Stephen Wolfram hosts an unscripted Ask Me Anything about the future of science and technology for all ages.
00:00 Start stream
1:57 SW starts talking
2:08 What will “doing science” look like in 20 or even 50 years compared to today?
38:08 Will future scientific knowledge be stored as papers, simulations, executable models or something totally new?
42:28 How do we decide what science is worth pursuing when almost anything becomes possible?
46:21 From your perspective, is this a good way to learn programming so the knowledge is stored in long-term memory and remains usable years later instead of being forgotten after short-term practice? And how did you personally figure out how you learn best when mastering complex subjects?
49:44 Learning and interacting with an LLM
52:39 In the context of LLMs, how important is to provide access to curated DB or labels of answers on particular topics like CS, Algos, etc. Is this something that Wolfram aims for?
54:38 How do we prevent students from outsourcing all thinking to AI without banning it outright? Should schools explicitly teach students how to work with LLMs, not just restrict them?
58:36 With LLM prompting you have to be critical of every answer and play your own devils advocate
59:48 I’ve seen so many examples of companies trying to use LLMs as search engines, and I’ve never seen it turn out well. Is it possible for LLMs to reliably retrieve factual information, or is it a fool’s errand?
1:02:03 Do you think there’s an opportunity for development of an educational LLM, i.e. a dedicated system specifically designed to aid discovery and understanding as opposed to simply providing an “answer”?
1:04:42 What’s a significant scientific discovery that you’d like to see being made during this year?
1:07:53 What do you think about the concept of technological singularity?
1:12:12 What if LLM were trained not to guess the next words, but to guess the previous words? that way you could give it an end state “and that is how we discovered the…”, is that harder?
1:14:33 Will we ever have cybernetically enhanced memory? Would this be a great leveller for people with otherwise similar intelligence?
1:25:37 End stream
Credit to : Wolfram
