OpenAI's LEAKED "Super Agents": The End of Human Work

So there was a recent article by Axios and it's pretty crazy. It talks about how we are going behind the curtain and coming soon there will be PhD level super agents. Now I want you guys to pay attention to exactly what's being said here. The key terminology being used, they don't just say agents are being created, they say super agents. And this is going to be one of the craziest articles you're probably ever going to see because the implications are truly profound. So it says right here that this is OpenAI's next major breakthrough. So it says architects of the leading generative AI models are abuzz that a top AI company, possibly OpenAI, in the coming weeks will announce a next level breakthrough that unleashes PhD level super agents to do complex human tasks. It says that we've learned that Sam Altman, the CEO of OpenAI, who in September dubbed this the Intelligence Age and is in Washington this weekend for the inauguration, has a scheduled closed door briefing for the U.S. government officials in Washington on January 30th. Now this is crazy because a closed door briefing is essentially a private meeting that's not open to the public or the media. It's basically like these government officials are having a super secret meeting that needs to be kept private. Now these types of meetings are common in government when officials need to discuss sensitive or confidential information. Now only specifically invited participants are allowed to attend. And in this case it means that Sam Altman would be privately meeting with government officials and he wouldn't be having a public hearing or press conference where, you know, news reporters could potentially attend or even listen in. Now what's crazy about all of this is that most people are currently speculating that this technology is so significant that they have to brief government officials ahead of time to basically give them a heads up on how society may change. And for historical context, we've seen this before. When ChatGPT was about to be released, there were similar briefings. When AI companies expect major capabilities breakthroughs, they often engage with the government first. And during the early days of social media, companies like Facebook had similar meetings about content moderation and election integrity. Now, what makes this one particularly stand out for us is the fact that the timing of this is really, really important. This is coming off the rumors after a major OpenAI breakthrough. And when we actually do take a look at the level of officials involved, it's in Washington D.C. with real government officials. And so whilst there might be a lot of speculation about whether or not these companies are completely hyping or larping having a scheduled closed door briefing with the US Government officials clearly states that something is going to be said in that meeting and that is of high importance. Now what's also very fascinating is that we do have the fact that Mark Zuckerberg and others have publicly talked about AI replacing mid level software engineers and other human jobs this year. Meta said that, you know, we are going to effectively have an AI that can be a sort of middle level engineer at your company that can write code. And this is what he said in.

The Joe Rogan interview, probably in 2025. We at Meta as well as the other companies that are basically working on this, are going to have an AI that can effectively be a.

Sort of.

Mid level engineer that you have at your company that can write code. And once you have that, then in the beginning it'll be really expensive to run. Then you can get it to be more efficient. And then over time we'll get to the point where a lot of the code in our apps and, and including the AI that we generate is actually going to be built by AI engineers instead of people engineers.

What was crazy was that the Salesforce CEO actually said something rather similar, talking about software engineers.

Salesforce, we're seriously debating maybe we aren't going to hire anybody this year because we have seen such incredible productivity gains because of the agents that work side by side with our engineers and making them more productive. And we can all agree that software engineering has become a lot more productive in the last two years with these basically these new models.

Now, crazily as well, we do have the fact that these AI systems have currently been exceeding projections. It says that our sources in US government and leading AI companies tell us that in recent months the leading companies have been exceeding projections in AI advancement. And I would say that this is probably accurate when we actually take a look at how quickly AI development has gone with the thinking models in the new paradigm. I would say that with all of the research that we've seen recently, including research papers, AI has definitely gone a lot faster. I think the search test time paradigm is definitely something that accelerated the area as a whole, because no longer are we relying on human data, we're relying on synthetic data and more neurosymbolic approaches that are akin to how humans learn and actually reason and solve problems. And this is where OpenAI last week released an economic blueprint arguing with the right rules and infrastructure investments, AI can catalyze a reindustrialization across the country. Now, of course, I do want to state that with every AI announcement, we always must distinguish the hype from the reality of said projects. And one of the things that this article actually talks about is the fact that AI investors have reason to hype small advancements as epic ones to juice valuations to help fund their ambitions. And this is pretty true because recently OpenAI actually did announce that they are going to need more money. You can see Here it says OpenAI says it needs more capital than we'd imagined as it lays out a for profit plan. In a blog Post On Friday, OpenAI laid out its plans for changing its corporate structure to allow for more capital and raising of funds. So we have to understand that whilst yes, AI is advancing rapidly, if OpenAI aren't very public about the kinds of advancements that they're making, and if they don't continually push the news that their software is going to do amazing and amazing things, then maybe investor sentiment might change and they may lose funding that they desperately need to keep the company afloat. I do think that this is something that is 50 50. Number one, they of course want to create hype around their product because that makes sense. And number two, of course they are going to need some investors to ensure that they can make it to the finish line of creating AGI. Now something that I have seen myself across the AI industry, mainly with these OpenAI employees and is that these employees are definitely spooked and jazzed by the recent progress. It says. Sources say that the coming advancement is significant. Several OpenAI staff have been telling friends that they are both jazzed and spooked by the recent progress. As we told you in the column on Saturday, Jake Sullivan, the outgoing White House National Security Advisor with security clearance for the nation's biggest secrets, believes that the next few years will determine whether AI advancements end in catastrophe. And this is where the article dives into super agents in more detail. It talks about how you can imagine a world where compet complex tasks aren't delegated to humans, but instead they're executed with the precision, the speed and creativity you'd expect from a PhD level professional. It says we're talking about super agents, AI tools designed to tackle messy, multi layered real world problems that human minds struggle to organize and conquer. They don't just respond to a single command, they pursue a goal. Super agents synthesize massive amounts of information, analyze options and deliver products. That's pretty different from traditional AI agents that we do know. Because even right now we could argue that AI agents are in a pretty rudimentary phase where they need a lot of scaffolding to achieve a lot of tasks. But we do know that OpenAI is usually at the frontier of when it comes to any kind of AI software, we've seen what they've been able to do with the voice, technology, real time video and vision. So having super agents is something that I do believe they'll be able to do because I've actually used Google's deep research, which I believe is on the back end. Just a bunch of research agents that are able to synthesize data from the web. And so far with that I've seen a tremendous amount of results and I do believe that that is going to be something that OpenAI is going to probably exceed. Now, when we do take a look at these super agents, we can take a look at what it says here. It says, build from scratch. Imagine telling your agent, build me a new payment software. The agent could design, test and deliver a functioning product, make sense of chaos. For financial analysis of potential investment, your agent could scour thousands of sources, evaluate risks, compile insights faster and better than a team of humans, and master logistics planning an off site retreat. The agent can handle scheduling, travel arrangements, handouts and more, down to booking a big dinner in a private room near the venue. Overall, we have three different scenarios that leverage the AI's capabilities to make decisions based on various pieces of information and come together, formulate a plan and deliver a result. And if this is true, if this is what super agents are going to be able to do, then it makes sense for OpenAI to be able to, you know, talk to the government to tell them that, look, the economy is rapidly going to change within the next 24 to 18 months and of course the government is going to need to prepare for that kind of change. Now, of course, this article does talk about the other tweets and the other side of this information and about how Noam Brown tweeted that there's lots of vague AI hype on social media, but there's good reasons to be optimistic about further progress. But plenty of unsolved research problems remain. And this is of course the fact that unfortunately we haven't managed to solve the hallucination problem in AI. And if that problem does persist in AI, then using these agents wide scale is going to be a very bad problem. Large issue when it comes to deployment. And remember, it was OpenAI that did say that they have begun to turn their aim beyond that to superintelligence in the true sense of the word. They love their current products, but they are here for the glorious future. With superintelligence, they can do pretty much everything. So this is going to be something that I personally do believe. OpenAI might not be overhyping in the sense that, you know, they're having conversations with the government. And, of course, we've seen the research papers that cause pretty much everyone to freak out at OpenAI. So with that being said, let me know what you think about these super agents, if you think that this is going to truly change the world and what you think OpenAI has in store for us.