I had a fairly rational conversation (literally) with a computer working for a deliver company recently, which compared favourably with a similar conversation I had with a human operator working for my mobile operator.

Ok the computer sounded like a computer, and sometimes asked me to repeat more clearly and restrict myself to a small vocabulary. The human operator just repeat by rote the contents of some manual he had in front of him. When I asked the human for further information and he just kept repeated himself word for word, with none of the little bits humans usually add to indicate an ongoing conversation. He had to be reading it, as I don’t believe someone could consistently repeat himself so accurately if he did not have a visual cue in front of him.

I my opinion the computer made a much better attempt at passing the turning test than the human, although both clearly failed.

For sure. I'd rather have a discussion with a voice that sounded like Hawkings telespeaker that could pass an upgraded Turing test than the 40% of the mellow toned human speakers that couldn't, and it's not close.

"I just won a pot, should I play the next one because it could be the start of a rush?"
"yes, of course."
"ooops, look at the time, ... gotta run, ttyl".

I want a Sklansky or a Bill Chen paperclip that pops up when I am about to make horrible decisions.

"Hello, it looks like you are trying to raise 22 UTG full ring.

Would you:
- Want to know the expected EV of this move?
- Want my estimation of opponents ranges?
- Want to know why this is a bad idea, and review pre-flop standard hands as described in the mathematics of poker page 127?

Ty
"

Anyway Piers this is speech generating technology from before 2004 I believe. AT&T has a demo online here: http://www.research.att.com/~ttsweb/tts/demo.php
Paste the following text, and select charles or ******. (charles is good, ****** is funny, most speakers are fairly good)
-----

Hello, how may I help you?

Yes, I can be of service.

No sir, the 35% rakeback deal is only for special customers.

Thank you for calling!

---------
They are clearly working on it, and since then new technologies have emerged.

This.

If we can make a robot who plays soccer as well as a human in 10-30 years like Phil suggests, then we've solved everything that is difficult to model about a human being.

What does Phil think is difficult to make in an AI? Our limited memory and calculative abilities?

Creativity and imagination, intuition, a sense of context and the ability to adapt to new situations, the ability to recognize complex arrays of patterns in disordered situations, facial recognition, understanding of language - inflection and slang and subtext, internal experience (reflection, also visualizations and dreaming), humor, emotion... Do I need to go on? Are you aware of a mathematical model we can plug in to make all these things happen?

You may be right about the soccer since intelligence is as intelligence does, even in humans there is not direct correlation between mechanics and output, no fixed brainstate. That's part of the nature of emergent properties. So the one thing we don't have to be concerned with is the how. This isn't a reference to phil at all, just underlining that the fact AI may not do it like humans is irrelevant.

But both qualifications for passing the turing test are wrong.

Computers dont have to be conscious to partake in a turing test. They also dont need to be of at least human intelligence, they need to to be able to feign human intelligence. We dont need Google to be sentient nor of human intelligence, to have it serve as valid knowledge-building blocks for a machine intelligence. We can model emotions, but why should we? Why would emotions be a prerequisite for general intelligence? If we can create intelligence, why would we want to model it on fallible human intelligence? An intelligence could do without prejudice and desire for war.

Here is the original article, its not very complex: Computing machinery and intelligence

And here the counter to the theological arguments against machines with souls:

The Theological Objection. Thinking is a function of man's
immortal soul. God has given an immortal soul to every man and
woman, but not to any-other animal or to machines. Hence no
animal or machine can think.
I am unable to accept any part of this, but will attempt to
reply in theological terms. I should find the argument more
convincing if animals were classed with men, for there is a greater
difference, to my mind, between the typical animate and the
inanimate than there is between man and the other animals.
The arbitrary character of the orthodox view becomes clearer if
we consider how it might appear to a member of some other
, religious community. How do Christians regard the Moslem view
that women have no souls ? But let us leave this point aside
and return to the main argument. It appears to me that the
argument quoted above implies a serious restriction of the omnipotence
of the Almighty. It is admitted that there are certain
things that He cannot do such as making one equal to two, but
should we not believe that He has freedom to confer a soul on
an elephant if He sees fit ? We might expect that He would
only exercise this power in conjunction with a mutation which
provided the elephant with an appropriately improved brain to
minister to the needs of this soul. An argument of exactly similar
form may be made for the case of machines. It may seem different
because it is more difficult to " swallow ". But this really only
means that we think it would be less likely that He would consider
the circumstances suitable for conferring a soul. The cir-
cumstances in question are discussed in the rest of this paper.
In attempting to construct such machines we should not be
irreverently usurping His power of creating souls, any more than
we are in the c roc re at ion of children: rather we me. in either
case, instrumenis of His will providing mansions for the souls that
He creates.
However, this is mere speculation. I am not very impressed
with theological arguments whatever they may be used to support.

Well, soccer in its form will put the following things to the test:

1. Visual and audial perception.
2. Senso-motoric ability.
3. Fuzzy logic ("estimating with presicion and efficiency")
4. Communication.
5. Referencing (not a necessity, but its an appealing thought that it makes a soccer player better).
6. Learning.

All these working together in union modelled in a single robot that can actually play soccer (ie it is not the size of a house and energy efficient to the level of a human) is heaps more difficult that making some model of our frontal lobes (our ability to 'think').

I've said it before on this board, but it probably bears repeating. From a 'computational' perspective - your lunch chat with your co-workers is probably more impressive than any display of calculative/logical ability ever displayed in science by a human.

All you'd really need for soccer is a topographic model and game theory.

A good set of robots could probably beat a group of humans with little more than the location of the ball and the locations of the human players, which wouldn't be too hard to find.

A simple "soccer-bot" the size of a dog could be much more stable than a humanoid, much faster, and much more powerful. It could also be accurate enough to nail most shots - not perfectly, but with basic calculus at least good enough to get it in the goal box every time (from any position on the field - and deliver it with superhuman speed, using a simple algorithm based on goalie position - probably at least 1 out of 3 shots would score). Intercepting the ball itself also wouldn't be too hard, and a goalie-bot designed for the task would probably be able to keep out virtually any human shot. I mean, superhuman speed and perfect accuracy - it wouldn't take much more than simple calculus to triangulate the observations of multiple units (or even calculate it from just the goalie unit's data) in order to get position and velocity. With position and velocity, intercepting it would be something even normal robots are capable of.

You seriously underestimate the difficulty of modeling our frontal lobes.

I took it for granted we were talking about a human-sized biped, that could go on a team without much difficulty.

If not, I can probably build a modified cannon that will play soccer quite well.

In comparison to picking up, daydreaming and digesting your coffee the modeling is easy.

(If by frontol lobe processes tame_deuces means the ability to plan our acts)

Nightrider's kit already runs faster than Carl Lewis. Computers are able to model pool tables and predict where and with what velocity to hit the ball. We dont need the hardware set-up to proof machines are faster and more accurate computers than us.

Playing robot soccer to proof anything is ridiculous. Robocop can win by skillfully eliminating all oponents including the referee, Deep Blue could lull them to sleep. An AI in the media is always portrayed as an android. The whole idea about the Turing Test for machine intelligence mimicking humans, is 'against' androids.

Wether the machine outputs: Kick the ball at an 43.2 degree angle with a velocity of 50 miles per hour, in the direction countering wind influences and spin. Or he follows this through as a command to issue a robotic kick, Both intelligences are already kicking our butt in soccer. Robotic mechanic acts are redundant to Artificial Intelligence.

No, I was assuming Phil meant simple bots designed expressly for the purpose and tied into the same hard-coded network. Something like scuttling beetles, in my mind.

We're not there yet, a cannon would have trouble keeping the ball. And I assume a "goalie" large enough to fill the entire goal box is outside the spirit of the game, but something similar to a spider with a plate designed to deflect the ball doesn't seem unreasonable to me.

If you mean a biped, well, I still think that would be easier than emulating thought, but I think we're a long way from being able to create humanoid robots capable of performing complex tasks (like soccer).

Yes this is true. I made the mistake of assuming a being who 'plays like us', a common mistake to make.

It is often hard to grasp how extremely VIRTUAL this entire field is. Even if you understand parts, it's like it is all quite unreal.

Well, the entire example was a little moot (ref 46:1's posts on this) anyway, so it's no biggie.

Anyway, our brain is an impressive thing. The ability for cognition is just a small part of its machinery. It is a very unique thing and very special, but from a certain perspective (a computer perspective I guess) there are actually things that are far more impressive and hard to grasp.

Game theory?

Sure. The main thing is just finding the optimal paths for each of the robots on the field to take, given the positions of the human players and the possible paths that the human players can decide to take. That isn't game theory? I imagine a game-theory optimal strategy would be more or less unbeatable by the human players in any case.

hahahahahahahaha

The experts in the field (the Japanese) believe that we'll have better-than-human soccer player robot by the late 30s. They don't believe that we'll have anywhere near something capable of being close to a human domestic servant, let along a functioning being of human intelligence. Why do you think that is?

The extremely parallel and integrative nature of the processing required for intelligence. You can't code that crap, it's orders of magnitude too complex. And you can't just stick a blob of learning goo (assuming we can discover how to make such a thing easily, which is very very unlikely) in a pile and feed it instructions or learning experiences...the human brain has a number of very specific ways in which it generates language understanding, context sensitivity, awareness, and so on. It's an *unbelievably* intricate system with hardware encoded learning capabilities that we can't even begin to understand. Psychologists have a history of misunderstanding the complexity of the brain, which seems bizarre since they're the ones actually studying it...

From wikipedia:
I don't know why you guys are so thoroughly brain dead when it comes to this stuff.

As I've said before, intelligence is an algorithmically hard problem, many orders of magnitude above speech recognition or visual processing, two things which we *still* can't do properly despite having had the processing power for a decade. And the integrative and creative thinking that humans have is so massively parallel and shaped by learning that any intelligent computer is likely going to require nano level changeable hardware such as what happens in the brain. That's a huge engineering challenge that might not be possible with rigid silicon.

Are you operating under the assumption that we need to understand the brain to produce a comparable AI? We don't and the algorithmically hardness is not an issue.

I agree about the flexibilty in configurable hardware (or simulation of) required but there's no reason to think we are limited to silicon, plaasics, organics etc are available.

Re the Turing test: I think Turing would have been satisfied with helpdesk software indistinguishable from people.

A robot about the same size as the goal would be pretty useful (and unexploitable from a game theoretic point of view).

I think some of the AIM chat bots that exist now could be pretty close to a five year old.

Not at all, simply that such a similar level of programmed complexity is going to be required. It's not a case of sticking a bunch of processors together in a parallel configuration and letting it do its thing, or writing an evolutionary algorithm. I think you grossly underestimate how programmed biological brains are.

Right, it'll all magically solve itself. They had evolutionary algorithms in the 60s too, you know. Apart from the douchebags I quoted above, who were experts in their field, similar predictions were made for software; many predicted that programs could be written that build themselves such that the hard work of programming would be solved by the 80s. But there was no silver bullet, despite the predictions of many smart people. The hardness remains as much as ever. I think it's human nature for people to underestimate (or not even comprehend) algorithmic complexity.

Of course, but again you underestimate the difficulty of working with these materials at a nano scale. It's 2008 and we don't even know how basic proteins fold, for crying out loud.

but not programmed by anyone who knew how to do it.

Not magically but yes that's the idea - just like the brain did.

we don't need to how they fold to make use of them