The 10 Biggest AI News Stories of the Past Month

Today on the AI Breakdown, we're counting down the 10 most significant news stories of AI from the last month. 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 about our YouTube, our Discord and our newsletter. Welcome back to the AI Breakdown. Today we are doing something a little bit different. As it is the first day of the month, we're going to do a little bit of a recap of what happened in February because it was quite the set of news. We are going to be counting down from 10 to one the most important AI news events. And kicking it off at number 10 was the announcement of a grand prize winner in the Vesuvius Challenge. The Vesuvius Challenge was announced in 2023. It was a $1 million prize initiative to try to crowdsource an AI solution to reading papyrus scrolls been lost in an ancient library in the eruption of Vesuvius. The scrolls have had a multi hundred year history of people trying and failing to unlock their secrets. And over the past year or so, a set of Silicon Valley investors helped fund this new initiative to try to unlock the power of the AI crowd. To figure it out, the team took high resolution CT scans of the scrolls imaged at a particle accelerator near Oxford in the UK and gave a worldwide community of people access to start exploring solutions for figuring things out. The journey of how these scrolls began to be unlocked, which involves a combination of human insight and AI training, is a fascinating one and I highly suggest you go check out my episode about that. But in terms of why this event ranks as a top news story for February, this was a month, as we'll see when we get to Google's Gemini backlash, where questions of the relationship between AI and history and how AI could be used to obfuscate, control and manipulate history were on full display. So the fact that we also had AI showing its capacity to unlock history that otherwise would have been completely lost to us, feels to me quite important. Next up, we have a story that only ranks as low as it is on this list because it remains highly theoretical. The Wall Street Journal reported at the beginning of the month that Sam Altman was going around the world seeking trillions. Yes, trillions with a T of dollars for a major global AI chip project. 3 the numbers were quoted as between 5 and 7 trillion, which would for reference be somewhere between 6 and 8% of the world's GDP. It sounded from the reports like the genesis for this was not Altman trying to beat Nvidia or anything as pedestrian as that, but instead to create a massive cooperative effort across business and governments to transform the compute capacity of the world. Still, the numbers were so eye popping that it got a ton of attention and and even led to Sam trolling all of us by tweeting eff it, why not eight? He added snarkily. Our comms and legal teams love me so much. Number eight on our list if you thought that the AI rally was going to peter off in the stock market, think again. Nvidia has continued to see major growth after a more muted second half of 2023. This month the company reached a $2 trillion market cap, jumping out ahead of Amazon and Alphabet and landing behind only Microsoft and Apple when it comes to US Companies in terms of market cap. To me, this story reflects two things. One, the continued insatiable demand for AI chips and computing power more broadly. This is not just a hype cycle. Nvidia's value is being driven by incredible revenue and profit growth because there is just such an insatiable demand for their products. When people were asking if this is just a big hype y bubble, some folks pointed out the that Nvidia's profit was growing faster than their market cap. The other side to the story is a very Wall street part of the story though. What the market thinks about AI and how it will impact business is going to shape this, given how much of this is not just a startup movement, but a movement of the big tech companies as well. Speaking of the big tech companies, number seven on our list is a recent one from this week. According to Bloomberg sources, Apple has canceled its electric car project, an initiative it spent over $10 billion in 10 years on. The reporting was that many of the engineers who were working on that project have now been shifted over into the group working on generative AI. And while one part of the story is an indictment of the state of the automobile industry in general, and particularly the challenges around electric vehicles right now, another part of the story is how much pressure there is to compete in this generative AI space. The market response to these reports was extremely positive and indeed, ever since it was reported, there's been even more pressure on Apple to actually get spec about what its AI plans are now. The reason that this is ahead of Nvidia on this list is that Apple moving full throatedly into the AI space, given the incredible install base of iPhones around the world, could be hugely significant in the shape of AI to come. At number six, we have another big tech company, Microsoft, partnering not this time with OpenAI, but instead open source leader and French startup Mistral. Now to be fair, Microsoft had started working with Mistral in some capacity last year, offering their previous open source models through Azure. This, however, is a more direct partnership and involves Mistral's first commercial offering that is not open source, Mistral Large being distributed through the Azure platform. This ranks high on the news list for a few reasons. The lowest of those is what it means for Microsoft's relationship with OpenAI. I think in general both of those companies view each other as very strong partners, but not exclusive partners, but this certainly does bring reasonable questions. The second reason that it was a big story is the EU's response to it. Mistral lobbied hard to water down certain provisions of the EU AI act when it came to foundation models. France and other countries like it argued hard, pointing to Mistral as an example of a company that couldn't really work in Europe if certain provisions were passed. Now some EU policymakers are calling foul, imagining that this deal was in the works the whole time, even as Mistral was talking about European solidarity when it came to AI models. Still, for me, the most important reason why this is a big piece of news is what it says about open source and the viability of open source. Mistral, to their credit, attacked this head on, basically saying they want to do as much open source as they can, but they also have to be a commercial company. That's going to be an interesting and tricky space to navigate, as I think this story reflects. And now we are heading to the top five of our list, Kicking it off at the beginning of the month With Google finally opening up Gemini Ultra at the end of last year, Google was under a ton of pressure to do something that caught up or even more hopefully got them back out ahead of OpenAI. That discussion was reaching a fever pitch, forcing Google, it feels like in some ways, to announce at the beginning of December their Gemini model. One of the big disappointments of that December announcement was that their most powerful model, the one that they claimed exceeded the performance of GPT4, wasn't at the time available. Instead, only a GPT 3.5 class model was. Well, that changed at the beginning of this month when Gemini Ultra finally started rolling out. And for a while, the main story of this month with Google and Gemini could have just been how Ultra actually compared to GPT4. Although as we will see, that's not really where the Google related stories ended. Number four and now we are moving into a category of stories which at one point this month I summed up as generative AI genuinely entering a new phase. And that was Grok Groq, no relation to Elon Musk's Grok Grok, which is a new hardware company that runs LLMs at a speed previously unimaginable. Grox inference engine produces up to 500 tokens a second, meaning near instantaneous comprehensive responses to queries of LLMs that frankly are so differentiatedly fast that it actually opens up different types of possibilities for use cases. True, no latency interactions in the form of conversations, for example, if this was combined with a voice model, would be available in a way that simply wasn't the case before. Because this company is relatively smaller, this news has flown more under the radar than some of these other stories. But if you're watching who's really at the edge of thinking about the state of the art in AI and the next important use cases of AI Grok, which you can test by the way, @groq.com was way, way high up on that list. At number three, we have a story which I have covered ad nauseam, and so we'll only go a little bit into here. And that, of course, was the extreme backlash around Google, Gemini's image creation and specifically its insert of diverse imagery and historical context where there was none. At first, one could be forgiven for thinking that this was just getting a lot of buzz because it was getting caught up in the American culture war, which tends to eat everything these days. However, I've argued that part of why it was so triggering wasn't just that, but that it gave us a moment of recognition of how powerful these tools could be in shaping people's perceptions of the past. It's scary to imagine a single company with its biases and priorities and prerogatives dictating what people know and think about our history. Ultimately, it was a big enough deal to wipe tens of billions of dollars off of Google's market cap to force a response from Google's senior leadership. And for right now, as we're recording this, there to be no human image creation available for Gemini. I think in the long run, all of the technology advancements that surround this story at the very top of our list will be more significant. But it's hard to deny how big a deal this was this month. And now we get to number two and number one. And you know, there are only two big stories left, two big technological advancement stories. And so you must be imagining, how am I going to rank order them? Well, at number two, I am putting Sora, but really one could make a very strong argument for Sora at number two or at number one. Sora is of course OpenAI's new video generation model. Instead of creating four second snippets, which is what we've seen from Runway and Pika, these videos were up to 60 seconds long. What's more, the quality is just something totally, totally different. Indeed, the videos were so remarkable that it triggered a response far beyond the AI space and bridging into the mainstream. TikTokers, YouTubers, everyone this month has been talking about the implications of having video generation this good, and frankly, a lot of that has been concerned. Concerned about deep fake nudes, concerned about political manipulation, concerned about a world in which the most significant communications medium, which is at this point video, has an AI generated version that is indistinguishable from reality. Now, in addition to provoking those sort of society level conversations, SORA has also majorly kicked up competition around AI video, with about a million new projects launching and getting attention in the weeks since this came out. Like I said, I think you could easily make an argument for this being the biggest story of the month, but for that I'm actually going to give it to Gemini 1.5 Pro, the surprise model that came out just a week after Gemini Ultra was released. You will almost assuredly already know that Gemini 1.5 Pro's big remarkable shift was a 1 million token context window that is enough for perspective to insert all three of the Lord of the Rings trilogy at once. I think as much as Google has been beleaguered by the Gemini Woke culture whatever conversation this month, the introduction of a model with a million token context window totally transforms what's possible with LLMs in ways that are going to be incredibly significant going forward. To me, I gave Gemini 1.5 the nod over Sora because one I think that at least in the short term, LLMs are going to be even more significant than video generation in terms of actual usage. But two for once Google actually gave us something that people could play with. Actual regular people like me have been able to get their hands on Gemini 1.5 Pro and actually start to use it. And so that was the number one story of the month. Once again, this was clearly a very wild month. Quite a kickoff to what looks like it's going to be another wild AI year. That, however, is going to do it. For today's AI breakdown. Until next time, peace.