DeepSeek was funded by a quant billionaire. Another Chinese billionaire funding another foundational project at the intersection of human memory and AI.
My background is in psychology and I have an interest in the interface between our humanity and our technology. The research in developing AI “memory” is exciting in that it is forcing us to look deeply into the processes which form what we call “consciousness” and examine many of the misconceptions and cultural perceptions we have accumulated over eons. This new research is forcing us to re-examine our concept of consciousness and to scientifically develop a roadmap of all the processes that somehow are sorted and combined to form a consensus we call consciousness.
Thomas Moore wrote a series of books on “Care of the Soul”. What I find lacking in AI is it’s stark “soulessmess.” Moore defines “soul” as the messy interface between our powers of reasoning and of sensing and experiencing. I remember the old computer term “fuzzy logic” which were early attempts to bridge the gap between machine logic and the ‘hot mess’ of human consciousness. It seems that dynamic is now being fully explored.
Can our machines help us to ironically become more human and less machine-like? Jacob Bronowski wrote a book and hosted an enlightening BBC series, “The Ascent of Man”, documenting how advances in technology have vastly altered human psychology and sociology. We have evolved to become more “machine-like” as mastering new technologies gave individuals and societies power and wealth. It seems in the process, we have sacrificed our humanity for the power of technology resulting in a plague of diseases of despair - depression, drug addiction, and suicide. Can we relearn from our technology as it better mimics our humanity to again honor and respect the wonder of our illogical humanity? Or, will we plunge ourselves into even deeper collective despair?
Robert, I'll encourage you to read Gerd Gigerenzer's excellent book, "How to Stay Smart in a Smart World: Why Human Intelligence Still Beats Algorithms".
Offers a very astute perspective on the differences between crystallised and fluid intelligence.
«large language models are hitting an observable bottleneck. GPT and DeepSeek are wonderful tools for general research about information accessible from the public internet, but internal knowledge (both personal and corporate) is still beyond reach»
LLMs are widely misunderstood:
* They are search engines actually, but search engines of "potential" documents.
* If you do a query with some keywords to a web search engine it will return hundreds or thousands of complete documents that relate to those keywords.
* If you do a query with the same keywords to a LLM search engine it will return one document which is one of the potential mixes of those hundreds or thousands of documents that a web search engine would return. Another way to say that is that LLM search engines do not search just existing documents but potential documents that are in some sense "between" actual documents.
Note: the underlying major differences is that web search engines associate keywords with whole documents while LLM search engines associate keywords with fragments of documents down to single word (or even characters) and that's why they can reassemble those fragments into plausible "potential" documents.
So LLMs cannot answer questions that are "obvious" but are not contained in the documents they have indexed. For example here is ChatGPT:
«> how many "r" letters are in the word strawberry?
ChatGPT said: The word "strawberry" contains 2 "r" letters.»
That is because there is no document that says something like "strawberry has 3 "r" letters".
Note: Some LLM search engines recently have got "reasoning" that is "expert systems" like algorithms so they have some ability to infer answers from existing documents.
Robert, great post. I believe the great Zen, Taoist, and Buddhist masters have a subtle but importantly different take on "Mind". We have those amazing paintings of mountains and gorges. Steeped in mist and mystery. They are attempting to tell us,something. Only in awe and silence do we cross over into a deep reality beyond our current awareness. Our current models of mind are too impoverished still.
My background is in psychology and I have an interest in the interface between our humanity and our technology. The research in developing AI “memory” is exciting in that it is forcing us to look deeply into the processes which form what we call “consciousness” and examine many of the misconceptions and cultural perceptions we have accumulated over eons. This new research is forcing us to re-examine our concept of consciousness and to scientifically develop a roadmap of all the processes that somehow are sorted and combined to form a consensus we call consciousness.
Thomas Moore wrote a series of books on “Care of the Soul”. What I find lacking in AI is it’s stark “soulessmess.” Moore defines “soul” as the messy interface between our powers of reasoning and of sensing and experiencing. I remember the old computer term “fuzzy logic” which were early attempts to bridge the gap between machine logic and the ‘hot mess’ of human consciousness. It seems that dynamic is now being fully explored.
Can our machines help us to ironically become more human and less machine-like? Jacob Bronowski wrote a book and hosted an enlightening BBC series, “The Ascent of Man”, documenting how advances in technology have vastly altered human psychology and sociology. We have evolved to become more “machine-like” as mastering new technologies gave individuals and societies power and wealth. It seems in the process, we have sacrificed our humanity for the power of technology resulting in a plague of diseases of despair - depression, drug addiction, and suicide. Can we relearn from our technology as it better mimics our humanity to again honor and respect the wonder of our illogical humanity? Or, will we plunge ourselves into even deeper collective despair?
Robert, I'll encourage you to read Gerd Gigerenzer's excellent book, "How to Stay Smart in a Smart World: Why Human Intelligence Still Beats Algorithms".
Offers a very astute perspective on the differences between crystallised and fluid intelligence.
Thanks!
«large language models are hitting an observable bottleneck. GPT and DeepSeek are wonderful tools for general research about information accessible from the public internet, but internal knowledge (both personal and corporate) is still beyond reach»
LLMs are widely misunderstood:
* They are search engines actually, but search engines of "potential" documents.
* If you do a query with some keywords to a web search engine it will return hundreds or thousands of complete documents that relate to those keywords.
* If you do a query with the same keywords to a LLM search engine it will return one document which is one of the potential mixes of those hundreds or thousands of documents that a web search engine would return. Another way to say that is that LLM search engines do not search just existing documents but potential documents that are in some sense "between" actual documents.
Note: the underlying major differences is that web search engines associate keywords with whole documents while LLM search engines associate keywords with fragments of documents down to single word (or even characters) and that's why they can reassemble those fragments into plausible "potential" documents.
So LLMs cannot answer questions that are "obvious" but are not contained in the documents they have indexed. For example here is ChatGPT:
«> how many "r" letters are in the word strawberry?
ChatGPT said: The word "strawberry" contains 2 "r" letters.»
That is because there is no document that says something like "strawberry has 3 "r" letters".
Note: Some LLM search engines recently have got "reasoning" that is "expert systems" like algorithms so they have some ability to infer answers from existing documents.
Robert, great post. I believe the great Zen, Taoist, and Buddhist masters have a subtle but importantly different take on "Mind". We have those amazing paintings of mountains and gorges. Steeped in mist and mystery. They are attempting to tell us,something. Only in awe and silence do we cross over into a deep reality beyond our current awareness. Our current models of mind are too impoverished still.