Books: A Thousand Brains and Code Breaker

Bill Gates's new winter reading list came out yesterday, and as usual he has some interesting choices.  I bought the two most interesting ones, which arrived already today (gotta love Prime!)

"A Thousand Brains" by Jeff Hawkins is subtitled "A New Theory of Intelligence" and appears it will be arguing that our brains are not just one thinking glob of grey matter.  Hawkins argues that our brains are committees, with competing sub-sections.  In effect our brains "vote" on which interpretations to accept, and which to reject.  It sounds like a very interesting premise, one that explains why some people accept different realities than the consensus of other brains.  It also neatly defines why AI feels and acts so "artificial."

The second book is "The Code Breaker: Jennifer Doudna, Gene Editing and the Future of The Human Race" by Walter Isaacson.  It describes the discovery, refinement and uses of CRISPR.  Fits right in with my recent reading (Carl Zimmer):

Quote :

Chapter 17 details the development and use of CRISPR to modify genes, in a targeted way, to snip out or insert specific genes to produce specific results.

Walter Isaacson, whose biography of Steve Jobs was such a fun read, is an excellent writer whose specialty is taking highly-technical subjects and explaining them in laymen's term, without dumbing them down.  So far (1/4 way through) his book about Jennifer Doudna is compulsively readable, almost unputdownable.  The staggering successes and advances these people made in biotechnology are awe-inspiring.  My sister worked in a genetics lab for twenty years, and I have some small idea of the incredible devotion it takes to tease out invisible results from ambiguous experiments.

I see Isaacson also did a bio of Ben Franklin.  I think thats on my "to-read" shelf.

The second half of the book, starting with chapter 35, delves into the ethics of genetically-altering germs cells, eggs or sperm.  This makes the alteration ubiquitous in the resulting organism, and heritable by its offspring. The temptation to improve human genetics by removing deleterious genes was nearly irresistible--and from there it's a short hop to making all babies blue-eyed and blonde and 6'4" and IQ 120.

Scientists were wary of crossing that line, and by mutual contentious debate agreed not to cross that red line before standards were established and in place.

Then a Chinese scientist went rogue, and created two genetically engineered babies in 2018.

Quote :

"The requirements of embryo [modification] are minimal: a micro injector, micro pipette, and microscope. All of these can be purchased on eBay and assembled for a few thousand dollars. Human embryos can be bought from fertility clinics for about $1000," he said. "You can probably have the embryo transferred to a human by a medical doctor in the U.S. if you don't tell him or her what you've done, or you can do it in another country."

Wait, what?  You can buy human embryos for a thousand dollars?  Really?  That doesn't seem right.

NoCoPilot wrote:

Wait, what?  You can buy human embryos for a thousand dollars?  Really?  That doesn't seem right.

I don't think it's right either, but I can well imagine that it's true.I can't swear to the exact details, never having tried it, but I think for IVF they usually create several petri-dish embryos, and pick the healthiest looking one to implant back into the mother. What happens to the left-overs?

In Chapter 42 "Who Should Decide?" the author makes a comparison.  In Orwell's "1984" he envisioned a society totally dominated by a centralized computer that controls everyone's movements and thoughts.  Big Brother is always watching you.

Instead, in the year 1984 Steve Jobs introduced an easy-to-use personal computer, giving millions of end-users the ability to upload their thoughts and opinions (and conspiracy theories and lies and misinformation) to a public forum with no rules or regulations whatsoever.  The result has not been 100% positive.

Isaacson sees the democratization of genetic engineering moving the same way.  Instead of a couple of regulatory boards and the medical establishment deciding who can use GE for eliminating which genetic diseases, he sees the process getting so easy and ubiquitous that wealthy individuals will be able to enhance their family descendants with better health, stronger bones, bigger muscles, smarter brains and maybe, some day, biological advantages that Mother Nature never envisioned for us: infrared sight.  Ultrasonic hearing.  Longer life.  The ability to digest cellulose, or glow in the dark.

The author worries that if everyone can go to "a genetic supermarket" and choose their child's susceptibility to genetic abnormalities, their child's height and weight and intelligence and mental stability, then we'll have no more Danny DeVitos in society, no more Leonardo DaVincis, no more Marty Feldmans.  He feels diversity gives us empathy and acceptance, whereas if we all look like Mark Spitz then poor people who are born naturally will be looked down upon as Moorlocks.

The "All men are created equal" axiom will no longer be true.

Part Nine, beginning at Chapter 48, details how all of this CRISPR, RNA, gene-editing expertise was called into emergency use to develop tests and vaccines for COVID-19 when the Trump Administration failed to do anything. A remarkable cross-discipline, cross-academic, cross-commercial collaboration sprang into being almost instantly, with zero help from the government. Everything was put in the public domain, with no patents, because the need was more important than profit.  Astonishing story.

Moderna was a relatively-small biotech company, 800 employees, investigating mRNA (messenger RNA) as a way to target various cancers.  It had not yet had any of its investigatory vaccines approved, or even completed the testing procedure on them.

Nevertheless, when Chinese scientists posted the entire COVID-19 genome (26,100 pairs) in early 2020 they realized their technology could be used to deliver a vaccine.  It took them only two days to create the vaccine.  In only 38 days they had a manufacturing process in place, and sent the first vials to NIH for human trials.

When the CEO of Moderna heard, on a Zoom call, that their vaccine had tested 90% effective, he said, "Say again?  Did you say 19%... or 90%?!?!"

The book remained compulsively readable all the way through; I had to pace myself to avoid rushing through it in a day or two and missing the details. Isaacson is a master at making science at least SEEM understandable to the layman, and his enthusiasm for bioscience is as contagious as the Omicron variant.

I did eventually finish "1,000 Brains" though I never came back to report on it. I put Jeff Hawkins' only other book, "On Intelligence" (2004) on my want list.

A copy turned up at Amazon last week.

His premise is that what we call "artificial intelligence" is really "expert systems" and not a generalized intelligence. In order to build a generalized intelligence, we first have to define what it is, and what it isn't.

Unfortunately the book is written in that pedantic style I hate, where he says "Here is what I'm going to teach you" and then states some opinion of his and then says "Now that we have that settled..."

I'll try to muddle hough. The subject matter should be interesting.

Getting back to "On Intelligence" after a couple diversions to clear my mind. The author makes some very good points, which I haven't seen mentioned in any other AI text I've read.

Our brains constantly map our surroundings.  We carry within us a mental map of where everything is and what everything does, including smells and tastes and how heavy things should be.  That's why we're immediately alert when something is out of place, why we feel disoriented in a place we've never been before.

It would be nearly impossible to program a computer with this detailed a representation of the world, because the level of detail would approach infinity.

At the same time, our brains don't "memorize" specifics.  We remember patterns and relationships.  We recognize a tune played on a piano, even if it's being played in the wrong key.  We recognize friends, even if they've dyed their hair or are wearing sunglasses.

Getting a computer to recognize patterns is even harder, because generally this is done by consulting a database of specifics.  Computers don't do so well with generalities.

Though he hasn't explicitly said so (yet) I can extrapolate the author's theory of intelligence to postulate that consciousness is the memory of our surroundings, that we all carry around in our neocortex, combined with our sensory input. By comparing what's going on around us, what we're seeing and feeling and smelling and hearing, we confirm that our mental model of the world is correct. Or not correct. This constant "bearings checking" is the basis of consciousness. It projects us a little bit into the future and a little bit into the past, so we're not living in a constant "right now" like most animals, including our pets and our politicians.

Back in the old days on DU, Keith and I came to the conclusion that the way the brain works is, at least in part, by simulation of the world. To figure out what is going to happen, you make a model, plug in what's happened in the past and how you think things work, and project what's going to happen next. At some level, this another way of saying what you just said, that the brain maps the environment.

But taking it a little further, this idea explains consciousness fairly easily. Most of this mapping and extrapolation, simulation, whatever you want to call it, happens at a subconscious level. (Most of everything that the brain does happens subconsciously.) But for some situations, the brain needs to know, or decide, what you should do, rather than just react by reflex (so to speak). Especially for those situations, the brain needs to simulate itself, so that it can extrapolate a future that is affected by it's own actions.

This is consciousness - the brain's simulation of the brain. When you are thinking, consciously, the precise piece that is doing the thinking isn't your brain, exactly, it's the simulation of the brain that's running inside your physical brain. Hence the Cartesian theater, the feeling that "you" are living inside your body and looking out through your eyes, but are somehow separate from your body. That's actually exactly true. But it doesn't invoke a need for something extra, a soul or whatever. It's a kind of program running on the flesh computer.

To me, this seems fairly obviously to be the logical conclusion that the simulation scheme leads to. But I don't know if anyone who actually studies this stuff has drawn the same conclusion.

That's a really compelling insight, Richard. The more I've thought about it, the more it resonates as "probably spot on." Consciousness isn't just awareness of the past and future, it's awareness that we are aware of the past and future. The Cartesian Theatre, I'll have to remember that.