The AI Daily Brief: Artificial Intelligence News and Analysis - Who's Winning the AI Race

Episode Date: November 29, 2024

A year on from OpenAI's board debacle, a look at the state of the AI race -- from OpenAI to xAI to Google and beyond. Brought to you by: Vanta - Simplify compliance - ⁠⁠⁠⁠⁠⁠⁠https://vanta.com/nlw Plumb - AI automation that just works - https://useplumb.com/ The AI Daily Brief helps you understand the most important news and discussions in AI. Subscribe to the podcast version of The AI Daily Brief wherever you listen: https://pod.link/1680633614 Subscribe to the newsletter: https://aidailybrief.beehiiv.com/ Join our Discord: https://bit.ly/aibreakdown

A year after Sam Altman's firing and rehiring at Open AI, who is actually winning the Gen AI War? The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, follow the Discord link in our show notes. Hello, friends, we are right smack in the thick of turkey and mashed potato-related hangovers, which means that we're going to do something a little bit different for today's episode of the AI Daily Brief. It was almost exactly a year ago when the Open AI board fired, Sam Altman, causing utter chaos not only at OpenAI, but totally opening up the sense of the possible when it came to the competition in the AI space more broadly. Now, of course, we know that Altman was

eventually reinstated, but now a year on, I think it's interesting to ask, as Axios did on Wednesday, who's winning the AI race? So what we're going to do today is go through company by company, specifically the frontier labs, discuss where they are and what it says about how the competition is evolving. And let's start with Microsoft. the partner who was potentially most going to be impacted by Altman's firing. You might remember that during that whole episode, CEO Satya Nadella effectively invited Altman to bring over the entire OpenAI team into Microsoft. In fact, they announced that it was a done deal.

That was part of the pressure that ultimately led to Sam Altman being reinstated, but there were clearly going to be consequences with that Microsoft relationship. The net of those consequences became clear, a few months later in March, when the team from Inflection, which had raised about a billion and a half dollars just a year before, was gutted to come join a new division inside Microsoft to coordinate all their AI efforts. That new division would be led by Mustafa Sullyman, who was both the co-founder of DeepMind and later the co-founder of Inflection. To me, this read very clearly like a hedge from Microsoft when it came to their relationship with OpenAI,

an unwillingness to let the vagaries of that company ultimately shape their place in the AI battle. One of the things, however, that I've said frequently is that just because they felt like they had to make this hedge doesn't mean that they're lying when they say things like they still want the relationship with OpenAI to be great. My sense is that, remember, Microsoft's deal with OpenAI has this very weird clause where when OpenAI hits AGI as determined by the Open AI board, Microsoft's commercial relationship ceases to exist. In other words, the Open AI board has a lot of power to shape Microsoft's financial destiny when it comes to their relationship with OpenAI.

And that's not a very comfortable position, especially after you've seen the board behave in a way that some might call capricious. In other words, it felt to me like from a sheer fiduciary duty point of view, Microsoft had to hedge that relationship a little bit more. And so bringing in this very senior team kind of just made sense. Now, obviously, Microsoft has done a lot more. They've continued to try to roll out co-pilot across all of their products

at varying levels of success. They're now hopping into the agent era. but by and large, I think that 2024 will be seen as a restructuring and infrastructure year when it comes to that particular company. What about OpenAI? There were a few notable moments. First, the company showed way back in February that they still had some serious chops when it came to the state of the art, blowing people away when they announced SORA.

Sora was, of course, their video generation model, although to this day we still haven't actually gotten access to it. Other, of course, than a leak earlier this week, where some of the artists, who had early access, shared discruntedly a version of it on Hugging Face. For much of the year, OpenAI has been unable to escape drama around leadership. Numerous senior executives, including CTO Mira Moradi, have left. Co-founders Ilya Sutskaver and Andre Carpathy left. Ilya specifically to build a competing frontier lab

that wouldn't concern itself with trifling things like revenue and business models. And more recently, OpenAI has been at the forefront of this question of whether we're facing limitations in our current scaling models for frontier model performance. OpenAI is currently investing a lot more in reasoning models and test time compute and a number of other strategies. But notably what we haven't gotten this year, and which Sam Altman has pledged that we will not get, is any version of GPT5. That in and of itself seems to suggest, or at least lend credence to the idea that there may be a plateau that we're reaching. Although with cryptic tweets like this one from Sam Altman, where he said there is no wall, there's also some denial of that.

that. One of the companies that has benefited the most from OpenAI's turbulence is lead competitor Anthropic. Menlo Ventures recently did a report called the state of generative AI in the enterprise, and they found that whereas OpenAI's share of enterprise business had dropped from 50 to 34% this year, Anthropics had doubled from 12 to 24%. In other words, Anthropic closed the gap with OpenAI from a 38% differential to a 10% differential, at least in the enterprise. Of course, OpenAI is still way ahead when it comes to consumer revenue, but it shows that Anthropic is absolutely not just an also-ran. For much of the year, their Claude 3.5 model was seen as the best. But even more than that, I think where Anthropic has really excelled and pushed the pedal to the metal and forced open AI

to respond to them is with features and user experience. In May, their new chief product officer was Instagram co-founder Mike Krieger, their artifacts feature, which separates the creation and input panel from the output panel using Claude, makes the previous LLM interface feel old and clunky and forced others like ChatGBT to copy them, and they continue to push out new features. I just recently talked about their new style presets, where you can both create a style that imitates your own writing style, as well as just create a set of presets that match any particular style you want. Importantly, though, this is embedded in the UIUX rather than being something that you have to prompt. Now, I don't think that with these features alone, Anthropic

would be gaining ground on OpenAI, it's also because their models are performing really well. One company that seemingly had a harder year is Mistral. Mistral is a French lab that in the second half of last year really took on the banner of open source, effectively competing for narrative mind share in that space with meta. And according to that same Menlo study, their market share of enterprise LLM use actually went down from 6 to 5% this year. From outside, it seems to me like they're struggling to keep up with the huge budgets that other companies have, while also retaining their open source routes. And so I think it'll be an important one to watch, if only because it might indicate something about just how challenging it is to compete in the frontier lab space.

Today's episode is brought to you by Vanta. Whether you're starting or scaling your company's security program, demonstrating top-notch security practices, and establishing trust is more important than ever. Venta automates compliance for ISO-2-GDPR and leading AI frameworks like ISO 402,001 and NIST AI risk management framework, saving you time and money while helping you build customer trust. Plus, you can streamline security reviews by automating questionnaires and demonstrating your security posture with a customer-facing trust center all powered by Vanta AI. Over 8,000 global companies like Langecane, Lila AI, and factory AI use Vanta to demonstrate AI trust and prove security in real time. Learn more at vanta.com slash NLW. That's vanta.com

slash NLW. Today's episode is brought to you, as always, by Super Intelligent. Have you ever wanted an AI Daily Brief, but totally focused on how AI relates to your company? Is your company struggling with AI adoption, either because you're getting stalled, figuring out what use cases will drive value, or because the AI transformation that is happening, is siloed at individual teams, departments, and employees, and not able to change the company as a whole?

Super Intelligent has developed a new custom internal podcast product that inspires your teams by sharing the best AI use cases from inside and outside your company. Think of it as an AI Daily Brief, but just for your company's AI use cases. If you'd like to learn more, go to Bsuper.aI slash partner and fill out the information request form. I am really excited about this product, so I will personally get right back to you. Again, that's Bsuper.a.ai slash partner. And then, of course, there is meta, the one among the big tech. techs who has anchored their flag to open source and has continued to push in that theme all year. As Fortune Sharon Goldman recently put it, Mark Zuckerberg has fully rebuilt meta around Lama.

The company released Lama 3.1 this year, including their large 405B model, and they are aggressively weaving generative AI from their models into all of their products. They are taking full advantage, in other words, of the distribution channels they have via WhatsApp, Instagram, Facebook, to get people interacting with AI. What's more, they're also starting to prove out the business use case by showing how AI-enabled ad production in those platforms is outperforming when people don't use AI. Meta hasn't necessarily been the most loud when it comes to AI this year, but they are carving out a very important, very powerful spot and should not be faded. Interestingly, there also have been recent reports that Meta is trying to get the U.S. to adopt Lama as their AI standard. Of course, in the wake of Donald Trump being elected president, Zuckerberg is hardly the only

big tech leader jockeying for position. No person better represents techs shift to the right into Donald Trump more than Elon Musk. Musk, of course, went all in with Trump towards the end of the campaign and has since been seen as being extremely influential during the transition process. There are, of course, questions already around how long that influence will last that are way outside the scope of this AI show. But it certainly is part of the texture of the AI. story for 2025 in the decisions that are made by the Trump White House will have a big impact on this

industry. Meanwhile, Elon's XAI, while not necessarily releasing a hugely disruptive model, Grock is still available primarily just in Twitter, raised money at first a $24 billion and more recently a $50 billion valuation, reflecting, I think, the power of a built-in distribution channel like Twitter, but also the strength of the conventional wisdom, which is not to bet against Elon. Apparently, XAI is also going to have a consumer app soon, which could give it more of a chance to actually compete with ChatGBT. One thing that Elon does appear to be influential in is who Trump is going to choose as his AIs are. The Trump team is apparently considering elevating a White House senior position to focus on artificial intelligence, which again could have a huge impact on how this

industry evolves over the next couple of years. For some other big tech CEOs, Musk is uncomfortably in the middle of their contact with Donald Trump. Most notably, he, recently joined the call where Google CEO Sundarpa Chai called to congratulate Trump, a highly uncomfortable conversation one has to imagine. But let's use this to talk about Google. Google's year in AI did not get off to a great start. First of all, the perception of being behind Open AI and having surrendered their lead in AI has dogged Google basically for the last two years. It felt rushed and forced when they announced Gemini last December, exemplified by the fact that the most performant version of the model wasn't actually going to be available for a number of months.

What's worse, Google's year got off to a very controversial start when Google's attempts to combat bias in AI ended up producing images that had, among other things, racially diverse Nazis. Google ended up having to take their image generation model offline for a while and ultimately apologize for, quote, missing the mark with it. And yet, none of that will be the story that people remember from Google from 2024 when it comes to AI. instead what they will remember is Notebook LM, which is many people's vote for these single best AI product of the year. Notebook LM has a very specific feature, the ability to take a bunch of written material and turn it into an audio podcast discussion between two hosts. It is fully automated and generated. They've added the ability also to guide and customize it.

And the results are really, really impressive. People are using it to change the way they study. They're using it to change the way they distribute information inside of work. It's an incredibly powerful tool and has put some win back in the sales of Google from a product perspective. At Super Intelligent, we deal exclusively with big enterprise customers, and Notebook LM has really got people excited. Even up to a couple days ago, you were getting stories like this one in ZDNet. Even NVIDIA's CEO is obsessed with Google's notebook LM AI tool. In other words, Google has a lot better momentum heading into next year than they did heading into 2024.

Speaking of Nvidia, we're not going to get into their continued dominance in the chip space, but I will note that quietly, they keep dropping their own models that perform really well, and one has to wonder how they're thinking about the long term of that company, and how much it's strictly focused on chips. Speaking of chips, one big tech player that is clearly focused on chips is Amazon. Amazon has invested $8 billion now in Anthropic this year, and it's very clear that the thrust of the investment is about Anthropics Collaborative, on Amazon's tranium chips. This is clearly the vector that they've decided to compete,

even though it's still quite early. And lastly, there's Apple. At the Worldwide Developer Conference, we finally got information around how they were going to approach AI, of course, rebranding it Apple Intelligence instead of artificial intelligence and weaving it into their iPhone. And the optimistic take was that Apple was going to really focus on AI for normal people, to bring AI to experiences that were simple, every day, and radically improved by the presence of AI. That vision hasn't flopped, but it also certainly hasn't fully come to bear yet. Many of those features still aren't available. The ones that are only available in the most advanced hardware.

And unfortunately for Apple, the iPhone 16 isn't selling as well as they had hoped. It is much too early to count Apple out of anything, but 2024 wasn't really a huge one for them when it comes to AI. And so that's the landscape from where I'm sitting right now. On the question of who's winning the AI race, it is still well and truly anyone's game. That in and of itself, I think, is a victory for all of us, and if nothing else, will keep this podcast interesting for some time to come.

For now that, that is going to do it for today's AI Daily Brief. Appreciate you listening or watching, as always, and until next time, peace.