In the series of posts of here I have made many critical comments about the work of science journalists, who all too often these days act like pushovers who treat with kid gloves groundlessly boasting scientists, while failing to apply critical scrutiny to the dubious claims of such scientists. So I was rather surprised to read a recent book by a science journalist who uses an impressive amount of critical scrutiny when discussing claims in the technoscientific sphere. The very readable and informative book is called "More Everything Forever," and its author is Adam Becker. The subtitle is "AI Overlords, Space Empires and Silicon Valley's Crusade to Control the Fate of Humanity."
Becker is someone with a PhD in astrophysics, but the book's dust jacket describes him as a science journalist. Some of his astrophysics background comes in handy when he gets around to criticizing some of the schemes and dreams of hi-tech titans that he examines.
We read about hi-tech titans such as Sam Altman who are predicting there will soon be "AGI" or "Artificial General Intelligence." The term "AI" stands for artificial intelligence. Strictly speaking, there does not really exist any such thing as artificial intelligence. So-called AI is mainly merely computer programming and data processing. There has been very much progress in computer programming and data processing. But no such progress ever justified the use of either the term AI (artificial intelligence) or the term AGI (artificial general intelligence).
What happened is a massive abuse of language similar to the massive abuse of language that occurred around so-called "natural selection" (which is not really any such thing as selection, because "selection" means a choice by a mind, and blind unguided unnatural processes do not make choices). Back around 1988, the term "artificial intelligence" was just a kind of aspirational term used for software toolkits that could be used to try to make progress at producing artificial intelligence. In the following years more and companies realized that it was very profitable to just stop talking in aspirational terms, and to start pitching whatever product they had as an example of "artificial intelligence." It was a deceitful escalation of hype, and the public fell for it.
By now we have examples of computer programming and data processing that are much more impressive than anything existing around 1988. But no such examples qualify as "artificial intelligence," using a good definition of "intelligence." The Merriam Webster dictionary defines "intelligence" as "the ability to learn or understand things or to deal with new or difficult situations : reason." The American Heritage Dictionary defines "intelligence" as "the ability to acquire, understand, and use knowledge." The Oxford English Dictionary defines intelligence as "The faculty of understanding; intellect." Computer systems do not understand anything, so they are not intelligent. So-called AI systems such as OpenAI and Google Gemini are not minds and not intellects. They understand nothing.
You may start to realize what a swindle the term Artificial Intelligence is when you realize that you cannot have an intellect and you cannot have intelligence until you first have experience, which means a mind that is living a life. Given a mind that is living a life, there may be some understanding of some concept by that mind. But where there is no experience and no life, there can be no understanding. Computer systems do not have experiences, do not have lives, and do not understand anything.
How do you infer that other people are living a life, that other people have experiences, and that other people understand things? Well, it works rather like this. You know for certain that you have a life, that you have experience, and that you understand things. And you know what kind of being you are, from viewing your body and seeing yourself in the mirror; and you know that other people around you look and move pretty much the same as you look and move. So it is entirely logical to infer that other people you see have real lives like you do, that they have experiences like you do, and that they understand like you do. But no such inference would ever be justified when thinking about a computer.
If you were a robot having experiences and living a life and understanding things, and you looked down to see a robot metal body below your eyes, and you saw other metal robots looking like yourself, then you would be justified in inferring: robots have lives, robots have experiences, and robots understand things. But you would never, ever be justified in making such a conclusion if you had only lived as a person of flesh and blood.
Around page 14 Becker starts to discuss longtermism. It's a philosophical approach in which various extravagant proposals are justified by means of dubious calculations in which someone tries to quantify the goodness of supposedly good things that can happen over extremely long time spans. So, for example, someone engaging in longtermism might justify spending billions on an interstellar expedition, because that expedition might be the beginning of a colonization of the galaxy that might result in 1,000,000,000,000,000 humans existing over the span of a million years. The longtermism reasoner will say things like this: "You must support my plan to launch the mission, because if it is not launched, that will be like killing 1,000,000,000,000,000 future inhabitants of the galaxy!"
As I point out in my post here, entitled "Longtermism Is Fueled by a Goofy Belief in Computer-Generated Lives," longtermism is centered upon the unreasonable cheat of assuming that there can exist computer-generated lives. In an article describing longtermism, we read this:
"Longtermism is a quasi-religious worldview, influenced by transhumanism and utilitarian ethics, which asserts that there could be so many digital people living in vast computer simulations millions or billions of years in the future that one of our most important moral obligations today is to take actions that ensure as many of these digital people come into existence as possible. In practical terms, that means we must do whatever it takes to survive long enough to colonize space, convert planets into giant computer simulations and create unfathomable numbers of simulated beings. How many simulated beings could there be? According to Nick Bostrom —the Father of longtermism and director of the Future of Humanity Institute — there could be at least 1058 digital people in the future, or a 1 followed by 58 zeros."
Such calculations are nonsense, largely because of the impossibility of using computers to generate a life. Computers do not and can never produce experience or lives. What happened is that people got interested in Nick Bostrom's senseless theory that we could be living in a computer simulation created by extraterrestrials, a theory predicated on the utterly irrational and groundless idea that computers can generate lives like the lives we are living. The same people who had got interested in Bostrom's nonsensical theory then kind of reasoned: "Well, if the extraterrestrials can generate computer-generated lives, then we can too!" So they started estimating that future humans with far more powerful computers will be able to generate far more lives than the flesh-and-blood lives that now exist on planet Earth -- maybe billions or trillions or quadrillions of times as many lives.
This was all the worst kind of nonsense, nonsense as bad as Bostrom's original speculation that we are living in a computer simulation. Computers cannot generate experience. Computers cannot generate lives. So we cannot be living in a simulation created by extraterrestrials. And we will not be able to create lives by having computer systems generate lives. And you will not be able to obtain immortality by downloading your mind into a computer. But that's no reason to be sad. The very reasons for concluding the impossibility of mind uploading into computers are part of the very many reasons for having confidence that your mind is something that cannot be produced by a brain, and something that will survive the destruction of your body.
Believing that progress in computer systems will one day lead to computers being able to generate lives is a kind of error similar to the error of thinking that progress in improving the resolution of TV sets justifies the claim that one day we will have TV sets that cause animals and people to literally jump out of the TV screen, allowing you to hold them or touch them.
Around page 39 to 47 Becker discusses the Singularity claims of Ray Kurzweil. It starts out on page 39 with a creepy discussion by Kurzweil about hi-tech means to "resurrect" his dead father. On page 39 the book quotes him as saying this about his father:
"We can find some of his DNA around his grave site...AI will send down some nanobots and get some bone or teeth and extract some DNA and put it all together. Then they'll get some information from my brain and anyone else who still remembers him...[They'll] just send nanobots into my brain and reconstruct my recollections and memories."
This is creepy nonsense, both because of the impracticality of nanobots doing such a task, and also because brains do not actually store memories. Many brains of the recently deceased have been examined with very powerful microscopes, as has recently extracted tissue from living people. But such microscopic examination of brain tissue has never discovered the slightest trace of anything a human ever learned -- not a single word, not a single letter, not a single image and not a single pixel. No scientist has a credible account of how a brain could learn or recall anything, given the known physical shortfalls of all brains (which neuroscientists senselessly ignore).
Becker's critique of Kurzweil's claims is not as strong as it could have been. He fails to make a list of incorrect predictions that Kurzweil previously made. A list of some of those predictions can be found in my post here. Becker also fails to criticize the incorrect assumptions about minds and brains that are a foundation of Kurzweil's claims about a Singularity. But Becker does make a good criticism of the flaws in Kuzweil's claim about a law of accelerating returns, the claim that progress is increasing exponentially. He points out that a pillar of such an idea (Moore's Law, that the number of transistors on a chip doubles every two years) no longer holds true.
Today's user of the Internet should laugh hard at the notion that progress is increasing exponentially. More and more times that I go to some web site I used without any hassle in the year 2021, I find that the website is no longer functional for anything other than viewing headlines. Upon clicking on one of these headlines, I so often get a screen telling me I cannot view that article without paying for a subscription. The web pages that I go to are very often very badly cluttered up with ads, both on the sides and in the main text. So reading is rather like this:
4 or 5 sentences of the article.
An ad.
6 or 7 more sentences of the article.
An ad.
4 or 5 more sentences of the article.
An ad.
A typical user of the Internet should laugh hard at claims of a law of accelerating progress, and maybe complain that the Internet worked better in the year 2000. Ask some so-called AI system some of the most important questions, and you will often get some very unwise or very false answers. What often happens is that materialist misconceptions and dubious socially-constructed triumphal legends of academia are being sucked up by web-crawling AI systems, with the errors and dubious dogmas being repackaged and regurgitated as AI slop. You might call the resulting answers "Artificial Stupidity."
Around page 70, Becker criticizes Kurzweil's claims that there would be exponential progress in brain scanning that would help allow computers to become as smart as people. He points out that such progress in brain scanning has not occurred, saying that "the maximum resolution of noninvasive human brain scanning in 2024 is not appreciably better than in 2000."
Becker's argumentation around this point misses the mark. He argues along the lines of "it's hard to exactly reproduce the brain electronically." He should have studied brain physical shortfalls, and argued "it's futile to try to reproduce the brain electronically." The real problem here is that no amount of brain scanning or brain emulation will do anything to help build superintelligent computers. Brains do not produce minds, and brains bear no real resemblance to computers or devices for storing or retrieving memories. No neuroscientist can give any credible explanation of how a brain could produce a thought or imagine anything or learn anything or recall anything. When neuroscientists talk about such things, they give us only the emptiest of hand-waving, or suggest ideas that cannot possibly be true given what we have learned about the physical shortfalls of brains and the rapid turnover and short lifetimes of their components such as protein molecules, dendritic spines and synapses.
Thus far the study of human brains has done absolutely nothing to help produce advances in computing and software. There has been progress in various types of software that are wrongly called "neural nets." But such software does not bear any close resemblance to anything going on in the brain. In so-called neural nets, information is passed along with 100% reliability. This bears no resemblance to the physical reality of the brain, in which electrical signals have to pass through endless numbers of synapses, each of which transmit a signal with a reliability of less than 50%.
Around page 223 Becker criticizes aspirations to colonize space by building the type of space colonies imagined by Gerard O'Neill. He says the scheme depended on the existence of a "mass driver" that would eject raw materials from the moon, but that "the technological advances that O'Neill was hoping would enable the mass driver simply didn't work out."
O'Neill's original proposal was for massive rotating cylinders which would give the inhabitants a magnificent view of a large landscape. His later proposal was for space colonies offering no such view, and we might ask, "Who would want to live in such a colony?"
Around page 226 Becker's background in astrophysics comes in handy, as he criticizes the impracticality of schemes such as Elon Musk's proposed scheme to colonize Mars. He criticizes the "Techno-Optimist Manifesto" of Marc Andreesssen, something I criticized in my post here.
Although he misses some opportunities to make a stronger critique, Becker has shown lots of ability to effectively criticize unwarranted ideas of powerful figures in the hi-tech field. It would be great to see Becker write another book making a critique of grandiose but unwarranted claims, with the focus being on the grandiose but unwarranted claims of today's biologists.
No comments:
Post a Comment