Has OpenAI Really Achieved AGI with o3? Unpacking the Hype and Reality

[@jeremiahg7925]

In my opinion, it needs a conceptual awareness of the world to be AGI. The ability to form and organize concepts, learn how they interact with eachother, predict the outcome of actions, and weigh competing goals for decision-making. Not just solve certain tasks. It needs a feedback loop of the key components of consciousness.

[@couchtaming23]

If o models can accelerate scientific research by 1000x in the future, the distinction between AGI and non-AGI becomes less relevant. Consider the progression from AlphaGo to AlphaZero, and then to AlphaFold and AlphaFold2 as a prime example of how targeted advancements in AI can drive extraordinary breakthroughs.

[@ANDMAX01]

Thank you for the common sense and explaining the tests and metrics, theres so much that is just a given by lots of people who talk about the subject, and that people who want to be informed are not familiar with so THANK YOU

[@JeredBlu]

Thank you for the feedback! Thats exactly what I'm trying to do - break down these complex topics in a way that makes sense for everyone. There's often a lot of assumed knowledge in tech discussions, and I believe we all benefit from clear, straightforward explanations. Really appreciate you watching and sharing your thoughts!

[@drormarkus7539]

This is really interesting - AGI as a concept makes us rethink what intelligence really means. The fact that there's no single definition might actually be a good thing, showing how complex and varied intelligence can be. So maybe instead of aiming for one big AGI, we should think about it as a range of different abilities, kind of like how people's smarts work in different ways....

[@ToolmakerOneNewsletter]

You had the story correct until the end, and then you failed to look at the way the Arc Prize questions are generated. They are unique for each test suite. So yes, it passed their definition for AGI, but notice how the representative for that prize was careful with Sam sitting beside him to NOT make an AGI proclamation. I'm sure he was "briefed" before the presentation, just like LLM's are "briefed" before release to the public.

[@razgaon3680]

if i understood correctly, he was referring to the dozens of twitter influencers who claimed o3 is AGI without ever talking to the model

[@JeredBlu]

@@razgaon3680 This was indeed my point!

[@amesasw]

At this point I don't even agree that working on AGI is school. All of these CEOs are getting ultra excited by attempting permanently eliminating knowledge based labor. They hype up the ability to build, but 100% downplay the real goal of eliminating all human labor with minimal thought or speculation of how to rectify a society where their AI can accomplish that. To be fair, I think Sam Altman has at least said the words UBI. But that is not a realistic way to run a society. If real AGI or ASI were achieved it should 100% be owned by the public who's information and work built it. That's assuming you can own a computer that is as smart as a human.

[@razgaon3680]

This is a great video! There really is fake everywhere

[@JeredBlu]

Thanks!

[@johannesdolch]

OpenAI has till mid 2025 to become the company to create real AGI or they are toast. They are getting outscaled by xAi on every level.

[@JeredBlu]

Interesting, I'm curious what you have found xAi does better than OpenAi?

[@EelkodeVos]

You are wrong, I'm afraid. You, like many others, have a tendency to antropomorphize AGI. The term AGI does not encompass human intelligence. These entities are not physical, are not raised by parents or a community, they have no "consciousness" like we have. So you should not raise the bar into a human cognition environment. That is like judging a spider to swim like a dolphin or a human to perform like a snail. What has been reached here is a system that can think about a question that it has not been trained on, and reach the correct answer. That, in essence, is AGI. All other factors that you stamp on AGI, are mostly human traits people expect these entities to have. Which they don't and perhaps never will. These systems are moralless unaligned psychopaths that will pick up any task you give them, regardless of their training. Once you factor that in to an AGI, you get an Artificial General Intelligent and Empathic system, and AGIE. We are not even close to having that.

[@UmeshBhardwaj-p5l]

Bro wow super you’r correct It’ a new hype bubble

[@UmeshBhardwaj-p5l]

We will never reach human intelligence in Ai because they don’t have consciousness

[@googm]

Thinking is also an anthropomorphization. In any case it's neither thinking nor reaching the correct answer on questions it hasn't seen (e.g., ARC-AGI 2).