Maybe world scope six would be learning how interacting with humans influences how they in turn interact with each other or their environment. Or is that encompassed by world scope five?
That is an excellent explanation. You are the best. Thanks. At what point information becomes knowledge? That's the central question. Is it possible that some organization puts way too much money into building a machine that can understand math & information? Then the question is can the AI device create new knowledge? If that is possible, then the 5+ steps described in the paper are significant, but the progress of AI could move unpredictably. What do you think?
This was great explainer! Does WS5 include AI AI socialisation? If not I'd offer that as WS6. You can have an awful lot of more interactions without having to rely on humans in the loop. Tho things would get very strange shortly after.
I believe this is a human-centric framework for learning human language. There's a bunch of RL work on communication language that emerges between agents (that is not necessarily human-understandable).
I would have included the LM "toys" - calculator, python repl, sims, search engine or external memory modules. Basically they are all simulators - of math, code, 3d or contain an extended dataset for retrieval. Probably WS6 will be a society of AI agents, this time AI interacting with AI in order to expand experience even more.
I wrote the "BULL" on a wooden sign on a fence as a narrative hypothetical for ChatGPT. I asked about intention, outcomes and what the human character would infer. ChatGPT passed with flying colors, both the potential dangers in detail and the reasoning behind hanging such a sign.
You are the only person in the world who deal with transformers like a baby.. Please keep doing..
Very Interesting Videos and Paper. Hope will get to hear more about the progress done in WS4 and WS5 in the upcoming videos.
Thanks
Maybe world scope six would be learning how interacting with humans influences how they in turn interact with each other or their environment. Or is that encompassed by world scope five?
That is an excellent explanation. You are the best. Thanks.
At what point information becomes knowledge? That's the central question.
Is it possible that some organization puts way too much money into building a machine that can understand math & information? Then the question is can the AI device create new knowledge? If that is possible, then the 5+ steps described in the paper are significant, but the progress of AI could move unpredictably. What do you think?
This was great explainer!
Does WS5 include AI AI socialisation? If not I'd offer that as WS6. You can have an awful lot of more interactions without having to rely on humans in the loop. Tho things would get very strange shortly after.
I believe this is a human-centric framework for learning human language. There's a bunch of RL work on communication language that emerges between agents (that is not necessarily human-understandable).
Great content... Keep it up
I would have included the LM "toys" - calculator, python repl, sims, search engine or external memory modules. Basically they are all simulators - of math, code, 3d or contain an extended dataset for retrieval. Probably WS6 will be a society of AI agents, this time AI interacting with AI in order to expand experience even more.
I wrote the "BULL" on a wooden sign on a fence as a narrative hypothetical for ChatGPT. I asked about intention, outcomes and what the human character would infer. ChatGPT passed with flying colors, both the potential dangers in detail and the reasoning behind hanging such a sign.
That's a beautiful test!
✨✨✨✨✨✨✨✨
We need LLMs to train on historic data, philosophical and god-related data. In general, feed books to models, not only the web.
yoo that’s my professors paper
First..