The AI Daily Brief: Artificial Intelligence News and Analysis - Is Pika 1.0 AI Video's Breakout Moment?

Episode Date: November 28, 2023

Pika Labs officially launched 1.0 and many think it might be representative of a coming breakout moment for text-to-video generation. Plus a look at the complicated geopolitical tension around AI between the US, China and the United Arab Emirates. Today's Sponsors: Notion - Notion AI. Knowledge, answers, ideas. One click away. - https://notion.com/aibreakdown ABOUT THE AI BREAKDOWN The AI Breakdown helps you understand the most important news and discussions in AI.  Subscribe to The AI Breakdown newsletter: https://theaibreakdown.beehiiv.com/subscribe Subscribe to The AI Breakdown on YouTube: https://www.youtube.com/@TheAIBreakdown Join the community: bit.ly/aibreakdown Learn more: http://breakdown.network/

Today on the AI breakdown, we're looking at the complicated geopolitical relationship between the U.S., China, and the United Arab Emirates, and what it has to do with OpenAI. Before that on the brief, PECA Labs 1.0 launches an idea-to-video platform, and the AI community is stoked. The AI breakdown is a daily podcast and video about the most important news and discussions in AI. We're to Breakdown Network for more information about our YouTube, our Discord, and our newsletter. Welcome back to the AI Breakdown Brief. All the AI headline news you need in Arirmed. round five minutes. We kick off today with an exciting new product launch. In fact, one that some

people are calling one of the most exciting product launches of the year. The launch is from PICA Labs, PICA 1.0, and what it is is a text video generation platform, or as they call it, an idea to video platform that brings your creativity to life. So there's a lot that is very cool about this tool. First of all, not only does it allow you to go from an idea written in plain text to a video, but it also allows you to edit your own videos and change aspects of them in the same way that, for example, generative Phil does for photos in Adobe Photoshop. The example they use in their launch video is a woman who's changing her top as she walks down the street, or adding sunglasses to a monkey. There's also the ability to resize videos and expand them automatically, which again, even the image version of this

feature was getting people excited just a few months ago. Now, in addition to the announcement that the company was launching PICA 1.0, they also announced a $35 million. Series A, led by Lightspeed. The company was founded by two women from Stanford, one of whom Demiguo recently left her Stanford AI PhD to build this company. Now, I think in many ways, Andre Carpathy from OpenAI, has the best description of why this is such an exciting development, and what this says about the state of this particular field. He writes, You know how image generation went from blurry 32 by 32 texture patches to high-resolution images that are difficult to distinguish from real and roughly the snap of a finger?

the same is now happening along the time access extended to video, and the repercussion boggle the mind just a bit. Every human becomes a director of multimodal dreams, like the architect in Inception. Coming back to Earth for a second, image and video generation is a perfect match for data-hungry neural nets because data is plentiful, and the pixels of each image or video are a huge source of bits, on the parameters of the network. When you're training giant neural nets in supervision-rich settings, your train loss equals validation loss and life is so good. So here we have Andre being excited about this both from the standpoint of technology and how this opens up new sets of information to training AI models, but also because of the unlocking of human creativity that this is going to create. Now, even other nominal competitors are excited about this launch.

Imam Mastok from Stability AI, who just launched their own stable video diffusion model last week, writes, Amazing work by Pika Labs. Congrats to Chenlin Meng and Demiguo, shipping great stuff with a small, talented, and focused team. I don't think this is just him being gracious. I think that the folks who are building these models understand that right now, it's a rising tide lifts all boats kind of moment when it comes to AI video generation. It really does seem like a very exciting product and one that I can't wait to dig into further.

Next up, yesterday we were talking about Q Star and how its capacity to do grade school math might have been the thing that freaked out the Open AI board sufficiently to actually go out and fire Sam Altman. Well, now there is a $10 million competition to fund the development of AI models that can reason mathematically. The announcement page writes, XTX Markets is launching a new $10 million challenge fund, the Artificial Intelligence Mathematical Olympiad Prize, or AIMO Prize.

The fund intends to spur the development of AI models that can reason mathematically, leading to the creation of a publicly shared AI model capable of winning a gold medal in the International Mathematical Olympiad. The grand prize of $5 million will be awarded to the first publicly shared AI model to enter an AIMO-approved competition and perform at a standard equivalent to a gold medal in the IMO. There will also be a series of progress prizes totaling up to 5 million for publicly shared AI models that achieve key milestones towards the grand prize. Now, it seems like there are two different goals here.

One is to actually get people working on this particular problem set and to actually dedicate their brain power to solving this challenge of getting AI to reason mathematically, but then the other other is to spur that development in a way that is open to the public. In other words, they don't want this innovation to be limited to open AI or anthropic or any closed lab and instead want it to be widely available. Now, if this is something that excites you, it appears that they are hiring a director to leave this AIMO prize right now. XTX Markets founder Alex Gurko writes, first order of business with AIMO prize, we're looking for somebody to run this exciting project. So if that's you, that is something to maybe check out.

Next up today, a big dust up around AI generated content. CNN Business writes, Sports Illustrated deletes articles published under fake author names and AI generated profile photos. So basically it appears like sports. Sports Illustrated was experimenting with AI-generated content. That means not only using AI to write content, but actually inventing fake authors as well. All of this came to light after the site Future Sim started investigating and found, as CNN put it, that the magazine had repeatedly published articles whose authors could not be found online outside the Sports Illustrated website.

The articles were all accompanied by AI-generated profile photos that Future Sim also found for sale on digital marketplaces that sell AI-produced headshots. Continuing, they write, During the course of Future Sims reporting, some of the alleged Sports Illustrated writers mysteriously vanished from the publications website. And their articles began appearing under the names of different authors who similarly didn't appear to exist online and whose likenesses were also being sold on AI headshot marketplaces. Now, obviously right now, there are no exactly agreed upon ethics when it comes to AI generated content in publications and content websites. Certainly this is something that people are experimenting with, often facing a lot of backlash when they do. One thing that is almost definitely for sure, though, is that people do not like being lied to and tricked. It's one thing to have a publisher say, hey, we're going to test AI publishing.

Check out this article, see if you like it. This might be something we do in the future. This might be something we don't do. It's another thing entirely to try to create fake avatars that trick people into thinking it's real when actually it's been generated by AI. In fact, this seems to be one of the areas that there is wide bipartisan agreement in potential upcoming AI legislation that publishers and platforms are going to have to identify when something is created by AI so as to not so disinformation and mistrust.

For their part, the company that operates in License Sports Illustrated, the Arena Group, told news outlets that all of this content had been created by a third-party company AdVon Commerce. The spokesperson said, quote, we have learned that Advon had writers use a pen or pseudonym in certain articles to protect author privacy, actions we don't condone, and we are removing the content while our internal investigation continues and have since ended the partnership. Now, according to this, the Orina Group also says that although they use those fake images, that, quote, Advon has assured us that all of the articles in question were written and edited by humans. I would just like to insert the raised eyebrow emoji here, as something about that doesn't really track.

Now, following the report, the union of Sports Illustrated writers, which is called the Sports Illustrated Union, said that its members were, quote, horrified by these allegations. In a statement they wrote, if true, these practices violate everything we believe in about journalism. We deplore being associated with something so disrespectful to our readers. Emma Basilieri, a staff writer for the magazine, also posted on X, saying, along with basic principles of honesty, trust, journalistic, ethics, etc., I take seriously the weight of a Sports Illustrated byline. It meant something to be long before I ever dreamed of working here.

This report was horrifying to read. Now, of course, this is not the first, nor will it be the last experiment with AI-generated human alternatives. You may have seen a story about a new AI model, Aetana, who was designed, signed by a Spanish agency and has accumulated over 120,000 followers on Instagram since launching over the summer. The AI generated model is apparently earning up to $10,000 a month on brand deals. There is basically an endless amount that we could discuss on this, and indeed my guess is that entire sociology thesis will be dedicated to this this year. Now, beyond just models or

sports illustrated writers, there is apparently a large sense among regular people that AI could be coming for their jobs. In fact, Fox Business published a piece this morning titled two-thirds of Americans say AI could do their job. The reporting was on a recent survey conducted by Spokio, which found that of over 1,000 respondents, 66.6% of them said that AI could carry out their workplace duties, and nearly 75% of them said that they were concerned about the technology's impact on their industry. Now, interestingly, Spokio CEO Harrison Tang pointed out that this might be about their actual job tasks being able to be automated, but it also might be about how much the media has presented AI as a great disruptor. Tang said, whether it's because people

realize that a lot of work can be easily automated or they believe the hype in the media that AI is more advanced and powerful than it is, the AI box has now been opened. Yes, indeed it has, and I am glad you are here exploring this new world with me. That's going to do it for the AI breakdown brief. Next up, the main AI breakdown. And now a quick word from today's sponsor. I am a huge notion user. We're talking multiple accounts for multiple projects. I use it for everything from applicant tracking to note taking to project management, to sharing public documents, to frantically capturing ideas I have while out hiking or just driving around. Given that and given the topic of the AI breakdown, I was excited to learn that they've launched a new AI tool called Q&A.

It's like a personal assistant that responds in seconds with exactly what you need. Notion AI can give you instant answers to your questions using information from across your wiki, projects, docs, and meeting notes. For someone like me who makes dozens of notes per day around a huge array of topics, having a built-in AI tool to help recall that is incredibly useful. Now, beyond that use case, think about this. Have an urgent question you normally turn to a coworker to answer? Just ask Q&A instead. It'll search through thousands of documents in seconds and answer your question in clear language

no matter how large or complex your workspace is. Plus, you can trust your data is secure because Notion AI is designed to protect your information. No AI models are trained with your information, the data is encrypted, and answers will never use information from pages you don't have access to. With Notion AI, it's even easier to do your most meaningful work. Try Notion AI for free when you go to notion.com slash AI breakdown. That's all lowercase letters, notion.com slash AI breakdown to try the powerful, easy to use notion AI today.

And when you use our link, you're supporting the show. One more time, that's notion.com slash AI breakdown. Welcome back to the AI breakdown. The most fraught relationship in geopolitics in general is, of course, the U.S. and China. It stands to reason then that the most fraught relationship in the geopolitics of AI is, once again, the U.S. and China. Indeed, that relationship is so fraught that this year, much of the efforts of the White House and various security agencies has gone into not China specifically but its proxies around the

world. The New York Times has just released a major research piece called Inside U.S. efforts to untangle an AI giant's ties to China. The story is about an Emirates-based firm called G-42 and why the U.S. is concerned about its work with Chinese companies. Now, to understand why this story might be interesting, especially in the context of recent events in the AI space, let's go back to October 18th and a story from Reuters titled ChatGPT creator OpenAI partners with Abu Dhabi's G42 as UAE scales up AI adoption. Now, this was a small announcement. It came from G42 themselves, and the details were fairly

scant. Reuters writes, G42 said it would utilize OpenAI's generative AI models in areas including financial services, energy, health care, and public services. That's literally it. That was all that was announced. However, that was not the first time this year that G42 was in the news around its AI efforts. Back in July, a new $100 million AI supercomputer deal was announced between G42 and Cerebris Systems. Cerebris, which is a company based in Silicon Valley, announced that G42 had agreed to purchase three of its first Condor Galaxy supercomputers, and that that deal was part of a larger strategic partnership where G42 could buy up to to six additional supercomputers by late 2024. The deal was so significant in size that the

CEO of Cerebrus Andrew Feldman was actually planning to move to the UAE for three months to work on developing G42's products around the systems. The only detail about what G42 was planning to do with it is this line where they said they planned to use the Cerebris systems to sell AI computing services to health care and energy companies. Fast forward a month. At the end of August news reports came out that the U.S. had been putting pressure on companies including invidia and AMD, not only to block exports of chips to China as per export restrictions that had gone into effect at the end of 2022, but also to try to block sales to parts of the Middle East. Now, this came from a regulatory filing from NVIDIA, which said that the restrictions would affect its A-100 and H-100 chips, and while they wouldn't

have a, quote, immediate material impact, they were a concern in the long run. Now, after the NVIDia story broke, Reuters confirmed that AMD had also been asked to place similar restrictions on their business by the U.S. government. In a statement, NVIDIA said, during the second quarter of fiscal year 2024, the U.S. government informed us of an additional licensing requirement for a subset of A-100 and H-100 products destined to certain customers in other regions, including some countries in the Middle East. Now, at the time, the supposition was widely reported as not the U.S. government taking issue with the Middle East itself, but the concern that these advanced chips that went to the Middle East would find their way to Chinese companies or the Chinese

government. Adding additional intrigue, however, the U.S. government then came out and said that they weren't blocking chip sales to the Middle East. After the regulatory filings, the departing Department of Commerce spokesperson said that the Biden administration had not blocked chip sales to the Middle East, adding just a whole lot of confusion to the situation. Okay, so where we were is one, there's seeming to be increasing U.S. government focus on how Middle Eastern countries, particularly in the Gulf region, could be conduits for AI technology getting to the Chinese. Two, we have the Gulf nations, particularly the United Arab Emirates, cutting deals with big Silicon Valley players, including cerebrus and open AI. So what is the substance of this New York Times piece about this company G42? The lead of the piece reads, When the secretive national security advisor of the United Arab Emirates, Sheikh Tannun bin Say had visited the White House in June, his American counterpart, Jake Sullivan, raised a delicate issue. G42, an artificial intelligence firm controlled by the Sheikh that American officials

believe is hiding the extent of its work with China. In public, the company has announced its staggering growth with a steady cadence of news releases. They've included agreements with European pharmaceutical giants like AstraZeneca and a $100 million deal with the Silicon Valley firm to build with the company's boast will be the world's largest supercomputer. Last month, G42 announced a partnership with OpenAI, the creator of ChatGPT. But in classified American intelligence channels, there have been more concerning reports about the company. The CIA and other American spy agencies have issued warnings about G42's work with large Chinese companies that U.S. officials consider security threats,

including Huawei, the telecommunications giant that is under U.S. sanctions. U.S. officials feared G42 could be a conduit by which advanced American technology is siphoned to Chinese companies or the government. The CIA even produced a classified profile. file of Peng Xiao, the chief executive of G-42, who was educated in the United States and renounced his American citizenship for an emirati one. Basically, the argument that this piece makes is that G-42 is at the epicenter of the U.S.'s goals to limit China's ability to both influence the Middle East, as well as to access advanced technology. The article also points out that the stakes for the UAE are high, arguing that they're, quote, building an artificial intelligence industry as an

alternative to oil income. Now, the article goes deep on all the different ways in which there might be some overlap between G42 and the Chinese government. G42's underlying technology infrastructure has been built with the help of Chinese companies, including Huawei, the company sells surveillance technology that's very similar to products that are popular with Chinese police. G42's portfolio includes a $100 million investment in bite dance, which is obviously the owner of TikTok, which has been at the epicenter of questions of the relationship between China and the U.S., and so on and so forth. Now, OpenAI and Cerebrus are not the only companies that have relationships with G-42. Microsoft and Dell technologies do as well.

What's more, Andrew Feldman, the CEO of Cerebrus, said he was, quote, surprised to hear about the Biden administration's concerns about G-42 and said that no American officials have raised those concerns with him or other officers at Cerebrus. For this article, OpenAI and Microsoft declined to comment. Now, I'm not sure that if you read this piece, you can really come away with many conclusions other than how complex the relationship between China and the U.S. really is, especially around this question of AI, that the Gulf region in general and the UAE in specific are part of a growing thread in that tension, which is countries that have relationships with both groups. Three, that part of this story is the blurring of the lines between state and the private

sector, which is causing even more challenges. And four, that G42 is, if nothing else, very ambitious. Now, part of what makes this interesting in the context of the recent Open AI scruffle was the reporting that came out between when Altman was fired and when he was rehired, that he had been, quote, looking to raise tens of billions of dollars from Middle Eastern sovereign wealth funds to create an AI chip startup to compete with processors made by NVIDIA. Security expert Matthew Pines wrote a threat about this. He writes, why did the Open AI board bounce Sam? Nobody really knows. But there are public reports of Sam's geopolitical activities that might have raised concerns. Below are some events that may explain the board's accusation of lack of candor. On June 6,

2023, Sam spoke at a Hub 71 forum in Abu Dhabi, an event put on in partnership with G42 and backed by the UAE sovereign wealth fund, Mubadala. Opening for him was the CEO of the Inception Institute for Artificial Intelligence, a G42 company founded by a Chinese national. Note that G42, which is chaired by the UAE's national security advisor, launched a $10 billion 42x tech fund in August 2022 in partnership with Mubadala. They hired a former executive of Chinese e-commerce giant JD.com to lead their Chinese investments and expand across Asia. On June 10, 2003, Sam gave the keynote virtually at the Beijing Academy of Artificial Intelligence, an event which involved Chinese firms sanctioned by the U.S. There, he emphasized the need for U.S. and Chinese researchers to collaborate on AI development

to manage global risks. AI has become the locus of U.S.-China competition in the Gulf, where KSA and UAE both looked to play both sides, security protection from the U.S., and tech transfer and trade with China. G42, led by the UAE's spy shake, is at the center of this swirling geopolitical. and tech vortex. In August 2023, G42 launched Jase, a chat-GBT-T-like LLM purpose built for the Arabic world. This LLM was trained using cerebrous chips, known to be

inferior to NVIDIAs. G42 also partnered with Chinese companies to boost COVID testing and vaccine development, aka mass DNA collection. The UAE has a national strategy for AI 2031, aiming to catch up on the global tech frontier, and strategically shift long-term away from an oil economy. This aligns

with China's national strategic ambition to become a major AI innovation center by 2030. In August 2023, the U.S. tapped Nvidia and AMD on the shoulder and said no moss to any advanced GPU sales to KSA and UAE. This was motivated out of an apparent concern for pass-through sales to China that bypass export controls, but it's likely G42's indigenous efforts also played a role. The U.S. has also expressed acute concerns over Mubodala and SoftBank's investment activities. CFIUS has intervened to scrutinize and block some of their recent acquisition targets on national security grounds. U.S. has also flagged a Chinese military facility, at Califa Port and UAE.

Matthew then flags what we had just talked about, with the October deal between G42 and OpenAI, and the report from Bloomberg, that the OAI board had been concerned that Sam had been looking to raise tens of billions of dollars from Middle Eastern sovereign wealth funds, and then he added on November 19th, Yahoo confirmed that Altman had been traveling

to the Middle East to fundraise for the project, which was codenamed Tigris, and raised tens of billions from SoftBank, Saudi's public investment fund, and the Mubudala Investment Company. Matthew concludes, this is all in the public domain. One has to imagine that the,

OAI board is privy to more details, some of which may have been shared by the U.S. intelligence community, who give defensive briefings to U.S. firms, which gave them serious concerns about Sam Altman's lack of candor. Now that Sam has been fired and rumors swirl about his reinstatement, I have no idea how the situation evolves. But the above hypothesis would be confirmed if the G42 and other deals were suspended as a condition of his return. Now, of course, now we are sitting here with the benefit of hindsight and know that that didn't happen. So perhaps it's all just coincidence. The Occam's razor explanation for all of this, is that Sam as the entrepreneur he is,

recognizes that the stranglehold that NVIDIA has over the entire AI industry is an opportunity, and that there are very few people who have the capital necessary to actually go and compete to take advantage of that opportunity, and thus partners with an aligned interest, including these big sovereign wealth funds and companies like G42, are, of course, the companies he's talking to. What I think this New York Times story shows

is that even if the explanation is as simple as that, it's not actually as simple as that. It's a great example of how AI is going to be a strong, not just of technology and not just a business, but of the shape and structure of the world that we move into. Interesting stuff to keep an eye on, but until next time, peace.