The AI Daily Brief: Artificial Intelligence News and Analysis - Is OpenAI Actually Going to Be Open Again?

Episode Date: April 2, 2025

OpenAI just raised the largest private funding round in history at a whopping $40 billion valuation. But the bigger news? They're returning to open source for the first time since GPT-2, promising an open-weight language model that anyone can run on their own hardware. Interested in the Disruption Incubator?Email [e-mail] Brought to you by:KPMG – Go to ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://kpmg.com/ai⁠⁠⁠⁠⁠ to learn more about how KPMG can help you drive value with our AI solutions.Vanta - Simplify compliance - ⁠⁠⁠⁠⁠⁠⁠https://vanta.com/nlwThe Agent Readiness Audit from Superintelligent - Go to https://besuper.ai/ to request your company's agent readiness score.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/1680633614Subscribe to the newsletter: https://aidailybrief.beehiiv.com/Join our Discord: https://bit.ly/aibreakdown

Today on the AI Daily Brief, is Open AI about to become open again? Before then on the headlines, a new, very impressive seeming video generation model. 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. After a week of discussing image generation, we are back over on the video side of the AI house as runway has announced their new Gen 4 model. The company claims it's one of the highest fidelity video generation model. to date, capable of incredible consistency in characters, locations, and objects across scenes,

what the company is calling world consistency. In a blog post, they wrote, Gen 4 can utilize visual references combined with instructions to create new images and videos using consistent styles, subjects, locations, and more, without the need for fine-tuning or additional training. Now, obviously, if this holds true, it completely changes the dynamics of what you can do as a filmmaker with Runway. And by filmmaker, I mean this in the new AI-enabled version of that term, which does not just mean people who make movies, but anyone who wants to use film or video as a creative medium.

The examples they give of this character consistency are pretty spectacular. Even as someone who spends everyday looking at this, I don't know that I would have identified this as AI if I had been presented with it in a different context. The other part of character consistency is that it allows for more dynamism when it comes to getting different angles on shots, giving people much more creative expression. Part of the upgrade and capabilities is that it's easier to use because this new Gen 4 model apparently understands language much better. And lastly, the company argues that they made serious

strides in what is, of course, the Achilles heel for AI-generated video, which is real-world physics. They use fire, wind, water, shadows and light to show the advances they've made in this domain. Overall, the model seems incredible. The big question remains whether standalone media generators will survive the next generation of LLM improvements. OpenAI and Google are pushing native multimodal architecture to add features like image generation to their standard LLMs natively, and you got to think that the same thing is coming for video models. Still, when it comes to the here and now, as the old saying goes,

a model available today is worth two on the roadmap, and a ton of people are excited to dig into Gen 4 right now. Runway is also reportedly looking to raise $450 million at a $4 billion valuation, so that question of competition obviously has some big financial implications. Next up, viral Chinese agent, Manus AI, has launched a fairly pricey subscription plan. The tool, of course, went viral last month, offering a state-of-the-art agentic assistant that could consistently execute tasks. The big constraint on the virality was Manus having to gate sign-ups through an invitation system, most likely because they tapped out their compute

resources. The new commercial release includes two tiers. One is $39 a month and one is $199 a month, with the more expensive option coming with five times as many credits, and the claim that they can execute five tasks simultaneously. There is still free access, but it's going to have lower priority access to resources during peak usage times, which could honestly be all the time at this point. Manus is also now available as an iOS app. Now, Manus AI is still in beta, and there's still some question of whether it lives up to the initial hype, but there is no denying that it's captured a lot of attention, and the broader availability that this represents should help it get into more people's hands, both to see how good it actually is and to help it improve.

Lastly today, AI Alexa is finally here, and they are taking the Apple rollout path, by which, of course, I mean that they've shipped the product without all the features. Amazon showcased Alexa Plus at a special unveiling event in late February. They showed videos of Alexa ordering takeout on Grubhub, generating stories to entertain kids, visually identifying people and reminding them to do chores and brainstorm gift ideas. If you had to guess how many of those features were shipping on release, what would you say? I'll wait a minute to let you guess. That's right, for those of you who said zero you are currently.

Correct. Amazon said that these, quote, agentic features don't yet meet Amazon standards for public release. So what can Alexa Plus do right now? It can order an Uber. It can identify objects. It can draft emails and it can search for particular products. And Amazon spokesperson said, we're releasing a bunch of features to start and will continue to launch new features in waves. Basically, the new AI-powered versions of Alexa and Siri find themselves in good company, completely under-delivering on what seems like obvious promises. Anyways, friends, that is going to do it for today's AI Daily Brief. Headlines Edition, next up the main episode.

Today's episode is brought to you by Vanta. Trust isn't just earned, it's demanded. Whether you're a startup founder navigating your first audit or a seasoned security professional scaling your GRC program, proving your commitment to security has never been more critical or more complex. That's where Vanta comes in. Businesses use Vanta to establish trust by automating compliance needs across over 35 frameworks like SOC2 and ISO-2.

centralized security workflows, complete questionnaires up to 5x faster, and proactively manage vendor risk. Vanta can help you start or scale up your security program by connecting you with auditors and experts to conduct your audit and set up your security program quickly. Plus, with automation and AI throughout the platform, Vanta gives you time back, so you can focus on building your company. Join over 9,000 global companies like Atlassian, Kora, and Factory who use Vanta to manage risk and prove security in real time. For a limited time, this audience gets $1,000 off Vanta at vanta.com slash NLW. That's VANTA.com slash NLW for $1,000 off. Hey, listeners, are you tasked with the safe deployment and use of trustworthy AI? KPMG has a first of its kind AI Risk and Controls Guide,

which provides a structured approach for organizations to begin identifying AI risks and design controls to mitigate threats. What makes KPMG's AI Risks and Controls Guide different is that it outlines practical control considerations to help businesses manage risks and accelerate value. To learn more, go to www.kpmg.org.us slash AI Guide. That's www.kpmg.comg.comg slash AI Guide. Today's episode is brought to you by Super Intelligent and more specifically Super's agent readiness audits. If you've been listening for a while, you have probably heard me talk about this. But basically the idea of the agent readiness audit is that this is a system that we've created

to help you benchmark and map opportunities in your organizations where agents could specifically help you solve your problems, create new opportunities in a way that, again, is completely customized to you. When you do one of these audits, what you're going to do. to do is a voice-based agent interview where we work with some number of your leadership and employees to map what's going on inside the organization and to figure out where you are in your agent journey. That's going to produce an agent readiness score that comes with a deep set of explanations, strength, weaknesses, key findings, and of course a set of very specific recommendations

that then we have the ability to help you go find the right partners to actually fulfill. So if you are looking for a way to jumpstart your agent strategy, send us an email at agent at B-Supert.aI, and let's get you plugged into the agentic era. Welcome back to the AI Daily Brief. A lot of big news out of OpenAI yesterday, all of which we're going to dig into today. The headliner is that they've closed what seems like the biggest private fundraising round in history, raising $40 billion at a valuation of $300 billion, basically doubling their valuation from about six months ago.

but still that is definitely not the story that has everyone talking. OpenAI has committed to releasing their first open weights model since GPT2. Here's what Sam Altman had to say about this. We're planning to release our first open weight language model since GPT2. We've been thinking about this for a long time, but other priorities took precedence. Now it feels important to do. Before release, we will evaluate this model according to our preparedness framework, like we would with any other model, and we will do extra work given that we know this model will be modified

post-release. We still have some decisions to make, so we're hosting developer events to gather feedback and later play with early prototypes. We're excited to see what developers built and how large companies and governments use it where they prefer to run a model themselves. Now, in terms of what exactly they mean by open weights, the company's head of API, Stephen Hydel, wrote that this meant a model that you can quote run on your own hardware. To me, this implies that the company is not just going to release the weights for old deprecated models like GPT3. Instead, it sounds like either a modified version of 01 or 03, or something that's closer to the state of the art, or more likely an entirely new model that's optimized to run on consumer-grade hardware. Given how much

OpenAI has staked their claim on reasoning models, and given how important reasoning is for driving agents and all other manner of advanced tasks like coding, it could also imply that Open AI is looking to compete on that category of open model as well. Now, obviously one of the running jokes with OpenAI, especially propagated by competitors like Elon Musk, is that the company is far from Open. And indeed, it seemed at some point along the way, Open AI had become concerned from a safety standpoint or perhaps had evolved their thinking for some other reason to move away from open models that people could actually build and modify. However, it's clear that there's been an evolution in this more recently. In a Reddit AMA in January, Sam Altman wrote, I personally think we need

to figure out a different open source strategy. Not everyone at OpenAI shares this view, and it's also not our current highest priority. We will produce better models going forward, but we will maintain less of a lead than we did in previous years. So for those who have been reading the tea leaves, this doesn't seem completely out of left field. More importantly, there is very clearly a genesis in a motivator for this that comes from outside the company, and that is, of course, Deepseek. While OpenAI and Altman didn't mention Deepseek in any of these announcements, Altman did share that change of hard on Reddit, a little over a week after the release of Deepseek R1, just as its viral moment was hitting. Since then, Deepseek's impact in China has been palpable.

We've seen the model integrated into online services. from each of that country's four big tech firms, who have also released their own cutting-edge models in just the last few short months. Keep in mind that Deepseek didn't just open-source their model. They also open-sourced a huge range of training optimizations that allowed them to build it. It is not unreasonable to view what has been happening with Deepseek in China as nationwide technology transfer, with basically all Chinese firms leveraging each other's breakthroughs to drive AI adoption. Singaporean poker player Wayne Yap posted his impressions on tech in China after a recent visit. Regarding AI adoption, he wrote,

Deepseek is integrated in Baidu Maps. Baidu Maps is the big Maps player in China, the equivalent of Google Maps. It was shocking but pleasant to see that when I was searching for food, I could just press Ask Deepseek and it would recommend me places to go. Again, point being that the interconnection of these services is happening very, very quickly, and also that that's enabled not by an extra active partnerships department on the part of Deepseek, but simply by virtue of the way their models have been released. Last week, there was a discussion on X about open source AI becoming the official position

for the Chinese government. Investor Balaghi Shrenovasen commented, I agree that it's surprising that the country of the Great Firewall is suddenly the country of open source AI, but it's consistent in a different way, which is that China is focused on doing whatever it takes to win, even to the point of copying partially abandoned Western values like open source, which seemed like the hardest thing to adopt. The broader discussion has, of course, been about China seeming to adopt a strategy of driving down the cost of AI as much as possible, and then exporting it to the world to out-compete Western firms, basically a new form of AI Belt and Road. In comments to the Financial Times, Zhang Yongian of the Chinese

University of Hong Kong coined the term open source modernization. He said, Western modernization is very exclusive. The West doesn't help other countries, poor countries to develop. Chinese modernization, I call it open source modernization. When you get rich, you help other countries to get rich. The point of all of this and why we're talking about it in the context of an open AI announcement is that deep-seek and Chinese AI in general are using open source as a way to push distribution as hard as possible. And it seems fairly clear that they're not just going to be competing for the Chinese market for long. The really interesting question is whether OpenAI is leading into open source to compete on that vector. Now, on the one hand, it doesn't seem strictly necessary given their insane

growth rate. In fact, also yesterday, Altman tweeted, the chat GPT launched 26 months ago was one of the craziest viral moments I'd ever seen, and we added one million users in five days. We added one million users in the last hour. Now, the ongoing giblification of everything is the obvious explanation, but the point remains that OpenAI doesn't appear, at least from the outside, to need to embrace open source to drive adoption. ChatjPT is basically already one of, if not the fastest adoption curve in the history of tech. Then again, if Albin considers Open AI to be more than just another tech firm, and if the ambition is more than simply making a boatload of money, if this is instead about a global AI competition, one that involves a competition for VALBORM.

then all of a sudden the open source shift makes a ton of sense. At the moment, open AI's models cannot be deployed absolutely everywhere like deepseeks can. Beyond just licensing costs, OpenAI is resource constrained. By offloading inference to local deployments using open source, OpenAI could decide to compete to become the de facto AI choice for the US and even the world. Altman even hinted at the idea that he wants this new open source model deployed in as many commercial applications as possible. Writing in a not-so-suttle knock-on-metta's licensing model, we will not do anything silly like saying that you can't use our open model if your service has more than 700 million monthly active users. We want everyone to use it. Like I said, that's a knock on meta,

which restricts the largest companies from using Lama models if they reach that size. Overall, what we're seeing here is potentially a very, very big strategic shift for OpenAI with very, very big implications, not just for the AI battle, but for the entire world. And the company is definitely going to be well-resourced as they make that shift. As I mentioned at the top of the show, OpenAI has officially closed their latest funding round, raising $40 billion at a $300 billion valuation. This is the largest private funding round for a tech company in history, beating out the $10 billion raised by Databricks late last year. In a very short blog post, OpenAI framed the round as allowing them to build towards AGI. They wrote that the funding will enable them to push the frontiers of AI research even further,

scale our compute infrastructure, and deliver increasingly powerful tools for the 500 million people who use ChatGPT every week. To my knowledge, that 500 million number is the first time that has been shared, so yes, a full half billion people now are using this tool every single week. Still, the fine print of the deal is extremely noteworthy, significantly raising the stakes, among other things, for OpenAI's for profit conversion. The deal is being led by SoftBank who are investing $30 billion. The $40 billion overall is split into two halves, with $10 billion to be received up front, and a further $30 billion to arrive by the end of the year.

The second payment is partially contingent on OpenAI completing their for-profit conversion this year. If they fail, SoftBank is allowed to cut their contribution by $10 billion. Further Venture Beat reports that $18 billion of the funding is earmarked for Project Stargate, so won't go towards supporting OpenAI's normal operations. These types of deal terms aren't necessarily super foreign for Open AI. For example, they had already committed to return the $6.6 billion they raised last fall to investors if they didn't complete the for-profit conversion within two years. Altman had already implied that the conversion was existential given their need to fund

costly infrastructure and training, and so honestly, what's another $10 billion among friends? at risk if that for-profit conversion doesn't go through. One more quick one today, another example of how OpenAI is thinking about potentially expanding its footprint more broadly. The company has announced that their OpenAI Academy is now live. The initiative is a free online resource hub to help, as they put it, support AI literacy, and help people from all backgrounds, access tools, best practices, and peer insights to use AI more effectively and responsibly. Friend of the show, an OpenAI general manager of education, Leobelsky, wrote,

When I first joined OpenAI, I kept circling one question. What's the best way to teach the world how to use AI at scale in real time as the technology keeps evolving? Coming from Coursera, I've seen firsthand how powerful online learning can be. But this moment is different. The tools are more powerful, the pace is faster, and the opportunity is much, much bigger. Do you build a next gen online course platform? Do you meet learners through the education ecosystem already in motion? Do you turn chat chachy-t the tool itself into the teacher? Or maybe it's some of all of that. OpenAI Academy is our first step. Bight-sized tutorial from chat. GPD to campus to store a video creation. Partner-led in-person workshops, build hours, and global

content collapse. Learning communities on the way built around students, teachers, and small businesses. The goal, make AI literacy accessible, practical, and global. Let the education ignite learners versus running them through long courses in phase one. This is just the beginning. This week we have another announcement coming, which takes a very different type of crack at driving AI literacy. Now, although superintelligent has evolved, now focused on helping companies audit their agent opportunities and then finding the right partners through a marketplace to actually deliver on those agent opportunities, a lot of the themes that Leah and OpenAI are exploring are exactly where we started. You can go and check it out now at academy.openAI.com, and I'm excited to hear

what you think. 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.