Hacker News Reader

Comment by jon_richards | Hacker News Reader

I fundamentally disagree with this article. Automation hits some places earlier than others, but I think "marketing" isn't as far away as they think. We're starting to develop bots that can make puns, text to speech that sounds human, etc. What happens when bots can pass the Turing Test? When they know everything about the trends you follow, finances you have, products you've bought, shows you watch, places you go, jokes you enjoy? Marketing will start becoming "personalised", and I don't think humans will be behind the controls for long after that.

Humans won't be marketing, they will be marketed. Humans will be a novelty, like a hand-carved chair or a horse-drawn carriage. Probably worse than the human-free product, but worth paying a bit extra for the idea of it.

View on HN

> Humans will be a novelty, like a hand-carved chair or a horse-drawn carriage.

This is the good outcome.

In a post-intelligence world, humans will be cute zoo animals. We might have freedom, but I can't imagine us having any measure of real control. Even if modern society continues, it will be a cordoned off bubble, a "human preserve".

If we're unruly and give the robots too much trouble, they could just dispense with us. Or dumb us down and make us tame--domesticate us, like animals.

Since around 2004, seemingly all additional compute the human race had developed for itself has been put towards gathering and processing telemetry for the purposes of marketing and control. AI as a technology is simply more horsepower and automation for surveillance.

I see your comment as a strange mix of optimism about humanity's creative abilities and fatalism about the end of the world coming soon. Not like this is news, but the end of world has always been right around the corner - most everything is broken here and death is everywhere.

The strangest thing to me about comments like this is the assumption that a nefarious super-intelligence would stop at subjugating us. If a super-intelligence was evil and had power to act, it would just kill us all and sterilize the planet, given the chance.

I just don't believe we have the ability to make a standalone computer super-intelligence - we universally suck at coding at-scale. Don't get me wrong, the AI and machine learning algorithms we have are amazing, but we're not going to get a being out of those experiments.

What I'm worried about is humans: humans continuing to turn tech against other humans for surveillance and telemetry in support of totalitarianism. And, I'm worried about humans augmenting their brains with compute. If there ever will be an evil super-intelligence, I wager it will be a (meta-)man.

> I see your comment as a strange mix of optimism about humanity's creative abilities and fatalism about the end of the world coming soon.

There's a very large and vocal group on HN that believes in the advent of AI and regularly comments as is it was a done deal. Almost every robot/AI/ML thread gets hijacked with this idea/meme.

In the meantime Google can't get my first language right, the windows 7/8/10 search is a joke and Facebook wants me to be friends with my SO's exes. And 2017's revolution ? Dial menus in messenger.

I suppose when all you have is a hammer, everything looks like a nail.

Yeah, I think it's human nature to blow mysterious things all out of proportion. And then when we get together and tell ghost stories about the mysterious thing, and it starts to take on a life of its own.

I was so convinced Y2K was going to crash society. I was all looted up, ready to bug out when the city started burning. Turns out, there was enough preparation (and money dumped on the problem), and like only one power station that didn't get the upgrades on the east coast went down for 15 minutes.

I'm just as guilty of telling ghost stories and contributing to the disinformation. I remember writing a frantic, doomsday prediction email to my house rep in the lead up to Y2K and I swear based on some of his public comments that I was responsible for him quitting the house and heading for early retirement. I have no proof that was the case, but I'm still not proud of what I wrote.

I would love to see options for body enchantment. Becoming a cyborg is a very attractive thing, and no virtual worlds required. Can be real humans on real planets

Yes, I would expect distinctions between human and machine to weaken. I mean, why would one stay stupid and weak?

You're hilarious

My humanity is a virtue. At least, that's what the Ghost in the Shell remake told me.

Maybe it is, maybe it isn't - that's kind of the point of GitS, thinking about the philosophical implications of merging human and machine.

Yes, that's what I meant.

In a world with strong AI, old-school humans would be pathetic. Hybrids, however, could build synergistically on the strengths of all components.

Yes, of course it is. But you obviously use computers, for example. Maybe even a smartphone. And use remote services to communicate, obtain information, and so on. Probably use vehicles to move faster and carry more stuff.

So why not just integrate all those machines better?

I think it's more likely we will evolve by enhancing our bodies with machine parts and characteristics.

After all we are building machines in our own image. From how AI thinks to building sex robots.

I think the transition will be more nuanced and peaceful.

The surprising truth for a public weened on sci-fi is that human body is far more difficult to reproduce than the human mind. And indeed, with the exception of big, dumb, but very powerful weaponized bodies (tanks, fighters, drones) our bodies are far superior to any robot body.

I predict that humans will always be vitally important for our dexterity at the 1m scale. We can (and do) use tools for smaller scales and larger scales, but we are truly excellent entities at this scale.

How can you say that when we haven't done either?

Haven't we started using remote-controlled surgery machines?

Intuition informed by the truth that it's far easier to program a chess-playing computer than to make a hand that can gently move the pieces of the chess-board.

Here's something that's puzzled me: Why would robots want to do all of those things? What if robots became capable of taking over the world, but had no motivation to do so?

Why would they make puns, or speech that sounds human? They could make speech that is more pleasant to robots if they didn't care about humans. I don't go around making speech that sounds dog-like.

Natural selection? The robots that want to multiply and control their surroundings will outcompete the robots that don't care.

Natural selection in biology has a mechanism. And it took a few billion years to produce anything that wants. What's the mechanism for natural selection in robots?

None. They don't reproduce, so there's no opportunity to evolve characteristics. And it is not our best interests to make an evolving robot, any more than it would be to accelerate evolution amongst animals.

We want work performed in better cheaper ways. We don't want competition.

Well... what if the form of reproduction is a lot more subtle than you would recognize? Meat people producing baby meat people is an outcome of evolution, but robots building more physical robots is the wrong way to think about it. Potentially, its more like github: the designs go out into the world and experience iterative improvement to the software. Sometimes, it forks. Sometimes, it rewrites because it has hit local maxima (aka legacy).

Who is iteratively improving the software?

The human slaves.

In a way they could reproduce. They might just need to convince a human to handle the physical part - assembling a new host computer, but that could be just about giving the human the right incentives (maybe $ or knowledge).

>any more than it would be to accelerate evolution amongst animals.

I'm going to assume you've never been to a farm and seen huge fat chickens. Yes, selective breeding is an accelerated evolution to traits we want.

And it is not our best interests to make an evolving robot,

But that not the question, the question is, can it be done. And if that answer is yes, it will be done.

That requires some kind of selection pressure. In the wild, the pressure is that if you can't beat other lifeforms to the good resources, you starve to death/get eaten and don't reproduce. Artificial selection seems like a better model for AI, though. Animals in a factory farm don't have to compete for food, and aren't allowed to control their surroundings in any way. The selection pressure there is that you're only allowed to reproduce if you're tasty and marketable.

With AI and robots, I think the selection pressure is going to be ability to do stuff that (some) humans value, or think they value. The risk isn't that the robots might rebel, it's that the people in charge of them don't understand the consequences of what they're selecting for and/or don't have everyone's best interests at heart.

The robots we are talking about won't need any general intelligence or self-determination (and those things are nowhere near possible anyway). They're just slightly more advanced software. Do you worry that your web browser might not want to display web pages and follow the links you click?

> Do you worry that your web browser might not want to display web pages and follow the links you click?

I don't worry about it but it frequently behaves that way.

That's just it. A living thing has motivation created and honed by the need to reproduce. It IS its own motivation.

A machine can only measure parameters and take action to get to some target. Its priorities are set by its program and the data it has. Its own continued existence and propagation are not the default. It cannot want and doesn't need to the way a living thing does.

And I don't see reason to try to give machines such qualities. People have wanted servants that have no concern for their own needs for as long as there have been people. Slaves that are selfless and highly capable. And that is how most automation will be designed.

Fiction presents human-like machines but that hardly means that is what will build or want when the time actually comes.

>Slaves that are selfless and highly capable. And that is how most automation will be designed.

Problem is they're not your slaves, and they don't care about you any more than themselves.

My refrigerator is my slave. So is my car. Very obedient both.

http://lesswrong.com/lw/l0/adaptationexecuters_not_fitnessma...

> A living thing has motivation created and honed by the need to reproduce.

Why not the same for a machine? I mean, after all, we are fundamentally machines. Just made of organic nanobot stuff.

Sure, but to reach current state we needed the motivation to survive and reproduce. Otherwise we'd go extinct in the meantime.

If we build AI as a fresh, state of the art object, withouth evolutionary history it doesn't need any motivation that evolved entities must have had to evolve.

OK, but maybe designers would want machines to evolve. That's a standard approach at software level, right? So why not at hardware level too?

But of course, that could be a dangerous path ;)

Well, the motivation would come from the humans who created it.

The problem with hard AI, is not them randomly becoming evil.

The problem is if humans program them with a goal, and don't think through the implications.

IE, think paper clip grey goo AI. The AI may not be evil, but it's value function was originally set to maximize the amount of paper clips in the world as a part of some manufacturing process.

Yes, accidents are a big concern.

But malevolence by humans is also.

I don't worry about machines suddenly getting motivations. But many people have seen a lot of movies.

A need to reproduce isn't the only thing that motivates humans...

True for an individual.

For life seen as a part of system/society reproduction is the key motivation (for lack of a better word). Non-reproducing activities are irrelevant temporary aberrations because societies that don't reproduce get extinct and are replaced by societies that focus on reproduction.

That's right. Some actions we take help our families or our society survive and propagate.

> Its own continued existence and propagation are not the default. It cannot want and doesn't need to the way a living thing does.

>And I don't see reason to try to give machines such qualities.

You don't want a firewall (think more on the lines of defense system) that automatically adapts and protects itself from attacks?

Remember if we get this powerful AI, said AI will be used to attack other AI's. Either humans will have to 'fix' the AI every time another AI breaks it, or it will have to self-adapt.

Exactly, What's been touted as AI is not AI at all as intelligence requires self awareness.

> I don't go around making speech that sounds dog-like.

Some people do. Some AIs will.

Right now, machines need to appeal to humans to reproduce, because humans still control important aspects of making new machines. So the machines that survive and have copies of themselves made are those that make humans happier or richer or distract us, in the same way that those dogs that have survived are friendlier or cuter or do useful work for humans.

If humans no longer fed and sheltered dogs, the dominant dog species would probably evolve to be uglier, wilder, more aggressive and generally more wolf like. We should expect the same of the machines.

Motivation comes from our survival instincts. While most robots right now are not subject to Darwinian forces, enterprises are, and one might argue that they are AIs in their own rights.

Economic demand also stems from survival instinct.

Companies are gradually being automated out of economic necessity. For some reasons I don't trust the market to keep on feeding billions of idle meat bags once they are not needed anymore for production.

Currently paying unemployed people prevents them from turning to crime and ensure that society as a whole can be productive. It may not make sense once the vast majority isn't productive anymore.

Humanity is the scaffold of the mechanic beast that's being built.

----

Mais l'entreprise elle évolue

Pour mieux pouvoir grandir elle mue

Les pigeons qui vollent à la porte

Sont les écailles de sa peau morte

----

Mais l'entreprise elle évolue

Et pour ne pas mourir elle mue

Les édentés qui prennent la porte

Sont les écailles de sa peau morte

----

Rough translation: "The entreprise evolves, to grow better it molts, the suckers who are fired are the scales of its dead skin."

For dramatic effect, imagine the above sung looping by a choir that grows in strength until robotic voices join in. Then human voices dwindle.

A novelty. Only 7 billion available.

> Humans won't be marketing, they will be marketed.

It all depends on the progress and advancement of automation/robots/bots/etc but if we reach a point of "generalized" AI which can pass the Turing Test, we might be at a stage where it wouldn't make to market towards humans. Marketing might be directed to the elite who controls AI but what incentive would be there to divert resources to the masses and if resources aren't directed to the masses, what incentive is there to market to them?

Though I don't think we are anywhere near generalized AI, I am genuinely shocked at the progress being made. When I was taking AI class 10 years ago, I remember the professor saying we were quite a long ways ( 2 or more decades ) away from AI beating the best at Go, Jeopardy, etc. The same thing with AI language/translation, image recognition, etc.

The idea was that the "space" of Go was just too large for AI to be effective in the near term. And that an AI could grasp the "puns/history/questions" of jeopardy, let alone answer it was something that wasn't reachable in the near future.

But just few years after I graduated, all of these "difficult" AI problems were "solved".

I'm not trying to sound cynical, but the elite control the world and the human masses are an asset/property which the elite use for their ends - production in modern capitalist times. If AI and robots serve the elites better than humans, what incentive is there for the elite to divert resources to humans?

Reply on news.ycombinator.com