The AI Daily Brief: Artificial Intelligence News and Analysis - The AI Arms Race Heats Up: US Considers Blocking AI Chip Exports

Episode Date: June 29, 2023

The Commerce Department is considering blocking AI chip exports from companies like Nvidia to China. Google DeepMind CEO Demis Hassabis claims their next model will eclipse ChatGPT. Unity releases AI platform including Unity Muse. NOTE: While NLW is traveling this week, The AI Breakdown will only be releasing The Brief each morning. We'll be back to our regular content at the end of the week.    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 talking about the acceleration of the arms race in AI. The U.S. is considering more restrictions on AI chip exports, and Google DeepMind CEO says their next model will eclipse chat GPT. The AI breakdown is a daily podcast and video about the most important news and discussions in AI. Go to breakdown.network for more information. Hello, friends. Welcome back to another AI breakdown brief. As you guys have heard, this week while I'm traveling in Europe, I am doing just the brief, not the main part of the episode. so at least you still get the headlines and can stay informed about what's going on

big picture-wise in AI. We are getting ready to head back, so we will be back to the normal episode soon. But in the meantime, if you're enjoying the AI breakdown, or in this case, the AI breakdown briefs, a really helpful thing you could do would be to just share it with one other person. It could be your sister, your brother,

your mom, your dad, your dog, your cat. Anyone who you think AI is going to impact their lives, share an episode that you think they should listen to. I love when communities grow organically, so this would be a major, major help. With that, let's go to the AI breakdown brief. Welcome back to the AI breakdown brief. All the AI headline news you need in five-ish minutes or less.

Today we start in the realm of the geopolitical. Since the launch of ChatGPT last November, OpenAI has clearly been in the lead when it comes to consumer-facing LLMs. Now, obviously in the U.S., they've seen increased competition from companies, including Google, as well as newer upstarts, including anthropic and inflection. But OpenAI's competitors aren't limited.

to the U.S. alone. In China, search giant Baidu has said that the latest version of their AI model, Ernie, is actually already beating GPT4 in many tests. In a statement on Tuesday, they said that their 3.5 version of their Ernie AI model had surpassed, quote, chat GPT and comprehensive ability scores and outperformed GPT4 in several Chinese capabilities. The evidence they cited was a test that was run by Chinese Science Daily, a state newspaper, that used a number of common benchmarks to evaluate the performance of AI models. Now, OpenAI did not respond to a Reuters request for comment, but this is not the biggest news when it comes to China and the U.S. in the battle for AI supremacy.

The Wall Street Journal reported yesterday evening that the U.S. is considering new restrictions on AI chip exports to China. As early as next month, the Commerce Department could halt shipments of AI chips from companies like Nvidia and others to customers in China that don't first obtain a license. Now, if this goes through, it would be part of finalizing the export control measures that were announced last October. This is not the first Chinese chip restriction that the U.S. has imposed. Last year, new restrictions came online that cut off Chinese companies' access to the most advanced AI chips from Nvidia and AMD,

which Nvidia responded to by making a new version of its AI chip specifically for the Chinese market that fell below the performance thresholds that had been outlined by the Commerce Department. The new restrictions that are being contemplated now would ban the sale even of that chip, the A800, for companies that didn't have a license. According to the Wall Street Journal, if the Biden administration goes through with this, it will likely happen after a visit to China by Treasury Secretary Janet Yellen in general. July. Bloomberg also has an article out today about the AI arms race between China and the U.S., describing how China is calling on the private sector to keep up and try to get ahead of the U.S. when it comes to

AI. Bloomberg writes, China's tech sector has a new obsession competing with U.S. Titans like Google and Microsoft in the breakneck global artificial intelligence race. Still, the road to walk is long. Bloomberg points out that AI investments in China this year stand at about $4 billion, as opposed to $26.6 billion in the U.S. But staying on this theme of competing with open. in AI and chat GPT, it is definitely not just Chinese rivals that have Sam Altman's company in their sites. Google DeepMind CEO Demi's Hizhabis has said in a new interview that their forthcoming system, Gemini, will tap into techniques from their AlphaGo and Alpha Zen AI to build a new model that will, in his words, eclipse Chat GPT. Hasabas said, at a high level you can think

of Gemini as combining some of the strengths of AlphaGo type systems with the amazing language capabilities of the large models. We also have some new innovations that are going to be pretty interesting. Still, when you read this piece in WIRE, to the extent that this was meant to be a PR coup for Google talking about how they're going to get out ahead of OpenAI, it kind of only serves to reinforce just how far ahead OpenAI actually is. And by the way, it's not like OpenAI is sitting still. According to new reporting from the information, OpenAI is planning to set up a personal assistant for work in their chat GPT suite. In conversations with developers, OpenAI CEO Sam Altman said that they want to turn chat GPT into a quote, super smart personal assistant for work. As the

information puts it, with built-in knowledge about an individual in their workplace, such an assistant could carry out tasks such as drafting emails or documents in that person's style and with up-to-date information about their business. Now, the theme of the information article is that this puts it on a collision course with Microsoft, who is of course its biggest investor. However, given that Microsoft is benefiting from both sides of this competition, it's not totally clear to me how concerned they'll actually be. There's also the question of whether when it comes to business, the LLM winners will be retrofitted versions of the main models or new startups that are custom built from the ground up for an enterprise use case. On that front, RACA has just emerged from stealth with a $58 million

financing round to try to go compete on this AI enterprise level. Their first product is a multimodal AI assistant that's trained to understand images, videos, and tabular data in addition to words and phrases, and is optimized to derive insights from a company's internal data. And all this competition certainly has markets hyped. A Wedbush analyst said that AI is the, quote, fourth industrial revolution and will see $1 trillion in investment. Dan Ives told CNB, I do not believe that this is a hype cycle. Instead, he compared it to a 1995 moment when the internet really broke out. There is evidence that other investors share Ives' point of view. The S&P 500 has risen 14% so far this year, largely on the back of AI excitement, but the tech

sector focused NASDAQ has gone up 36% since the beginning of January. Last up today, speaking of how AI changes everything, Unity's new AI gaming platform suggests just how much text-based and natural language inputs are going to change the way that we create things with computers. Unity has just announced two parts of their AI platform. Unity Muse, they call a platform of AI capabilities for content creation. And Unity Sentists, they describe as AI-powered experiences at Unity runtime. Basically, from their announcement post, it looks like Unity Muse is going to be a tool by which developers can use AI to build things in a much more natural language-aligned way.

As they write, the eventual goal of Muse is to enable you to create almost, anything in the Unity editor using natural input such as text prompts and sketches. So imagine being able to create characters, worlds, environments, and more, simply by typing in what you want that world or character to be like. They also announced a closed beta feature called Muse Chat, which leverages AI to search across Unity documentation, training resources, and support content to help people who are building games get answers faster to the questions that are holding them up. Now, when it comes to Unity Sentis, it appears to be a tool that will allow developers to embed AI models directly in their games.

Unity writes, Sentis enables you to embed an AI model in the Unity runtime for your gamer application, enhancing gameplay and other functionality directly on end-user platforms. Centis allows AI models to run on any device where Unity runs. It's the first and only cross-platform solution for embedding AI models into a real-time 3D engine. The last part of Unity's AI announcement was that third-party AI packages that meet Unity standards are now available on the Unity Asset Store.

so other companies that are building AI solutions for game development can now get those tools directly in front of Unity developers. Given how universal gaming is, it's hard not to imagine that this is going to be a primary way that people experience the outputs of AI and generative AI specifically over the next couple of years. Anyways, friends, that is it for today's AI breakdown brief. Go check out the podcast version of the show or subscribe to the newsletter. And until next time, peace.