Will GPT Agents Reach AGI?
Can we just calm down a little? Actionable A.I. is coming but in its own time at its own speed.
I’ve been a bit blindsided by the hype around AutoGPT and GPT Agents, and after the ChatGPT wave, I thought this would settle down, I think I was wrong. Silicon Valley seems determined to spin the story that AGI is just around the corner.
Let’s try to think about this a little. It’s a lot of smoke and mirrors.
The best guide to autonomous agents might be this one:
Is this an open-source movement or an automation movement or something else that could unlock the capabilities of LLMs even further? Recent developments involving Auto-GPT and BabyAGI have demonstrated the impressive potential of autonomous agents, generating considerable enthusiasm within the AI research and software development spheres.
If GPT as a foundational model was being compared to a General Purpose Technology by serious emerging tech pundits, perhaps they are missing the point and they are really General Autonomous Disruptors (GADs ?).
But I think we are all getting way ahead of ourselves. I’m on the Yann LeCun side of the fence with regards to LLMs even not being capable of reaching AGI.
Christof Koch: AGI is ill defined because we don’t know how to define intelligence. Because we don’t understand it. Intelligence, most broadly defined, is sort of the ability to behave in complex environments that have multitudes of different events occurring at a multitude of different timescales, and successfully learning and thriving in such environments.
Certainly with a handful of companies, giving LLMs the power to act onlin autonomous is half the battle. Eric Newcomer, who seems preoccupied by A.I. recently in Venture Capital trends, put it like this recently in a LinkedIn post:
I think it’s actually going to get a lot more complicated as good as these startups and companies are.
It’s highly doubtful to me that AutoGPT is one more step towards AGI as it’s being out to be made. Read my GPT Agents category on our Subreddit. It’s going to take years for this stuff to churn out real-world disruption in tasks at scale across industries.
AI is certainly boosting productivity and getting better at automating certain tasks. But AGI is another league of self-learning and common sense in real world and digital contexts across systems that we aren’t seeing any signs of yet.
Let’s dive into it for a bit more context:
Keep reading with a 7-day free trial