” the report said.

The videos the group cited come from a variety of sources, including pundits, podcasters and advocacy groups.

They also included industry giants like Exxon Mobil, which has been accused of “greenwashing” its contribution to carbon emissions, though its videos did not explicitly violate YouTube’s policies; and mainstream conservative media like Fox News, whose videos sometimes did. (In one, Fox’s recently fired anchor Tucker Carlson dismissed the fight against climate change as “a coordinated effort by the government of China to hobble the U.S. and the West and take its place as the leader of the world.”)

Exxon Mobil and Fox did not immediately respond to a request for comment

Almost all the videos featured ads, the researchers found, which meant YouTube was generating revenue from the content and, in some instances, may have paid creators for the videos. The placement of ads is an automated process. The platform’s videos are often targeted for particular viewers, meaning that different users will see different commercials before the same video plays.

Creators can receive compensation from YouTube as members of the company’s partner program, after they accumulate 1,000 subscribers and users watch 4,000 hours of their videos. It was unclear how many videos featuring climate misinformation were from creators in the program.

“What makes YouTube especially dangerous is that they profit share per video,” said Claire Atkin, a co-founder of Check My Ads, an advocacy group that studies advertising online and was not involved in the research. “When someone posts this information to Facebook, they don’t make money, but when someone posts a video to YouTube, they have the opportunity to make a full salary on disinformation.”

She said that YouTube was a powerful force for radicalizing people online and needed to work harder to govern content on its platform. “The fact that they haven’t changed that, that they are still funding — not promoting, funding — by sending advertisers to sponsor climate change disinformation is yet another proof point of their ineptitude.”

Source

It seems as if people are easily able to articulate a whole host of serious social, political and cultural problems that are likely to arise from the widespread use of GPT. But it seems much less easy for people to articulate specific potential benefits on the same scale. Should that be a huge red flag? The question is: Do the benefits outweigh the costs? The intellectually honest answer is that we don’t know. Some of us would like to slow this down because we are seeing more costs every day, but I don’t think that means that there are no benefits. We know it’s useful for computer programmers. A lot of this discussion around the so-called Pause Letter is a fear that if we don’t build GPT-5, and China builds it first, somehow something magical’s going to happen; 5 is going to become an artificial general intelligence that can do anything. We may someday have a technology that revolutionizes science and technology, but I don’t think GPT-5 is the ticket for that. GPT-4 is pitched as this universal problem solver and can’t even play a decent game of chess! To scale that up in your mind to think that GPT-5 is going to go from “can’t even play chess” to “if China gets it first, the United States is going to explode” — this is fantasyland. But yeah, I’m sure GPT-5 will have some nice use cases. The biggest use case is still writing dodgy prose for search engines.

Do you think the public has been too credulous about ChatGPT? It’s not just the public. Some of your friends at your newspaper have been a bit credulous. In my book, “Rebooting A.I.,” we talked about the Eliza effect — we called it the “gullibility gap.” In the mid-1960s, Joseph Weizenbaum wrote this primitive piece of software called Eliza, and some people started spilling their guts to it. It was set up as a psychotherapist, and it was doing keyword matching. It didn’t know what it was talking about, but it wrote text, and people didn’t understand that a machine could write text and not know what it was talking about. The same thing is happening right now. It is very easy for human beings to attribute awareness to things that don’t have it. The cleverest thing that OpenAI did was to have GPT type its answers out one character at a time — made it look like a person was doing it. That adds to the illusion. It is sucking people in and making them believe that there’s a there there that isn’t there. That’s dangerous. We saw the Jonathan Turley incident, when it made up sexual harassment charges. You have to remember, these systems don’t understand what they’re reading. They’re collecting statistics about the relations between words. If everybody looked at these systems and said, “It’s kind of a neat party trick, but haha, it’s not real,” it wouldn’t be so disconcerting. But people believe it because it’s a search engine. It’s from Microsoft. We trust Microsoft. Combine that human overattribution with the reality that these systems don’t know what they’re talking about and are error-prone, and you have a problem.

So have Sam Altman or Satya Nadella been irresponsible in not speaking more clearly about the actual capabilities — or lack thereof — of their companies’ technologies? Sam has walked both sides of that fence — at times, I think, inviting the inference that these things are artificial general intelligence. The most egregious example of that in my mind is when DALL-E 2 came out. He posted pictures and a Tweet saying, “A.G.I. is gonna be wild.” That is inviting the inference that these things are artificial general intelligence, and they are not! He subsequently backed down from that. Also, around that time, he attacked me. He said, “Give me the confidence of a mediocre deep-learning skeptic.” It was clearly an attack on me. But by December, he started to, I think, realize that he was overselling the stuff. He had a Tweet saying these things have trouble with truth. That’s what I was telling you back when you were making fun of me! So he has played both sides of it and continues to play both sides. They put out this statement about dealing with A.G.I. risk, inviting the inference that what they have has something to do with artificial general intelligence. I think it’s misleading. And Nadella is certainly not going around being particularly clear about the gap between people’s expectations and the reality of the systems.

And when Sam Altman said that ChatGPT needs to be out there being used by the public so that we can learn what the technology doesn’t do well and how it can be misused while “the stakes are low” — to you that argument didn’t hold water? Are the stakes still low if 100 million people have it and bad actors can download their own new trained models from the dark web? We see a real risk here. Every day on Twitter I get people like: “What’s the risk? We have roads. People die on roads.” But we can’t act like the stakes are low. I mean, in other domains, people have said, “Yeah, this is scary, and we should think more about it.” Germ-line genome editing is something that people have paused on from time to time and said, “Let’s try to understand the risks.” There’s no logical requirement that we simply march forward if something is risky. There’s a lot of money involved, and that’s pushing people in a particular direction, but I don’t think we should be fatalistic and say, “Let’s let it rip and see what happens.”

What do you think the 2024 presidential election looks like in a world of A.I.-generated misinformation and deepfakes? A [expletive] show. A train wreck. You probably saw the Trump arrest photos. And The Guardian had a piece about what their policy is going to be as people make fake Guardian articles, because they know this is going to happen. People are going to make fake New York Times articles, fake CBS News videos. We had already seen hints of that, but the tools have gotten better. So we’re going to see a lot more of it — also because the cost of misinformation is going to zero.

You can imagine candidates’ dismissing factual reporting that is troublesome to them as being A.I. fakery. Yeah, if we don’t do something, the default is that by the time the election comes around in 2024, nobody’s going to believe anything, and anything they don’t want to believe they’re going to reject as being A.I.-generated. Aand the problems we have around civil discourse and polarization are just going to get worse.

So what do we do? We’re going to need watermarking for video. For text, it’s going to be really hard; it’s hard to make machines that can detect the difference between something generated by a person and something generated by a machine, but we should try to watermark as best we can and track provenance. That’s one. No. 2 is we’re going to have to have laws that are going to make a lot of people uncomfortable because they sound like they’re in conflict with our First Amendment — and maybe they are. But I think we’re going to have to penalize people for mass-scale harmful misinformation. I don’t think we should go after an individual who posts a silly story on Facebook that wasn’t true. But if you have troll farms and they put out a hundred million fake pieces of news in one day about vaccines — I think that should be penalizable. We don’t really have laws around that, and we need to in the way that we developed laws around spam and telemarketing. We don’t have rules on a single call, but we have rules on telemarketing at scale. We need rules on distributing misinformation at scale.

You have A.I. companies, right? I had one. I sold it to Uber. Then the second one is called RobustAI. It’s a robotics company. I co-founded it, but I’m not there any longer.

OK, so knowing all we know about the dangers of A.I., what for you is driving the “why” of developing it? Why do it at all? Rather than lobby to shut it down?

Yeah, because the potential harms feel so profound, and all the positive applications I ever hear about basically have to do with increased efficiency. Efficiency, to me, isn’t higher on the list of things to pursue than human flourishing. So what is the “why” for you? Since I was 8 years old, I’ve been interested in questions about how the mind works and how computers work. From the pure, academic intellectual perspective, there are few questions in the world more interesting than: How does the human child manage to take in input, understand how the world works, understand a language, when it’s so hard for people who’ve spent billions of dollars working on this problem to build a machine that does the same? That’s one side of it. The other side is, I do think that artificial general intelligence has enormous upside. Imagine a human scientist but a lot faster — solving problems in molecular biology, material science, neuroscience, actually figuring out how the brain works. A.I. could help us with that. There are a lot of applications for a system that could do scientific, causal reasoning at scale, that might actually make the world of abundance that Peter Diamandis imagines. I don’t think, however, that the technology we have right now is very good for that — systems that can’t even reliably do math problems. Those kinds of systems are not going to reinvent material science and save the climate. But I feel that we are moving into a regime where, exactly, the biggest benefit is efficiency: I don’t have to type as much; I can be more productive. These tools might give us tremendous productivity benefits but also destroy the fabric of society. If that’s the case, that’s not worth it. I feel that the last few months have been a wake-up call about how irresponsible the companies that own this stuff can be. Microsoft had Tay. They released it, and it was so bad that they took it down after 24 hours. I thought, oh, Microsoft has learned its lesson. Now Microsoft is racing, and Nadella is saying he wants to make Sundar Pichai dance! That’s not how we should be thinking about a technology that could radically alter the world.

Presumably an international body governing A.I. would help guide that thinking? What we need is something global, neutral, nonprofit, with governments and companies all part of it. We need to have coordinated efforts around building rules. Like, what happens when you have chatbots that lie a lot? Is that allowed? Who’s responsible? If misinformation spreads broadly, what are we going to do about that? Who’s going to bear the costs? Do we want the companies to put money into building tools to detect the misinformation that they’re causing? What happens if these systems perpetuate bias and keep underrepresented people from getting jobs? It’s not even in the interest of the tech companies to have different policies everywhere. It is in their interest to have a coordinated and global response.

Maybe I’m overly skeptical, but look at something like the Paris climate accord: The science is clear, we know the risks, and yet countries are falling well short of meeting their goals. So why would global action on A.I. be feasible? I’m not sure this is going to fall neatly on traditional political lines. YouGov ran a poll — it’s not the most scientific poll — but 69 percent of people supported a pause in A.I. development. That makes it a bipartisan issue. I think, in the U.S., there’s a real chance to do something in a bipartisan way. And in Europe, people are like, “This violates the G.D.P.R.” So I’m more optimistic about this than I am about a lot of things. The exact nature is totally up for grabs, but there’s a strong desire to do something. It’s like many other domains: You have to build infrastructure in order to make sure things are OK, like building codes and the UL standards for electrical wiring and appliances. They may not like the code, but people live with the code. We need to build a code here of what’s acceptable and who’s responsible. I’m moderately optimistic. On the other hand, I’m very pessimistic that if we don’t, we’re in trouble.

There’s a belief that A.I. development should be paused until we can know whether it presents undue risks. But given how new and dynamic A.I. is, how could we even know what all the undue risks are? You don’t. It’s part of the problem. I actually wrote my own pause letter, so to speak, with Michelle Rempel Garner. We called for a pause not on research but on deployment. The notion was that if you’re going to deploy it on a wide scale, let’s say 100 million people, then you should have to build a safety case for it, the way you do in medicine. OpenAI kind of did a version of this with their system card. They said, “Here are 12 risks.” But they didn’t actually make a case that the benefits of people typing faster and having fun with chatbots outweighed the risks! Sam Altman has acknowledged that there’s a risk of misinformation, massive cybercrime. OK, that’s nice that you acknowledge it. Now the next step ought to be, before you have widespread release, let’s have somebody decide: Do those risks outweigh the benefits? Or how are we even going to decide that? And at the moment, the power to release something is entirely with the companies. It has to change.

This interview has been edited and condensed for clarity from two conversations.

Source