Oh no, sorry about that, I guess it depends on some cookie context. Does this link work? Could someone with the ability change the content of the link on the post, or should I just make a new post?
Oh no, sorry about that, I guess it depends on some cookie context. Does this link work?
Could someone with the ability change the content of the link on the post, or should I just make a new post?
Following some interesting discussion of a retrospective on Douglas Lenat's life from the perhaps somewhat tendentious Stephen Wolfram, I thought it might be interesting to some folk here to...
This piece communicates the central thread of his life's work. The overall thrust of it is that heuristics, or generalisations, are critical for humans to solve tasks that involve intelligence, supporting efficient search in very large solution spaces. Therefore, we should be focusing on implementing heuristics and capability for generalisation in AI systems, to better implement intelligent approaches to problem solving, i.e. supporting efficient reduction in search space.
I think it's quite a classic academic AI 'thinkpiece' of its time. Clearly written, idealistic, with some sparkle, really grappling with very fundamental concepts, still working first-hand through the implications of Turing. It can border on specious, arbitrary, and over-simplified, as these kinds of computationally focused explorations of human intelligence so often can. I think that the author is more aware than average that his model of human problem solving is a 'cast' or a 'narrow view.'
I like that he uses examples from a range of scientific fields, and as a piece of history it's great to see where those fields were at that time.
Points for discussion:
What do you think about the idea of human intelligence being cast as "very efficient searching of a priori immense space?" What the benefits and limitations of this model?
How can we interpret these views in the context of the modern AI environment? Can LLMs be comprehensively understood as agents that efficiently reduce search space through heuristics?
What were the fundamental limitations of the kinds of expert systems he describes in the latter part of the article, as well as the wider set of AI approaches of the time, that meant they could not efficiently achieve this vision?
That they could not achieve what LLMs can?
What is missing? Does it speak to something other than heuristic search, or is it something in the approach to implementation of heuristic knowledge, or its use in solving problems?
Having read Lenat's writing itself, does this affect your opinion on Stephen Wolfram's article on him?
How do the observations on human intelligence and behaviour hold up to modern (or even contemporary) understandings of human intelligence? Does the author suffer from a kind of "computer science brain" understanding of intelligence that misrepresents what real evidence shows?
The link provided gives an Access Denied XML response to me. Is there something additional needed to view it?
Oh no, sorry about that, I guess it depends on some cookie context. Does this link work?
Could someone with the ability change the content of the link on the post, or should I just make a new post?
Substituted your new link in.
Following some interesting discussion of a retrospective on Douglas Lenat's life from the perhaps somewhat tendentious Stephen Wolfram, I thought it might be interesting to some folk here to investigate Lenat's work through his own words.
This piece communicates the central thread of his life's work. The overall thrust of it is that heuristics, or generalisations, are critical for humans to solve tasks that involve intelligence, supporting efficient search in very large solution spaces. Therefore, we should be focusing on implementing heuristics and capability for generalisation in AI systems, to better implement intelligent approaches to problem solving, i.e. supporting efficient reduction in search space.
I think it's quite a classic academic AI 'thinkpiece' of its time. Clearly written, idealistic, with some sparkle, really grappling with very fundamental concepts, still working first-hand through the implications of Turing. It can border on specious, arbitrary, and over-simplified, as these kinds of computationally focused explorations of human intelligence so often can. I think that the author is more aware than average that his model of human problem solving is a 'cast' or a 'narrow view.'
I like that he uses examples from a range of scientific fields, and as a piece of history it's great to see where those fields were at that time.
Points for discussion: