The Ultra Intelligent Machine

[peter]

Someone once said (and it is a sobering thought) that the ultra intelligent machine will probably be the last invention that humans will ever need to make.

A short time ago in a live event in San Francisco IBM's Project Debator drew in debating contests against award winning human Debator on the subjects of subsidized space travel and telemedicine. The audience voted that, while the computers delivery was poor in comparison to its human opponents, the substance of its arguments (having access to millions of articles and research papers) was superior. In one particularly significant statistic it was reported that twenty percent of the audience said that the machines arguments had persuaded them to change their position on the topic.

Let's consider this - and play a game at the same time. What do you think, are the problems that the ultra intelligent machine would solve in the first week without even breaking a sweat? Would it immediately devise a system to clean up the air, halt global warming in its tracks. Would it solve the overcrowding problem by getting the colonisation of space back on track .......

What do you reckon?

[Ur Dead]

Douglas Adams.

[wayfriend]

It would get rid of the pesky humans, of course.

Don't confuse intelligence with wisdom. That Debator could probably have persuaded another audience of the opposite position with equal ease. Leaving us none the wiser (pun intended) about what is the right thing to do. IBM's AI can't choose the wiser course of action.

The UIM, left to it's own devices, wouldn't do anything until someone figured out what they wanted it to figure out. Hopefully, whoever that someone is makes a wise choice. "Solve world hunger" rather than "make me rich".

If you want to postulate that it is also wise and self-directing ... it would probably choose to do something that half the people would be against doing. E.g. Curtailing emissions to reduce global warming. The reason we HAVE the problems we do is we have people who COVET more than they need. They won't give it up.

Conclusion: the first problem it would need to solve is Greed.

[Ur Dead]

Conclusion: the first problem it would need to solve is Greed.

The problem with that is it was greed that got us to those machines.

[peter]

The machine would have to be sentient to qualify wouldn't it? And the moment it was sentient it would also be self-interested ...........

..... And that could only lead to a conflict of interests couldn't it?

[wayfriend]

Conclusion: the first problem it would need to solve is Greed.

The problem with that is it was greed that got us to those machines.

I knew ... just knew ... that some conservative would intentionally misinterpret my comment to mean "eradicate all forms of ambition". Which I did not say, and which I say now is not my meaning.

[Hashi Lebwohl]

Can intelligent machines make illogical choices like people can? Can an intelligent machine *understand* people's illogical choices?

[Vraith]

Can intelligent machines make illogical choices like people can? Can an intelligent machine *understand* people's illogical choices?

If it is truly intelligent, yes to both. [[Even a non-intelligent thing could do the first]].

There's an important point in there, though. An implied multiplicity [a thing I think fundamental]---Logic is a necessary, but wholly insufficient, ingredient in intelligence. And it has a partner--Intelligence is a necessary, but wholly insufficient, ingredient for logic. [not, perhaps, to execute logic anymore than a calculator needs to understand math to execute addition--but for meaning/knowledge application to reality/existence].

But, to peter's question...I think there are quite a large number of things such a machine could solve in next to no time.

The one most likely to happen and be put into effect [assuming a machine that gives a shit about people], because the PR campaign would be easy and sell itself and the vast majority of folk would be ecstatic in anticipation/participation, would be the curing of almost all diseases/sicknesses...possibly even aging itself [at least massive life-extension by many means].


Some [maybe many] would say "What about overpopulation?" and such...that answers itself, too---because even easier than solving sickness for such a machine are BOTH perfect birth control AND the terraforming tech for the entire solar system [[at least]].

The "evidence" underpinning my belief in the ease of medical solutions is extrapolation from this:
The very first time some people wrote some code that could read papers about a particular biochemical process and turned it loose on existing research, it accurately discovered all the known things AND a couple unknown ones.
In hours [or less] such a machine could read more studies on any given illness [in fact, on EVERY given illness] than all the experts on that/those illness[es] have/can/could in their entire lives.
AND it would UNDERSTAND all that research/all those pages better than all the experts combined.

For the space-stuff, that's even easier---because we already know almost all of it. All that's needed to make it go fast is money and a reason [[and money is just printing---the reason would be decades and more of youth and health---even for the currently old.]]

[Hashi Lebwohl]

Can intelligent machines make illogical choices like people can? Can an intelligent machine *understand* people's illogical choices?

If it is truly intelligent, yes to both. [[Even a non-intelligent thing could do the first]].

There's an important point in there, though. An implied multiplicity [a thing I think fundamental]---Logic is a necessary, but wholly insufficient, ingredient in intelligence. And it has a partner--Intelligence is a necessary, but wholly insufficient, ingredient for logic. [not, perhaps, to execute logic anymore than a calculator needs to understand math to execute addition--but for meaning/knowledge application to reality/existence].

A couple of other questions, then.

Can a machine make an illogical or irrational choice using either "because I felt like it" or "for shits and giggles" as its reasoning?

Can a machine arrive at the *wrong* or *incorrect* solution because it wants to or because it wants to troll us?

[Zarathustra]

We don't need an "ultra intelligent machine" to solve global warming, world hunger, overpopulation, or cleaning the air. We already know how to do those things. We just don't want to. So unless we empower this machine to be a dictator that forces us to do things we don't want to do, its answers would be no more intelligent or desirable than the choices we now face.

Solutions are easy. Getting everyone to like them is not. No machine can ever make a solution that all humans are going to like, no matter how wise that solution is.

The best solution--the one that the most people will like--is (ironically) making people rich. People in rich countries have less starvation, less overcrowding, cleaner air, all the rest. We don't need to make machines smarter. We need to make ourselves smarter, because there is a lot of ignorance, emotion, and prejudice keeping us from accepting the solutions we already have to these problems.

Why would we assume that a machine would want to stop global warming? Machines can function just fine with a slightly warmer earth. Why would an "ultra intelligent machine" have exactly the same values as the political Left? Maybe it would conclude that it's pretty stupid to worry about slowing down technological development to suit organisms and ecosystems that are the antithesis of technological progress.

[Vraith]

A couple of other questions, then.

Can a machine make an illogical or irrational choice using either "because I felt like it" or "for shits and giggles" as its reasoning?

Can a machine arrive at the *wrong* or *incorrect* solution because it wants to or because it wants to troll us?

Again...if it is really intelligent, I'd say yes.
In fact, I'd say humor and other a- or non-logical things self-evolved not simulated/coded would be the first sign a thing was true AI.

[Ur Dead]

Conclusion: the first problem it would need to solve is Greed.

The problem with that is it was greed that got us to those machines.

I knew ... just knew ... that some conservative would intentionally misinterpret my comment to mean "eradicate all forms of ambition". Which I did not say, and which I say now is not my meaning.

Greed, or avarice, is an inordinate or insatiable longing for unneeded excess, especially for excess wealth, status, power, or food.

It wasn't misinterpret, it was exactly what the word means. Wanting an excess
that includes information which is a form of power and status.
How could you make the tools that make the tools that make the tools which makes the basic parts without it?
Even the Curies who study radiation got status and power even through (she) wasn't wanting it. She wasn't rich. Look what her work lead to.
Her status in the world of science is paramount and she knew it.
Of course two Nobel prizes didn't hurt.

Greed doesn't mean money only. It's a slew of goals and object(s) to be attained. It is a main motivating factor. I havn't seen discoveries using hubris.
Nor sitting around the campfire singing Kumbaya.

[Zarathustra]

The problem with defining greed in terms of coveting more than you need is that it renders every ambition greater than bare subsistence into "greed." UrDead is right: we don't need ultra intelligent machines. If greed is defined this way, then greed led to such machines. It's one of many ironies here.

I don't think greed is necessarily bad, no more than gluttony, which is just hunger without restraint. Both of these arise from natural drives. It's like any human drive, something that can be either abused or harnessed for good. The thing that's good about capitalism is that it harnesses mankind's natural tendency towards greed in a fashion that spreads benefit around in the best, most efficient manner.


If you wanted an "ultra intelligent" solution to the problem of greed--an inherently human trait which we can't abolish without denying who we are (sound familiar, Donaldson fans??)--you could do no better than fashion a system that channels this drive to excess into producing an excess that spreads around to everyone _ all without a central organizing, top-down (i.e. "inefficient") human manager.

Until machines provide everything for us, and we still have to rely upon ourselves to provide things, you can find no greater motivator than greed. How else are we going to be motivated to produce more than subsistence existence? But once you eliminate human input, greed is a nonissue. If anyone can have anything without any effort, greed vanishes. Greed is therefore not "coveting more than you need," it is instead more ambition than is necessary to survive. Those who want more out of life than mere survival (or need) are the ones who lift ourselves out of a brutish existence. Only by comparing their success to those still in need does this success seem unseemly. But when everyone has all they need and more, this very same drive will be the only thing left that keeps us moving forward. Without it, we will stagnate in a world of plenty.

There is no difference between greed and ambition except in a 3rd party observer's moral indignation.

[peter]

Given instant access to all of our understanding of the 'way things work' (ie. all of our understanding of the processes of physics, chemistry, biology and all the rest - all of the research papers published to date - the sum of human knowledge) and an ability to sort out the erroneous from the valid, make new and innovative connections, draw conclusions previously hidden or undiscovered by us - could the machine carry out any research experiment or indeed observation based study virtually (if perhaps heuristically) and instantly? If so it's knowledge would simply rise exponentially and infinitely until it in short order became (effectively) ...... well .......God (for want of a better word)? Or would (as an alternative) small errors, uncertainty at the quantum level of prediction, render the system chaotic and subject to collapse in on itself?

[Hashi Lebwohl]

Again...if it is really intelligent, I'd say yes.
In fact, I'd say humor and other a- or non-logical things self-evolved not simulated/coded would be the first sign a thing was true AI.

My thoughts exactly. An AI which cannot tell a joke, or lie, or troll its programmers isn't really an AI, just like an AI which cannot produce a result or arrive at a conclusion which its programmers did not program isn't truly an AI.

[Ur Dead]

Very good point..
Is it super intelligence or just doing what is required of it's program. It can catalog, store, retrieve data and even find bits and piece of the puzzle that
is within the data. But.. There is always a butt...can it discover
knowledge we as bio entities have not discover yet? That takes creativity, the
initiative to surpass just handling and manipulating data that is not known to
anyone or the machine itself. If the machine can formulate new proven knowledge then it a few steps away from full blown conscience.
I say this because it programs and hardware have been built by humans.
And we know there is flaws in those being.

[peter]

Z's point is interesting; the creation of the machine as 'ultra intelligent' and the creation of the 'machine as servant to our needs' are two very different things. Perhaps it would be that the moment you constrain it by making it think in terms of our specific problems, you simply turn it into one of us and place our own limitations on its ability to be ultra intelligent at all.

[wayfriend]

For thousands of years humans have been creating beings who are smarter, more capable, and with more potential than themselves. They are called 'children'. Certainly it would be a rather 'adult' thing to do, if humans as a species do what they have done as individuals, and create something better than themselves.

Speaking hypothetically, where we have eliminated the possibility of creating a 'monster' that only wants to destroy us, enslave us, or marginalize us, and are discussing only something that is smarter than us, we should not fear. The reflexive need to be the apex intellectual is only hubris.

The premise of the question is an AI that can "solve" problems. Let's presume the solutions are good solutions - ideas that we did not imagine but, which now that we see them, we nod our heads and go 'of course!' I don't see anything to fear there. A solution to the depredations greed - the use of power and duplicity to enrich oneself at others' expense - wouldn't take away our incentive to progress, because it is a good solution. We can't imagine what it would be ... but that's the point! And who would want to preserve the capacity to use power and duplicity to enrich oneself at other's expense? Only the ones enriching themselves in this way, of course.

[peter]

If I have it correctly, in the Second World War much advantageous work in respect of problem solving was made by taking experts from different fields and throwing them together to 'brainstorm' a given situation. The fresh inspiration that this kind of mixing of fields can result in, the 'thinking outside the box' that a person from a different discipline can bring to the table, is an area poorly understood and even less capitalised on in the insular world of 'knowing more and more about less and less' of todays scientific world. our machine would surely, at a stroke bring forth the knowledge we currently have right before our faces - but are not 'blue-sky' enough to see.

[Vraith]

There is no difference between greed and ambition except in a 3rd party observer's moral indignation.

I'm gonna go ahead and assume you've borrowed my liking of hyperbole on occasion.
That's a bit like the idea [Avatar likes this one, I believe] that altruism is just a different flavor of selfishness.
Or a bit like saying lung cancer and lung tissue are the same. [[but they aren't...because one kind is sick and the other healthy]].
Neither of those is exact...but that doesn't matter. Cuz your statement isn't true neurologically, psychologically, sociologically, or philosophically.

But, to return to the more topical:

our machine would surely, at a stroke bring forth the knowledge we currently have right before our faces - but are not 'blue-sky' enough to see.

This, in fact, has already happened numerous times...without any ultra-brilliant [or even near- human-intelligent] machines. Or any awareness/consciousness/creativity.
Machines, with just the rules and data, and training have discovered, predicted, and even described how to create, previously unknown biologicals, chemicals, materials, alloys, processes. They've delivered mathematical proofs that no living person can even CHECK. Most of the employers and employees in the financial sector could be replaced right now. They can beat the best of us at most anything---now, or soon---

The problem [such as it is] is that the coding, so far, is highly specialized/narrow. So any particular machine/software, can only do one kind of thing. [and usually only one explicit, particular thing].

[[[that "problem" may really be a virtue, for those with a fear of true AI. I don't think it matters---cuz one fucking idiot with control over such a machine [[[with the processing capacity of AI, but no innate purpose/will/desire/awareness/selfishness or survival coding]]] can, and likely will try, because an idiot, to run things their stupid way and commit any and all atrocities that people fear from a non-human intelligence]]]

That flavor of AI is getting closer, though, and rapidly. The code and deep learning and all are broadening/generalizing. And it's happening fast.