Science and Inquiry discussion

For July 2022, we will be reading A Thousand Brains: A New Theory of Intelligence by Jeff Hawkins.

Please use this thread to post questions, comments, and reviews, at any time.

I have started reading this book. The deeper I go into this book, the more fascinating the story becomes. The author has some really good insights into the functioning of the neocortex.

I downloaded this book from the library and it turned out to be a summary. I did read through it but I need something that takes me more than 10 mins to read if I want it to stick. One thing I remember is that the author thinks it more likely that human behavior will lead to the doom of the human race. So I guess at present, that threat of death by nuclear weapons > threat of death by intelligent killer robots.

Is anyone else reading this book?

I started last night and find it fascinating. As a person without any formal science training, I think he explains a difficult topic in a way that's easy to understand.

I have had my library copy for a few weeks. I hope to begin it soon.

I finished this book this week. I agree with David that it gets better and more interesting as it progresses. Hawkins explains how, rather than simply processing external stimuli, the brain operates using mental models, which is what allows for human intelligence. The "thousand brains" in the title refers to Hawkins' findings that the models are generated from literally thousands of separate models using "reference frames." I don't know enough about brain science to evaluate the validity of those claims but did find his explanations intriguing.

From there, the book examines the future of intelligence, including artificial intelligence (which Hawkins believes will come about when we model machine learning on how the human brain actually works) as well as how humans might store and transmit knowledge over vast time and distance. I found this latter half of the book fascinating and appreciate Hawkins' ability to speculate about such matters in a clear and reasoned manner. In particular, I liked the sections on why we have little reason to fear advanced AI or be afraid of broadcasting our existence to the cosmos. I also appreciate his naturalistic orientation when discussing topics such as consciousness, which he believes results entirely from physical processes that will eventually be understood (and replicated in AI) rather than appealing to something fundamentally mysterious. Overall, I found the book thought-provoking and am glad this was selected for this month's group read.

I just finished this book. The entire book is fascinating, but it just doesn't seem to hold together. He doesn't really have a "theory", but just a reasonable "hypothesis" about how the neocortex works. He doesn't go into much detail about the hypothesis, and doesn't seem to support it very well. His 4 attributes of AGI (Artificial General Intelligence) are interesting. But I've written software with these 4 attributes, and I would never claim it to be intelligent! Here is my review.

David wrote: "I just finished this book. The entire book is fascinating, but it just doesn't seem to hold together. He doesn't really have a "theory", but just a reasonable "hypothesis" about how the neocortex w..."

Thank you for the review. I probably will not read this book beyond the summary I already read but your review helped me understand a little more what the book is about. I'm actually kind of glad that you thought it was a "mishmash" because even from the summary, I felt like it was kind of all over the place without a common theme. I wasn't sure if that was me not understanding or if it was the author. Now I know!

Regarding Hawkins' description of how AGI might work, it is hugely lacking in specifics. I understand given that, and the variety of other concerns through the latter half of the book, that it could leave readers feeling whiplash. I took it as a popular science book exploring a range of far-future considerations, which I personally enjoyed. David - I think your idea re how Bayesian credences might play a role in cognition is super interesting. This could potentially be an important insight.

I'm making slow progress, and I've finished Part 1.

It's been occasionally frustrating. First, Hawkins' theory is really a hypothesis as it hasn't been replicated & thus verified by other scientists. Second, as I progressed, his tone shifted and he wrote as though his "theory" is actually fact. For instance, ch 2 presented Mountcastle's theory of cortical columns; it had at least received some validation from other neuroscientist. Hawkins then arbitrarily assigned each cortical column to be 1 square millimeter in size so that therefore 150,000 cortical columns exist, but this is an assumption.

Also, it's not faring well in comparison to my previous science book - Immune: a Journey into the Mysterious System that Keeps You Alive. While Philipp Dettmer employed many metaphors and similes, I always felt as though I was learning a great amount. But Hawkins spent a lot of time on his cup analogies for a much lower payoff of information.

Slowly progressing...

I wondered whether Vernon Mountcastle's theory about cortical columns is widely accepted. I'm not going to wade through the ocean of scientific articles so the targeted online search yielded this --> https://www.ncbi.nlm.nih.gov/pmc/arti...

As this is from the NIH, I accept its credibility.
This paper inquires how the column came to be viewed as an elementary unit of the cortex from Mountcastle’s discovery in 1955 until David Hubel and Torsten Wiesel’s reception of the Nobel Prize in 1981. I first argue that Mountcastle’s vertical electrode recordings served as criteria for applying the column concept to electrophysiological data. In contrast to previous authors, I claim that this move from electrophysiological data to the phenomenon of columnar responses was concept-laden, but not theory-laden. In the second part of the paper, I argue that Mountcastle’s criteria provided Hubel Wiesel with a conceptual outlook, i.e. it allowed them to anticipate columnar patterns in the cat and macaque visual cortex. I argue that in the late 1970s, this outlook only briefly took a form that one could call a ‘theory’ of the cerebral cortex, before new experimental techniques started to diversify column research. I end by showing how this account of early column research fits into a larger project that follows the conceptual development of the column into the present.

And then I found this, also by NIH -- https://www.ncbi.nlm.nih.gov/pmc/arti...

This year, the field of neuroscience celebrates the 50th anniversary of Mountcastle's discovery of the cortical column. In this review, we summarize half a century of research and come to the disappointing realization that the column may have no function.

[Disclaimer - I did not read these articles in their entirety as they're both 100+ pages in their online forms..]

So Hawkins is an intelligent person and he's staking his professional reputation on the line. It remains to be seen whether he'll be proven correct as he was about personal, handheld computers.

Part 2 for the most part struck me as the musings of the big questions.

According to Wikipedia - https://en.wikipedia.org/wiki/AI_take...
Physicist Stephen Hawking, Microsoft founder Bill Gates, and SpaceX founder Elon Musk have expressed concerns about the possibility that AI could develop to the point that humans could not control it, with Hawking theorizing that this could "spell the end of the human race".[33] Stephen Hawking said in 2014 that "Success in creating AI would be the biggest event in human history. Unfortunately, it might also be the last, unless we learn how to avoid the risks." Hawking believed that in the coming decades, AI could offer "incalculable benefits and risks" such as "technology outsmarting financial markets, out-inventing human researchers, out-manipulating human leaders, and developing weapons we cannot even understand." In January 2015, Nick Bostrom joined Stephen Hawking, Max Tegmark, Elon Musk, Lord Martin Rees, Jaan Tallinn, and numerous AI researchers, in signing the Future of Life Institute's open letter speaking to the potential risks and benefits associated with artificial intelligence. The signatories "believe that research on how to make AI systems robust and beneficial is both important and timely, and that there are concrete research directions that can be pursued today."

Again, Hawkins boldly stakes his claim that AI will not threaten human life. Apparently our human brains, specifically the existence of our "old brains," are sufficient to threaten the future of humanity. Well, I have no doubt about the latter point.

I’ve finished the book today under the Croation sun, enjoying a lovely Holiday.

The first part of the book about our old brain and neo cortex was fascinating including Hawkins view how to build AI so it is able to learn for itself.
The last part of the book was less interesting, where to many scenario’s of our future and actions we should take. A bit to pessimistic in my view. Conclusions and view a bit quick in my view.

Therefore only 4 stars

Sorry for any spellingmistakes, not my native language

Started it today, but I’m a little bit concerned it will be a stupid read after some of the comments above.

One thing sticking in mind from Part 3 -

Hawkins estimated that brains have 100 billion neurons. A Google online check estimated 86 billion neurons. Either way, both estimates are staggeringly vast.

Back to this book -- with billions of neurons and trillions of synapses, the computer industry doesn't have the capability to "upload" a human brain into a computer. So this will probably remain pure science fiction for quite some time to come.

I struggled with writing this review. I found the book to be okay - I'm neutral about it. But if I go with 2 stars, GR regards that as equivalent to disliking a book, Maybe I don't know enough about AI or neuroscience to be excited about Hawkins' content, even though Hawkins wrote this for lay readers.

My review - https://www.goodreads.com/review/show...

I don’t think the author is actually providing proofs of what he theorizes in this book, although he indicates he has the proofs which has led him to his theories. It is possible the suggested readings he suggests at the back of the book provide the proofs which has led to the tentative ideas in this book.

It has been some time, and many other, often trashy, books I’ve read in between this book and The Alignment Problem: Machine Learning and Human Values, plus I am no scientist or even brainy, but this book and ‘The Alignment Problem’ are at odds with their conclusions? Didn’t ‘The Alignment Problem’ discuss how any determination of how the wet matter of the brain works, which I presume means “columns”, is not the way to AGI?

Hawkins poo poos the danger of “black boxes”, where the machine begins working on information designers have no clue of how the computer decided on to use, while in ‘The Alignment Problem’, didn’t the author mention how computers began erroneously deciding on issues, after receiving input which was supposed to be helping it learn, but the computers learned the wrong thing to be learned from the inputs? And programmers had a very very difficult time figuring out why the computer was outputting obvious biased or incorrect conclusions?

It’s interesting if I am getting this correct, that the two books diverge quite a bit in their, dare I say it, “old brain” feeling, about what is evidence? However, imho, ‘The Alignment Problem’ IS based more on actual facts. Hawkins seems a touch Pollyanna to me, well, ok, a lot optimistic, about building computers based more precisely on human brain matter functioning, instead of on what information the brain sees and working on an approximation of that ‘seeing’ the brain does with information, what it throws out and what it thinks important, and translating it into silicon functioning, not trying to recreate a machine version of brain neurons at all.

Am I making sense? I’ve been reading for hours so I’m a bit woozy. But I think Hawkins thinks if we don’t give computers our “old brain”, it will be just fine. ‘The Alignment Problem’ says computers have a computing problem, with or without the “old brain”.

Anyway, I believe there are other areas where these two books diverge. I think.

My review: https://www.goodreads.com/review/show...

Still catching up with the summer reads, and I just finished this one! I generally agree with David on the book being mostly around one idea and without providing a holistic model of intelligence, but overall it’s an enjoyable read because the author has quite a creative way to imagine how human knowledge and the species itself might persist on their instinct to prevail. I did miss more insights into the “old brain” which is mostly a black box throughout the book, and some connections with psychology or mental health overall, which are completely disregarded. I still gave it 4 stars because I really enjoyed it and it is quite thought provoking in most chapters. And I would recommend the read to those in doubt!