The Seven Deadly Sins of AI Predictions

[mtdoc]

Interesting article in the current MIT Technology Review: The Seven Deadly Sins of AI Predictions

Nice to hear a sane voice on this topic...

[adras]

Hmm, this is a very complicated topic. I'm scared of AI, too. The main reasons are as follows.

The scenario in IRobot. It appears like Asimov's laws of Robotics are perfect, at least I thought they were. But then that movie found something to cause harm because those rules aren't perfect.

If you look at why there is war in this world you can come to the conclusion that it's just because somebody was treated unfair. The main reason why we create robots is to enslave them. We don't want to create a new live form, we want somebody to do our work. The movie Bicentennial Man shows how hard it can be for a robot to get the same rights as a human.

The way we're currently achieving intelligence is based on neural networks, which try to mimic our brain. So I think at one point robots will be as intelligent as us. And then we will need to find a way to integrate them into our society which is basically impossible. Look at some countries in the world, there are lots of groups which cannot be integrated which results in war again.

We're starting to get to the edge of knowledge the human brain can handle. Evolutionary our brain hasn't changed in the last 2000 years. Yet look at all the knowledge we've got right now. A few hundred years ago, some people had degrees in 3-5 areas. Try to achieve that now. Every area is starting to get so complex that a single person already can't handle it anymore. That's why there are now specialists who focus on a tiny bit of a whole field. Which makes it impossible to see the overall picture. That could be easily solved with AI assistance however. But then we're enslaving a highly intelligent being. Which I consider even more riskful.

We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.

That's a nice saying. But all it's saying is: You can't predict the future. Which is kinda obious.But still we should try to do it as much as possible to try and prevent catastrophies. I'm pretty sure we predicted every danger of nuclear fission before it happend, but we didn't take it serious enough.

All in all, I think we're going to experience a very interesting future, and if we do it right we can solve a lot of problems we've got at the moment. But we need to be careful. We've got a very dark past. And we should try to get to a bright future without harming anyone on the way.

[cdev]

This is AI, but we are more like David - (Not Dave here, David in AI, its mecha protagonist) and I mean, we're in a similar mess, now, more so than we realize  - on multiple levels.

That is one of maybe dozens of reasons  why this is my all time favorite film.

Three Kubrick films have occupied that position at different times in my life..

We live in interesting times.



------------

[ebclr]

People already can go to jail based on computer decision


https://epic.org/algorithmic-transparency/crim-justice/

[Tomorokoshi]

From the article:

Could Newton begin to explain how this small device did all that? Although he invented calculus and explained both optics and gravity, he was never able to sort out chemistry from alchemy. So I think he would be flummoxed, and unable to come up with even the barest coherent outline of what this device was.

I disagree. He would look at the lack of the 1/8" headphone port, and conclude that it was obviously the flawed work of fallible people, without the influence of magicians or gods.

[Brumby]

How can anyone notice the "lack" of something when they never knew such a thing was possible, let alone desirable?


(Yeah, I realise it was tongue-in-cheek.)

[vk6zgo]

I've alway felt that Clarke's Third Law was a bit  condescending.

"Primitive people" at various times in the past,have been confronted by Colonists/invaders from a more technologically advanced civilisation.
The "It's magic" response may happen initially, but it wears off rapidly, until in a very short time they are using the new technology as a matter of routine.

Indigenous people in Australia didn't have steel, but it didn't take them long to see the possibilities of such a material  & incorporate it into spearheads.
They also learnt how to fire muskets, & use other alien technology.

When you find that you can do the same things the invaders do, with their equipment, the "sense of wonder" quickly disappears.

[MrW0lf]

When you find that you can do the same things the invaders do, with their equipment, the "sense of wonder" quickly disappears.

...together with your nice cozy civilization centered around humanitarian values, natural resources and most of your family and pals

[IanMacdonald]

[Zero999]

Hmm, this is a very complicated topic. I'm scared of AI, too. The main reasons are as follows.

The scenario in IRobot. It appears like Asimov's laws of Robotics are perfect, at least I thought they were. But then that movie found something to cause harm because those rules aren't perfect.

If you look at why there is war in this world you can come to the conclusion that it's just because somebody was treated unfair. The main reason why we create robots is to enslave them. We don't want to create a new live form, we want somebody to do our work. The movie Bicentennial Man shows how hard it can be for a robot to get the same rights as a human.

The way we're currently achieving intelligence is based on neural networks, which try to mimic our brain. So I think at one point robots will be as intelligent as us. And then we will need to find a way to integrate them into our society which is basically impossible. Look at some countries in the world, there are lots of groups which cannot be integrated which results in war again.

I can't see this happening any time soon. AI has made great strides in areas such as pattern recognition, game theory and learning a specific task, by trial and error, but general intelligence and actual understanding have barely progressed over the last 50 years! Computers still don't actually understand anything. Even after many years of AI development, they're still just following instructions. With AI, they can learn, based on past experience but it isn't the same as understanding. It's basic trial and error, with no reasoning whatsoever: monkey see, monkey do.

A computer can be programmed and to some extent learn, to play chess to a higher standard, than any human, but it doesn't really understand the concept of the game. To the machine it's just a big mathematical algorithm, with set rules and several inputs and outputs. A human who's good at chess will most likely learn how to play a similar game fairly quickly, to a fairly high standard. A machine will need to be given a totally different program, written by humans because it doesn't understand the game.

We're starting to get to the edge of knowledge the human brain can handle. Evolutionary our brain hasn't changed in the last 2000 years.

I think the human brain stopped evolving, long before 2000 years ago.

Yet look at all the knowledge we've got right now. A few hundred years ago, some people had degrees in 3-5 areas. Try to achieve that now. Every area is starting to get so complex that a single person already can't handle it anymore. That's why there are now specialists who focus on a tiny bit of a whole field. Which makes it impossible to see the overall picture. That could be easily solved with AI assistance however. But then we're enslaving a highly intelligent being. Which I consider even more riskful.

Good point about specialisation, however that also highlights what I was saying about general intelligence. A human with a good degree in a STEM subject will be able to specialise in any other STEM subject, because they posses general intelligence. Given current levels of AI, even multiplied by a factor of 10, reasoning wise, a computer is incapable of actually understanding any of it.

Current AI can help engineers and researches find information more effectively. Image recognition technology may help a doctor diagnose a patient's rare skin condition, but the computer doesn't actually understand anything about the disease or how to cure it. All it can do is help the doctor find information, based on a photograph.

I'm aware that it's impossible to predict the future and that artificial general intelligence may be possible. It's just not possible with current technology or likely to become possible, in the immediate feature, without some huge breakthrough. If machine understanding is invented, then it'll probably be much different to what we expect and will have many applications we can't even imagine.

[Vtile]

......

We're starting to get to the edge of knowledge the human brain can handle. ...... Every area is starting to get so complex that a single person already can't handle it anymore. That's why there are now specialists who focus on a tiny bit of a whole field. Which makes it impossible to see the overall picture. That could be easily solved with AI assistance however. But then we're enslaving a highly intelligent being. ......

Isn't there actually happening just otherway around, the human being (the monkey) is enslaved to machine (the God like).

... In some extend this outsourcing of the thinking have already happened. Thinking which have said to differentiate us from other animals on this balloon.

[Zero999]

......

We're starting to get to the edge of knowledge the human brain can handle. ...... Every area is starting to get so complex that a single person already can't handle it anymore. That's why there are now specialists who focus on a tiny bit of a whole field. Which makes it impossible to see the overall picture. That could be easily solved with AI assistance however. But then we're enslaving a highly intelligent being. ......

Isn't there actually happening just otherway around, the human being (the monkey) is enslaved to machine (the God like).

... In some extend this outsourcing of the thinking have already happened. Thinking which have said to differentiate us from other animals on this balloon.

What do you mean? Can you give any examples of humans being enslaved by machines, other than Sci-Fi?

[cdev]

I think many people overestimate human employ-ability after the kinds of changes we will likely be seeing.

We have to do that, its human nature.

I think superintelligent AI would likely either see us as worth preserving in zoos for scientific or ethical reasons, or as kind of a irrelevancy, they may even see us like we might view rust or mold or termites.

As something they don't want around.

I think the way we treat one another, and the way we treat animals will have a huge bearing on how we would be treated. We can expect similar treatment.

We have to start being good to one another, as well as to animals, as well as to AI's when they emerge.  Only if we are good to them, will they will be good to us.

[Brumby]

......

We're starting to get to the edge of knowledge the human brain can handle. ...... Every area is starting to get so complex that a single person already can't handle it anymore. That's why there are now specialists who focus on a tiny bit of a whole field. Which makes it impossible to see the overall picture. That could be easily solved with AI assistance however. But then we're enslaving a highly intelligent being. ......

Isn't there actually happening just otherway around, the human being (the monkey) is enslaved to machine (the God like).

... In some extend this outsourcing of the thinking have already happened. Thinking which have said to differentiate us from other animals on this balloon.

What do you mean? Can you give any examples of humans being enslaved by machines, other than Sci-Fi?

My first thought on this brought to mind the "smartphone generation" where people are continuously plugged in and adapt their normal activities and social interactions to fit around that technology.  Are these people "enslaved" to their devices?  I can see how some observers would say "Yes".

The counter argument is that they have simply adapted their behaviour to make use of the facilities provided by the technology.

As I see it, the pivotal issue is whether their behaviour is voluntary or not ... and that can come down to a range of personal qualities.

[cdev]

I think the simplest explanation, the economic one, is the best one.

People have no choice but to work. They have to eat. So the more machines do, for less and less, the harder and better and longer people will have to work to make the same amount of money as they did before.  And the more education they will need to have. or perish.

Do you know the story of John Henry?

https://en.wikipedia.org/wiki/John_Henry_%28folklore%29

What do you mean? Can you give any examples of humans being enslaved by machines, other than Sci-Fi?

My first thought on this brought to mind the "smartphone generation" where people are continuously plugged in and adapt their normal activities and social interactions to fit around that technology.  Are these people "enslaved" to their devices?  I can see how some observers would say "Yes".

The counter argument is that they have simply adapted their behaviour to make use of the facilities provided by the technology.

As I see it, the pivotal issue is whether their behaviour is voluntary or not ... and that can come down to a range of personal qualities.

[vk6zgo]

I think the simplest explanation, the economic one, is the best one.

People have no choice but to work. They have to eat. So the more machines do, for less and less, the harder and better and longer people will have to work to make the same amount of money as they did before.  And the more education they will need to have. or perish.

Do you know the story of John Henry?

https://en.wikipedia.org/wiki/John_Henry_%28folklore%29

What do you mean? Can you give any examples of humans being enslaved by machines, other than Sci-Fi?

My first thought on this brought to mind the "smartphone generation" where people are continuously plugged in and adapt their normal activities and social interactions to fit around that technology.  Are these people "enslaved" to their devices?  I can see how some observers would say "Yes".

The counter argument is that they have simply adapted their behaviour to make use of the facilities provided by the technology.

As I see it, the pivotal issue is whether their behaviour is voluntary or not ... and that can come down to a range of personal qualities.

A lot of discussion around the impact of AI & other sophisticated technologies seems to be driven by people without much knowledge of how industry works.
The image seems to be of thousands of people making everything by hand, whereas the reality is that "dumb automation " has taken over many jobs already.
 
This technology is a hard act to follow, as AI may only offer a very small increment of improvement over whatever "clunky" set up is doing the job now.

Slightly off topic:-
Many people get wildly excited about 3D printing, assuming it is the wave of the future.
Maybe it is, but only for very small quantity production, or for manufacturing patterns for use in casting or extrusion, which can produce hundreds to thousands of products in the time it takes to make one with 3D printing.

[Brumby]

Slightly off topic:-
Many people get wildly excited about 3D printing, assuming it is the wave of the future.
Maybe it is, but only for very small quantity production, or for manufacturing patterns for use in casting or extrusion, which can produce hundreds to thousands of products in the time it takes to make one with 3D printing.

It is ... with today's processes.

[vk6zgo]

Slightly off topic:-
Many people get wildly excited about 3D printing, assuming it is the wave of the future.
Maybe it is, but only for very small quantity production, or for manufacturing patterns for use in casting or extrusion, which can produce hundreds to thousands of products in the time it takes to make one with 3D printing.

It is ... with today's processes.

It will need a huge breakthrough to compete with extrusion, in terms of throughput, & casting/ forging in terms of material characteristics.
3D printed crankshaft anyone?

[Zero999]

With smartphones, there's usually a human at both ends, unless the user is just playing games and some people get paid a lot of money, just to game.

I don't know what all this gloom and doom is about. Technology has been stealing jobs from humans for around 200 years now, yet living standards have generally improved over the same time period.

What if labour was free and humans didn't have to do anything? Would the economy collapse? I doubt it. The Roman empire lasted a long time, propped up by slave labour.

Even if there is some miracle breakthrough resulting in general AI, which is self-aware, I don't see why it would be implemented for menial tasks. That would be counter-productive. Humans get bored of repetitive tasks and that's one of the reasons why they were automated in the first place. I doubt we humans would ever want self-aware AI for anything. AI with general intelligence would be good: it could solve the world's problems, without getting bored, tired or frustrated, when something is difficult, but emotions would get in the way.

As far as war breaking out is concerned: that can only happen if the borg is physically capable of doing it. It can be more intelligent, than any human, but it can't harm anyone, unless it's given the necessary hardware to do so.

[Vtile]

......

We're starting to get to the edge of knowledge the human brain can handle. ...... Every area is starting to get so complex that a single person already can't handle it anymore. That's why there are now specialists who focus on a tiny bit of a whole field. Which makes it impossible to see the overall picture. That could be easily solved with AI assistance however. But then we're enslaving a highly intelligent being. ......

Isn't there actually happening just otherway around, the human being (the monkey) is enslaved to machine (the God like).

... In some extend this outsourcing of the thinking have already happened. Thinking which have said to differentiate us from other animals on this balloon.

What do you mean? Can you give any examples of humans being enslaved by machines, other than Sci-Fi?

I did use a bit wrong wording. "In some extend" should be "In tiny extend" to be more descriptive.

The cases where the decision making of the process is given to a certain system is what I did mean with the "some extend..." such things exist ie. Maeslantkering the Rotterdam floodgate. This is of course in several magnitudes smaller case what it would be (or will) if and when the real AI is involved.

Another thing is if you do want to extend the philosophical idea and count a bureaucracy (heck this is hard word to spell in english) as a machine like structure.

[TerraHertz]

https://chess24.com/en/read/news/deepmind-s-alphazero-crushes-chess

DeepMind’s AlphaZero crushes chess

20 years after DeepBlue defeated Garry Kasparov in a match, chess players have awoken to a new revolution. The AlphaZero algorithm developed by Google and DeepMind took just four hours of playing against itself to synthesise the chess knowledge of one and a half millennium and reach a level where it not only surpassed humans but crushed the reigning World Computer Champion Stockfish 28 wins to 0 in a 100-game match. All the brilliant stratagems and refinements that human programmers used to build chess engines have been outdone, and like Go players we can only marvel at a wholly new approach to the game.

That it's Google/Alphabet doing this, really worries me. Eric Schmidt (CEO of parent company Alphabet) has a very evil worldview. He's been going on lately about how hard it is to program an AI to recognize 'truth' (which he defines as conforming to leftist-Globalist views), and as a result Google/youtube is currently hiring 10,000 human censors to rate youtube content. I'm sure he would much rather have been able to program a machine to achieve his goal, and will keep working on ideologically constraining an AI until he doesn't need those pesky humans and their unreliable opinions, that differ from his.

[cdev]

Globalization and the aggressive and dishonest creation of a harmonization downward of environmental standards, workplace protections, standards of living and wages by creating a single common labor market for corporations and a gutting of democracy with FTAs and its race to the bottom near term will cause orders of magnitude more disruption than AI.

[Buriedcode]

I have noticed that, the 'fears' about AI - specifically about creating a 'intelligence greater than us' as opposed to the fears about job losses - tend to come from the older generation, and those who really don't begin to understand what we call artificial intelligence.  Either that, or see it still in sci-fi terms, rather than the modern 'pattern recognition' algorithms of today, as the article alludes to.

Algorithms are everywhere and have been for decades, and really can determine life or death situations. Think about modern Aircraft, railway control systems, I'm sure most don't see these as examples of AI, but simple control systems and electronics.   We put our trust in these systems every day and they very rarely let us down (when they do, tends to be human error anyway).  Modern 'AI' is pretty impressive, and has already moved into to many areas of life but mostly in terms of voice recognition, face recognition and using 'big data' to try and tease out hidden information (like political persuasion from facebook likes).

Also, much of this is about predicting the far future - something we are notoriously bad at.  As with any discussion about 'evolution', or technological advancement, there is often a tendency to assume what is considered 'advanced', which is an assumption about what the future will be.  Like any discussion of human evolution.. ask people to state what they think an 'advanced' human will be like and you'll get all sorts of bizarre answers like bigger brains, or telekinesis, like somehow evolution follows what we think rather than environmental pressure.

The only 'danger' I can see about AI is humanity giving it more power without appreciating what it actually is, or without fail-safes.  Even today there are stories where AI fails terribly - often because of human error, such as only training face recognition on a certain race, leading to 'racist' results.  Ironically it is those who see AI as 'magic' or ''dangerous' who will allow it to be used in critical situations without controls.

[BrianHG]

Here you go:


Enjoy!

[cdev]

I have noticed that, the 'fears' about AI - specifically about creating a 'intelligence greater than us' as opposed to the fears about job losses - tend to come from the older generation, and those who really don't begin to understand what we call artificial intelligence.  Either that, or see it still in sci-fi terms, rather than the modern 'pattern recognition' algorithms of today, as the article alludes to.

Algorithms are everywhere and have been for decades, and really can determine life or death situations. Think about modern Aircraft, railway control systems, I'm sure most don't see these as examples of AI, but simple control systems and electronics.   We put our trust in these systems every day and they very rarely let us down (when they do, tends to be human error anyway).  Modern 'AI' is pretty impressive, and has already moved into to many areas of life but mostly in terms of voice recognition, face recognition and using 'big data' to try and tease out hidden information (like political persuasion from facebook likes).

There is a major concern that we're automating the human race out of employment, but that's for a reason, because we are.  You never mention how all those people will eat, or find water, or if they even will. Nobody is, actually, and it's because we're setting up a system to deny it to them. Which has nothing to do with automation or AI except that its a reaction to the impending changes.

The world of the future will be one of abundance of all kinds, but those currently in power are trying to create a false scarcity by triggering a race to the bottom for the remaining jobs.

The labor saving tools we build now save lots and lots of labor. Should the wealth they create be shared, and not concentrated, or not?

 Right now, that's being prevented, by eliminating democracy except in name only.

Thats because knowledge and technology is naturally generous and that generosity is seen as a threat by the insanely hierarchical system whose core principle- if it could be said to have one is the concentration of wealth and power in those already wealthy and powerful. (So, it won't ever rationally handle what's happening. That's not its way.)

Yes, we're saving lots of labor. Sure, a relatively tiny number of jobs are created building things like web applications that are designed to run themselves.

We build more of them and better and better tools and they also get segmented according to the global value chain ideology. That leaves low skill workers in high skill countries in an impossible situation. Before we know it, billions of people are going to be out of work in a way that has never happened before. They won't be able to just take less money and get another job at lower pay. They just wont be needed.

Also, much of this is about predicting the far future - something we are notoriously bad at.

Yes, and we keep getting worse at it too as the rate of change increases. Because its human nature to think that the rate of change in the future will be proportionate to the rate of change in the past, when in fact the two curves are best approximated as more like perpendicular to one another.

The biggest problem as I see it is the reaction thats being forced on the world which has occurred due to the positive changes in the late 20th century. There is a project now, a global one thats trying to push us back to a form of feudalism. Think "divine right of kings" feudalism. Thats because corporations, who are writing the rules for countries ow, made a deal with countries that necessitates them treating al countries the same and as if all governments were legitimate, when they are not.  Also, around 23 years ago they started putting agreements into place that make democracy more and more impossible.

They are very sophisticated in their methods but what they are creating is still feudalism. With all that implies.

Its goal is keeping the planet divided.

As with any discussion about 'evolution', or technological advancement, there is often a tendency to assume what is considered 'advanced', which is an assumption about what the future will be.  Like any discussion of human evolution.. ask people to state what they think an 'advanced' human will be like and you'll get all sorts of bizarre answers like bigger brains, or telekinesis, like somehow evolution follows what we think rather than environmental pressure.

The only 'danger' I can see about AI is humanity giving it more power without appreciating what it actually is, or without fail-safes.  Even today there are stories where AI fails terribly - often because of human error, such as only training face recognition on a certain race, leading to 'racist' results.  Ironically it is those who see AI as 'magic' or ''dangerous' who will allow it to be used in critical situations without controls.