Sunday 6 August 2023

So-called AI is all about human beings - Not about "thinking machines"

Artificial Intelligence (AI) research began by claiming to be able to make a thinking machine: an intelligent computer that was intelligent in the way human beings are intelligent. 

It has ended-up by fobbing us off with computer programmes that use human intelligence to mimic human intelligence (in ways that the uninformed and inattentive users do not always and instantly detect - this being the current ultra-low bar criterion of successful AI!...)...

And then claiming that they have done the job they set out to do! 

It seems that people simply do not notice that human beings are involved at every step of the current so-called AI!

The currently vaunted text AI is first programmed by human beings; then it uses human-produced texts in order to generate human-mimicking texts; then it "learns" by being subjected to continual positive and negative reinforcement by human beings!   

In other words; the "Learning" of AI by "feedback" gathered during usage; amounts to a continual process of re-programming!

It has reached the bizarre situation that a massive and continual process of human-dictation, direction, and correction of computation; is somehow being edited-out from our awareness - and all the 'credit' is being allocated to Artificial, not human, intelligence! 

People are fooled because they want to be fooled; and they fail to learn from experience because they fear to learn. 

Obvious failure may thus be 'sold' as success; and System-sustaining propaganda gets lapped-up by masses of willing dupes who want nonsense to be true.


Note added: Why do so many people want the AI nonsense to be true? Perhaps because they (rightly) mistrust and fear human beings; and have rejected faith/trust in the Christian God so deeply and thoroughly that they are almost insensible to their own consequent psychosis -- therefore they grasp at the straw of potentially wise-and-benign AI, as the only hope they know for avoiding the nihilism of utter despair.


Inquisitor Benedictus said...

AI is magical thinking. If you read Alan Turing's original paper (which isn't overly technical or hard to follow) on what constitutes AI, you'll see he brackets out any immaterial or substantive or ontological notion of the [i]mind[/i], and replaces it with a purely functional notion that, "If it looks like what we call 'intelligence", then for all intents and purposes it is intelligence." Notice how the concept of mind or intelligence has been thrown back to pure linguistics, to pure naming/labelling. If we would call it intelligent, it must be an intelligence. In other words, it's all about spelling, casting a spell. If you can make a machine put on a convincing enough performance that an unaware person might think it intelligent, it therefore must be intelligent. After all, modern philosophy has no metaphysical concept of the Mind (the AI theorists themselves dismissively refer to it as a "mystical" concept of mind), so mental or intelligent activity becomes a pure naming convention; so it's all about how we can get that naming convention to apply to machines. So really, it's not about making machines mindful, but about proving our own minds mechanical. It's an experiment being practiced on humanity under the guise of computing.

Here's a very good video explaining why it's a sham.

The author makes the very crucial distinction between AI, the science fiction concept, and machine learning, which is a useful branch of computing. I did a bit of computer science myself and can say that, while machine learning is indeed a very ingenious form of technology and computing, so-called "AI" is incredibly stupid. It is no more intelligent than a switch, because all it really is is a series of switches arranged in an ingeniously complicated way (which is of course how materialists view the brain).

I could see this magical thinking at work in an interview on AI recently where a robotics team left their object-finding robot on over a couple of weeks, and it suddenly went from having no object-finding capability to very accurate capability. The magical thinking is to say that the robot somehow acquired a mind over those couple of weeks, and that if it were left on for longer it would grow into a human like intelligence. This is complete nonsense and anyone who has done even a small amount of computing knows it. The reason the robot acquired the capability so suddenly is just that its boringly iterative programming finally broke through a certain barrier in the data processing method. The robot itself was really no more intelligent than it was on day 1, it's just that the data had filtered down the pipes long enough to finally land in the right baskets. But, again, this is precisely the crude concept of mind which materialists have, and AI is all about proving that crass materialism in action and experiment.

Bruce Charlton said...

@IB - Excellent comment!

I wrote this about Turing five years ago:

Not really relevant, but a little known fact wrt Turing's current sanctification:

Inquisitor Benedictus said...

Wittgenstein is very interesting because he took the (sometimes called "autistic") mode of intelligence associated with philosophical rationalism, and managed to loop it back round to an understanding of the universe and man which was open to mystical and super-rational understanding. This goes to show that, in a certain sense, all philosophical paradigms *can* lead you back to the truth eventually, if you are willing to follow your own premises with enough honesty and perseverance. The problem is that lesser thinkers than Wittgenstein, like Bertrand Russell and like Turing, are not willing to follow their own arguments far enough to expose their ultimate insufficiency. Wittgenstein was willing to do that.

Bruce Charlton said...

@IB - I regard Wittgenstein as wrong, but wrong on the basis of his metaphysical assumptions; which are those held by almost everyone (I mean, everyone except me!). Therefore, he is a very interesting figure.

I am currently (i.e. c 15% through) reading a very interesting and intense "not-biography" of W. by Miles Hollingworth (Ludwig Wittgenstein, OUP, 2018), which takes him absolutely seriously, and is a philosophical work of unusual commitment and focus.

You can get a flavour of Hollingworth and his book from this really excellent interview, which is where I came across it:

Epimetheus said...

I've been reading more contemporary sci-fi, and lots of it is intensely religious. The yearnings for guidance and transcendence sometimes come through with painful sincerity - once people have been heartbroken to the point that they assume the total malevolence of the human race, they naturally wish there was an entity that did not judge them or criticize them or reject them, an entity that could tell them what they ought to do and love them.

william arthurs said...

When I started out in technology almost 40 years ago, I worked in a milieu in which colleagues anthropomorphised their machines, and conversely treated their own mental processes as register operations. For example, failure to remember some or other fact was accompanied by hand gestures suggesting a spinning tape drive, and the error message that we were so familar with from our mainframe: "Error 5: Can't find file or account."

I never thought this would go further than (fairly feeble) I.T. dept humour. But the idea of poorly-chosen and yet apparently-compelling metaphors has been on my mind ever since. In his inaugural lecture at LSE, Michael Oakeshott reeled off some well-known but misleading metaphors for the state: an estate that needed a business manager (most recently, UK plc), a ship that needed a Great Helmsman... And there was Bill Phillips' famous hydraulic model of the economy, which suggested that a few wise hands on the taps and valves could provide "fingertip control" of GDP growth, employment, inflation, etc.

I think the thing that all these poor metaphors have in common is that they all deny the possibility of order emerging (given certain constraints) from a complexity that no one mind could ever fathom.

Jason said...

"If it looks like what we call 'intelligence", then for all intents and purposes it is intelligence."

The video game definition. Some of the enemies in Super Mario are that kind of AI; they are "intelligent" enough to walk to the end of a platform and not fall off but turn around and go the other way. Or moving to more advance video games, the monsters in first person shooters like Doom in 1995 can follow you around and shoot at you. In more advanced games they have pathfinding algorithms and can navigate mazes. In the Deus Ex games on PS4 the cops can hear you fire a gun outside on certain levels and come after you, and they do a search pattern which eventually forgets you in hide effectively for long enough. There seems to be some "intelligence" so there is, in terms of your suspension of disbelief to make the game enjoyable. But AI as real intelligence for what the government technocrats want to pretend it can do is nonsense.

Alexeyprofi said...

I also don't understand why it has intelligence in its name. The algorithmic executor doesn't really think, it just transforms your input by applying a specific algorithm to its database

Bruce Charlton said...

From what I gather, "AI2 is being imposed in the same way that most things are imposed nowadays: by bureaucratic fiat. As an aspects of someone's current job (say, half day a week) people are being instructed to help prepare (programme, select inputs, "train" etc) "AI" programs to replace them - i.e. your job (or some of it) is to to build the instrument of your own execution.

These AI-systems are never as good as a good human, they are always worse - but the situation is that the replacement of humans by such systems is seem as "managerially inevitable" - non-negotiable; and it is always short-term expedient to make the best of the bureaucratic imperatives (e.g. make a niche in this line of work, be the imposer rather than the imposed-upon) - and for modern, Godless, hollow Men whatever is expedient always carries the day, because there is nothing else...

Thus AI replaces people and the world is made worse - without anybody actually wanting it, and despite everybody being "concerned" about it.

All incentives and sanctions are a part of The System (except the demons who manipulate it); and the bottom line "reality" for modern Man is The System.

a_probst said...

Slightly off topic.

Thomas Sowell once complained that too much attention was paid to artificial intelligence and too little to the artificial stupidity being produced by our public schools.