Andy W
@aweissman
Over the past week or so I realized I have been thinking about LLMs incorrectly. I had considered them to be their own "things" that will let us do "stuff" Instead, LLMs are components of "things" that will "collectively" let us reimagine e stuff My colleague @mandel wrote about this last November (https://matthewmandel.com/2024/11/20/collaborative-intelligence/) but I didn't fully grok it until the other day (and then it lit my chaotic curiosity on fire)
2 replies
12 recasts
30 reactions
Andy W
@aweissman
For example, instead of thinking that a foundational model can practice medicine, what if there were a series of models that collectively act together to provide medical care. More specifically, you interface with an AI/model and tell it what's wrong and then - either in the background or transparent to you via chain of thought (your choice!) - a series of AIs/agents work together. To wit: the first model comes up with a differential diagnosis. A second model reviews and debates this. Maybe it sends it to a specialist model. Then another model intakes the data into a database to store that you can access later. And so on. All this can happen in seconds and, yes, eventually for pennies.
1 reply
1 recast
5 reactions
Andy W
@aweissman
Another example of a thing that takes a collective organization: making a movie. Imagine Director AI, coordinating with Cinematographer AI, and Lighting AI, and Sound AI, and then throw in wildcards like Contrarian Idea AI. The scale is endless, bounded only by, I guess, our imagination. I can think of many other examples too. And thinking about this is occupying my time - outside of music, books, art and movies :)
0 reply
0 recast
3 reactions
