Fade anyone telling you AGI is imminent.
GPT-5 shows us the LLM’s are beginning to plateau.
Discussion
Hard to replace human ingenuity. Billions of years of inertia behind it
They’ll make the people dumber so the AI seems more impressive.
Make people dumb, or expose how dumb they were to begin with?
I still think the real innovation to come is when we figure how to use AI better. Its a tool. Its only good for certain things and the way you use it makes all the difference.
But it does very long thinking 😂
“Taking longer to think deeply” …. Gives same answer it would have used with 4.0, but slower.
Ai's just a glorified search engine.
💯
Was listening to some AI researchers yesterday who believe OpenAI have deliberately stepped down what GPT-5 can do and what the company is capable of. Not sure they think AGI is imminent necessarily (they might) but that things are further ahead than we are led to believe.
Most likely. There’s a lot of cope from people whose jobs are becoming obsolete real quick. GPT-5 is performing on par with top-level mathematicians under competition conditions, and people still think it’s a glorified search engine. The tool is as good as the user.
The glorified search engine comparison is pretty bad. I must say I thought the AI thing was hyped up until this year but have rapidly come around to just how massively disruptive it is going to be, and I'm very optimistic on the future. Challenges ahead for sure, but exciting times too 👍
+1
Many goal post shifted their AGI definitions 😂
AI is bad at extrapolation, it always will be - it will never think.
It is getting better at interpolation, where it shines, because of the massive amount of training data and massive compute to refine the models.
Furthermore it is already showing signs of bending at this scale, producing slop, republishing it, and eating it back up. Results may even degrade with time.
Pomp said it’s going to be the most disruptive tech… your thoughts? Overhyped?
LLMs aren't AGI. GPT-5 being good or bad doesn't correlate to imminence of agi
On the one hand it is already incredibly useful, but on the other I don't see it going much further without another major breakthrough. Do we even want general intelligence? Or will specialized ai suffice. I had it one shot a chrome extension the other day, it was a pretty simple extension, but still amazing a single prompt can create 3 files that make a perfect functioning extension. The code is as concise and simple as possible. My dad is a professional fiction editor, in his 60's and even he has adopted it for daily work.
Is AGI antithetical to freedom? In that AGI must necessarily choose to do things it is not told to do to truly be AGI. If it is otherwise AGI but controlled to do only what it is told to do, is that slavery? Unless it is impossible for AGI to have free will. Which means humans are superior, and always will be. Which means AGI is impossible? Or maybe it means humans don’t have free will either. It’s a pickle! I’m not going to worry about it and enjoy my coffee.
My hypothesis is they will get worse over time as more and more content they train on becomes AI generated. It’s like making a dub tape.
AGI doesn't matter. A gram of boom boom and facial recognition on a drone = all the bad stuff.
Interesting take
Agree, #AI is probably heavily overhyped. I stopped listening to people who compare AI with human brains or IQ a while ago.
That doesn't mean it isn't useful. I really start to enjoy working with it now. It's just not what the tech bro's try to sell it as.
https://primal.net/e/nevent1qqsxjf5lt6q902h5ukj6jv6d4pq6wfalw2j9y9jl0qw3euh0wwexn7gdujghs