An excellent analysis. Your clearly expressed thoughts reinforce and add to what I have been arguing for years. I often point back to the 1980s and 1990s "IT productivity paradox" in which greatly increased computing power for years and decades stubbornly refused to show up in clear productivity increases.
I think you are correct in emphasizing the "AI economy" rather than a single, all-power AI. In fact, I'm currently writing partly about that for the upcoming Beneficial AGI Summit. This alternative view of what is likely and feasible (given LLMs and not a different approach such as "cognitive AI" or "neurosymbolic AI") has strong implications for AI safety discussions. I think it further underlines the implausibility of catastrophic AI scenarios.
The Beneficial AI Summit is organized by (non-LLM) AI developer Ben Goertzel and colleagues. I participated in last year’s event in Panama City, and will be speaking at this year’s event in Istanbul. Information here:
The focus is on thinking about how to use AI for human benefit. There is plenty of discussion of risks and regulation but also plenty of more encouraging perspectives. Ben is working on a form of AI very different from current LLMs/LRMs, as is my friend Peter Voss:
As for what I’m writing: I have written a fair bit on AI and risks and benefits. I am currently writing a series on the AI Singularity idea from a critical and questioning perspective despite my generally optimistic expectations. The first two parts are out:
An excellent analysis. Your clearly expressed thoughts reinforce and add to what I have been arguing for years. I often point back to the 1980s and 1990s "IT productivity paradox" in which greatly increased computing power for years and decades stubbornly refused to show up in clear productivity increases.
I think you are correct in emphasizing the "AI economy" rather than a single, all-power AI. In fact, I'm currently writing partly about that for the upcoming Beneficial AGI Summit. This alternative view of what is likely and feasible (given LLMs and not a different approach such as "cognitive AI" or "neurosymbolic AI") has strong implications for AI safety discussions. I think it further underlines the implausibility of catastrophic AI scenarios.
Thanks very much.
What is the Beneficial AGI Summit? What are you writing?
The Beneficial AI Summit is organized by (non-LLM) AI developer Ben Goertzel and colleagues. I participated in last year’s event in Panama City, and will be speaking at this year’s event in Istanbul. Information here:
https://bgisummit.io
The focus is on thinking about how to use AI for human benefit. There is plenty of discussion of risks and regulation but also plenty of more encouraging perspectives. Ben is working on a form of AI very different from current LLMs/LRMs, as is my friend Peter Voss:
https://petervoss.substack.com/
As for what I’m writing: I have written a fair bit on AI and risks and benefits. I am currently writing a series on the AI Singularity idea from a critical and questioning perspective despite my generally optimistic expectations. The first two parts are out:
https://maxmore.substack.com/p/putting-brakes-on-the-singularity
https://maxmore.substack.com/p/the-non-singular-singularity
Thanks. I subscribed to your substack. Looks really interesting. I share your skepticism on the possibility of the AI apocalypse.