I for one, don't welcome our Robotic overlords (Artificial Intelligence).

I think the enemy would quickly issue Red Spray cans.

A program to play a game (even Chess or Go) or especially to pick a colour isn't real AI.

If your Connect 4 used "real" true AI, you'd have a Nobel prize.

I've been programming over 20 years and worked with so called AI including "Expert Systems". "learning" is a bit misleading to describe how computer systems work

We know lots about what different bits of the brain do, and which bits "light up" according to different thoughts. You can even train people or animals to control computers/machinery by "thought", but the computer is not "reading your mind"

We are as far from understanding the Mind and Intelligence as Descartes was when he finally admitted it wasn't the Pineal Gland and he had no idea how brain and Mind was connected or related. I'm not arguing "Duality" is true, though it may be. I'm pointing out that we are like a non-techincal production worker who could take a computer PCB and make a list of parts and draw the schematic just from the description of each part and how many connections it has.
Quite an achievement.
But it wouldn't tell him/her how to make one from scratch, or a better one, smaller one and doesn't even address the issue of the OS and applications to make it go which he/she has also labourously copied as a table of 1s and 0s from the ROMs and HDD. The function of them is unknown, except some changes to parts of the 1 and 0s seem to have no effect (unused parts), others make it not even start and others have little effect or only occasionally halt it.

yes, but it's like Chess Playing Computers.

Once they thought Chess would require Artifical Inteligence.

It's the Car's PROGRAMMERS that are learning and writing better programs embodying the Programmer's Inteligence.

It's contentious that EVERYTHING a Human does "comes down to experience and rules", that's the Meat Machine argument and it's never been proven.

I've no doubt that based on experience, teams will improve Automous vehicles to the point where they outperform people on average, because machines don't do drugs, alcohol, distractions, too tired etc. But most times when the "unexpected" occurs that the Programmers have not thought of occurs the Machine performance is abysmal, and the human may perform OK (or may not).

No Computer or Program "learns" in the Human or even Animal sense, not even Neural Nets. You can't give a Computer a basic skill in anything and have it learn to be more skilled (like Arithmetic to Calculus), it is preprogramed with measurable goals and possible actions. It can be programmed to measure how closely the goal is met and then have a data relationship between a particular event and possible action. It's totally deterministic could be done slowly with sufficiently complex clockwork. Insects, animals nor humans reliably "learn" this way, though pavlov showed that simple conditioning is possible.

Yet dogs easily learn much more complex things without pavlovian Conditioning. We can't even begin to understand why some parrots and similar don't just Mimic words but appear to understand language.

Corvids, even the ordinary local "eat the road kill" kind show amazing ability to construct tools to get food in the lad, with no prior conditioning. It seems in the wild it's not seen because they don't bother, food is easy enough for them without tools. I've seen video Rooks or crows that could start a car, open a padlock and fish. Our locals decided quite out of the blue one year to use the peanut feeder like a tit.

Guinea Pigs (cavies) we found in tests we did with suitable rewards are absolutely hopeless at mazes. Rats can get good at it. Robots can do it quite well with a preprogrammed algorithm. However the Guinea Pigs (and apparently sheep) despite poor eye sight and less than brilliant sense of smell can tell if someone is a stranger instantly and respond differently to people they know. If they have run out of food or water they do a special squeal only used to humans (a variation of a baby squeak to adult cavy), but ONLY when they see you or hear the back door.

A dog however if needing food, water, or company may howl incessantly for hours, even though its owner is never back before 6pm. The owner will be surprised and disbelieve that their dog howls all day.


The reality is we haven't a clue what Intelligence really is or even how we manage to learn things. Until we know, we won't have the Programming spec for "true AI".

JC 2K3 wrote: »

I'm a determinist, so I just view intelligence as a very complex program.

That's certainly a valid viewpoint, but where is the evidence? It's only a theory. (But you do claim to be an Official Cyborg )

Also given sufficient Memory (if only 64K, paged to disk), even a 4MHz Z80 8bit CPU can run just as complex programs as a Supercomputer, just a lot slower.

My point is that even if these Animals are differently programmed and it is simply a program in the Turing sense, such as could be programmed into a Machine, there is no evidence that we are any closer to Descartes, Pascal or Babbage in knowing what the program might be.

If someone will give me a real spec for Intelligence I will give them a proof-of-concept program in C, C#, C++, VB6, Java or Modula-2 on Windows in whichever of those languages THEY choose in less than 6 months.

Most AI research is about Mimicking useful intelligent behaviour. We have no idea how the parrots, Cockatoos, Rats, Dogs, Chimps, Guinea pigs or people do what they do.

Why are some animals apparently self aware and some not?

Humans are not the only creatures who are self-aware. Thus far, there is evidence that bottlenose dolphins, some apes, and elephants have the capacity to be self aware. Recent studies from the Goethe University Frankfurt show that Magpies may also possess self-awareness.

Magpies are Corvids. It's probable that that Rooks, Jackdaws, hoodie Crow etc are too.
http://en.wikipedia.org/wiki/Self-awareness
http://en.wikipedia.org/wiki/Aware

It's a very interesting subject. Maybe one of the Greatest mysteries of Philosophy there is. If you sieve the Universe will you find a single grain of Justice, Mercy, Hate, Love, Revenge, Envy, Loyalty, Regret, Nostalgia, Dislike or Indifference?

What is the relationship between Creativity, Intelligence, Emotion, Cognition, Learning Ability or Awareness. What are any of these things? How would you put them in a Computer Program?

Why can people learn different things easier than others?

What do IQ tests measure exactly? Or Psychometric tests?

Are there different kinds of Intelligences as some suggest in last few years?

Does a truly AI system need to be self-aware? Is there a relationship between Self-Awareness and a threshold of some kind of Intelligence?

Are we using a too Anthropomorphic approach to measuring Cognition, Self-Awareness, Learning Ability and Intelligence?

Maybe some animals that are rubbish at solving mazes are really good (using Intelligence, not just some special nose) at spotting poisonous food. You get more than one go at mazes and only eat deadly food once, so I know which Skill I'd rather have if I was a small wild furry animal.

More about rooks
http://news.bbc.co.uk/2/hi/science/nature/8181233.stm

One of Aesop's fables may have been based on fact, scientists report.

In the tale, written more than 2,000 years ago, a crow uses stones to raise the water level in a pitcher so it can reach the liquid to quench its thirst.

Now a study published in Current Biology reveals that rooks, a relative of crows, do just the same when presented with a similar situation.

The team says the study shows rooks are innovative tool-users, even though they do not use tools in the wild.

(or they hide them in the wild when they see us coming. Gary Larson and Cow Tools)

He said that the complexity of the task made it unlikely that the crows were solving the problem using trial and error.

He added: "We are aware that the animals probably do it by putting together, in creative ways, things that they have learned individually."

You are making the assumption as to what is possible.

We been writing High Level Language Programs for about 65 years. For over 300 years we have struggled with the modern conceptions of Philsophy.

I can't see that we have made any progress other than the biology of Brain processes which seems to tell us nothing about Inteligence.

We have made progress on on Cybernetics, Robotics, Expert Systems and AI by clever programming which in reality is nothing to do with Intelligence or Cognition. Like Humpty Dumpty in "Through the Looking Glass" the Computer Researchers make words mean what the want them to mean such than even Computer Science Graduates are conned into thinking we have made progress toward "thinking" machines. yet there is no definition of Thinking or Intelligence.

Like Chess programs that can beat Grandmasters or Facial recognition software, very clever programmers have solved the problem by going round it. A computer has no idea what a face is, yet we will recognise a or a random rock on Mars as like a "face".

We recognise chairs we have never seen before from almost any angle and lighting as chair. Somehow after meeting a few chairs we understand "chair-ness". We have no idea how we do it either. Not only that but we can decide to use things we know are not chairs as chairs without thinking consciously.

I'd really love to know how we do all the marvellous things we do and take for granted.

f _ strt wrtng lk ths ftr _ whl y cn rd t qut wll! Hw cn w d ths?

Captain Furball wrote: »

A few reasons why that's not possible.The programming languages are not suited for such a thing, nor is the hardware.

A better language or more suitable HW may make it easier. But neither the current languages or HW would stop a computable specification being implemented. It just might run slower or take longer to implement. Only the Speed of Mechanical Intelligence is HW limited providing enough memory is supplied for the program. How "intelligent" it is however is only a factor of the Program, not the language or the HW.

How do you know this?

Have you a proof?

Anyhow it's possible to have self modifying code. One way of implementing it nearly safely is Virtual functions (Easily done in C, C++ and Modula-2, a bit harder in VB, Java and C#).

Current HW is only inadequate if "thinking" is many thousands of time slower than Human. For a Demo, no-one will care if it takes a week or too for something we do instantly, for the "proof of concept". Then work can be done on improving speed and efficiency.

Assuming that there is any problem. Know one knows. Maybe a 16MHz 486 Is fast enough (though it would seem unlikely). Since we currently have no spec. to work with, any projection of performance is fiction, just guessing.

Before the time Doom and similar engines were written, it was thought FPUs and CPUs 100s and 1000s faster were needed. But the original FPS 3D used a trick and Integer maths thus work on 386 with no FPU or 486SX.

JC 2K3 wrote: »

You say there's no spec for AI, and then use the phrase "rather than AI", this makes no sense.

It's very logical.

Since we have no spec for AI, then no programs can be doing it. Chess was only thought of as something that might need AI till successful Chess software was written. Chess software uses Brute force. I can beat loads of people at chess. Almost any Chess program (even one on an 8bit MPU) can beat me. Grandmasters maybe have all the openings memorised, but beyond that they don't play like a Computer plays. Nor do they play like I do either.

We don't know do build a successful Fusion Powerstation, thus logically any arbitrary Commercial Power station is not a Fusion one. We don't need the missing spec for a Fusion Power Station to prove it, but it's proved because we don't have the spec.

But we can go further. You can examine the source for a Chess Program and you will also conclude that while that have got good enough to beat anyone (and this was true for simple PC programs 20 years ago against non-Masters) they are not using anything that even remotely resembles creativity or Intelligence or cognition or self awareness. This is clear even though we have not figured how to program any of these.

Like when you look inside a real Power station, even though you don't know how to build a Fusion Model, you recognise the plain simple burning fuel, making steam, spinning a turbine models. Though the parts are very large.

Same with Starships. We hope someday we figure how to build one. But if you look at all the current HW we have for space, it's mostly big fireworks. Definitely nothing that can do Warp drive or FTL travel. We have some ideas, but basically no good ones whe can start building. All of our current Technology is Reaction drive (Burn fuel in a rocket motor, Reflect solar wind, accelerate Plasma or fire lasers). It's all useless for Interstellar travel except in a "Generation" Ship.

So far, the silver screen has been quite accurate, except when it comes to robots, says Dr Kaku.

In fact, he says Hollywood has misled us into thinking that smart robots are just around the corner.

Fans of the emotional Sonny from I-Robot, or the depressed and bored Marvin from The Hitchiker's Guide to the Galaxy, may be disappointed to hear that they are more likely to find a cockroach composing a symphony or writing a poem than a robot.

"Stupid cockroaches are smarter than our robots," says Dr Kaku.

"Robots are nothing but tape-recorders, pre-scripted moves ahead of time.

"Digital computers have a hard time learning and that's the fundamental problem. They don't learn new skills."

Renowned physicist Dr Michio Kaku says that the world of science fiction may be closer to reality than fantasy.

So if you thought that invisibility cloaks, time travel and teleportation were for the silver screen only, think again.

Dr Kaku is a theoretical physicist and the co-founder of string field theory, a branch of string theory, often referred to as "the theory of everything".

http://news.bbc.co.uk/2/hi/science/nature/8221009.stm