i didn't say computers couldn't be programmed to learn...ya could write a sophisticated program that could display intelligence without further learning. at any given point in time even a learning computer has a program that could have been written. ya could remove many or most of the brain's neurons and still see displays of conscious intelligence. animals, split brain studies. brainists are unable to articulate reasons for thinking intelligence is something more than the replicable displays we see in humans. there are extant computers that can give a better description of computer intelligence than is provided in the letters. the definition of sentience is ad hoc and obviously false.

logic, los angeles, ca

apare athought for m/c that has a self presevation chip, humans would become surlpus to requirements, the rats in a m/c world, pet is wishfool thinking in an academic world.lets face fact humans still think a light bulb is agood adea, place your bets on the human race evolving to pet status.

michael joseph heavey, cahersiveen>adams towns, m

I refuse to worry about machines getting up to anything nefarious while they continue to need any sort of power source from which they can be disconnected if they get uppity - plugs pull out, batteries remove, solar panels can be covered.....

alexandria, Sheffield, UK

So far, software can only simulate intelligence. That is to say, algorithms which perform logic operations that are in no way related to intelligent thought (I can't stress that enough) APPEAR to exhibit intelligence in their manipulation of input stimuli or output. The most obvious example of this is AI (Artificial Intelligence) in computer games. What's important here is the fact that this appearance of intelligence is just that, an appearance.

The only way true sentience can exist within software is by creating a neuron-by-neuron software simulation of the brain. That is to say: every single basic component of a brain is replicated within data structures manipulated by software. The state of the neurons is managed by the software and the result - theoretically - is the equivalent output of a brain. However, given that this has never been done at all, and that the human brain has 100 billion neurons and 100 trillion synapses, I wouldn't get your hopes up any time soon.

James M, Southport, UK

There is no way to stop the progression to computers that are smarter than humans. To say that computers cannot learn is short sighted. The brain in a box concept ignores the multitude of useful tools that computers have or will have like hands, legs, tenacles and anything else you can think of. Years ago I wrote a program that learned from scratch by pain and pleasure. It was only 9 boxes with X's or O's but it could learn simple games like "Poison Dot" or "Find the L" from experience alone.

Bob Miller, Scottsdale, Arizona

we have tried humans as pets many times but, despite their often playful nature, the problem is with both toilet training and their inability to remain in the same spot in a car park whilst we shop.
this "find the pet" game is a puzzle that many in my race find mentally stimulating in the manner of your countdown tv programme; the one with the very badly dressed human female.

andy ofield, smallville, england

I used to bolster my faith by referring to the fact that a lot of people, far more intelligent than I, could live with doubts that were racking me at the time. I eventually decided that intelligence does not equate to sense, decided that my doubts were valid, and left a large wodge of learnt superstition behind me. M E Duban would do well to ask the simple question posed to him by Galapagos Pete. Mathematics as a religion makes more sense than any religion extant, although the fact that it failed the test of time says even more about religion.

Bill Q, Derby,

we couldn't learn to learn without already knowing how to learn. a robot could be programmed to perform a vast array of behaviours without further learning. in many regards computers and robots reached singularity a long time ago, can calculate and think much more capably than humans--chess, number crunching, among the most impressive displays of intelligence. for most tasks ya perform now ya could write a program that would enable computers or robots to perform the same task more quickly, and with access to a wider range of knowledge. robots don't randomly generate intentions but of course they could be programmed to do so. a sophisticated computer with level five intelligence could outthink a human on any topic but it wouldn't make much difference if that intelligence were trapped in a box, in one regard like being out-thought by a paralyzed person, minus the intentionality. intentionality in humans isn't as mysterious as many think--in fact it's simple and predictable.

logician, los angeles, california

M-E: First, buy a sense of humor. Second, why is math that works the same every time more miraculous than math that gave you different answers every time? I mean, if one object plus another object was sometimes two objects, other times three, other times 23; wouldn't that be amazing? 1 + 1 = 2 every time? Boring. Non-miraculous.

GalapagosPete, Weed, California

AA's comments make generally intesting reading, but she ought fight the temptation to join the bash on creationists, particularly with regard to the rel'ship between math and the Creator's mind. In the first place, the unnamed Texas Baptist school--perhaps Baylor, so-founded, or E. Texas B.U., a USN&WR best in '08?--breaks no new ground. The anonymous Prof, perhaps unwittingly (likely not) follows in a philosphical vein to >er or <er extent consistent with that of Plato, Euclid, Kepler, Galileo, Poincaré, Blake, Hobbes, Pascal, Gauss, Ramanujan, Jeans, etc, and one not divorced from the intellectual meanderings of Einstein, Hawking, et al. Second, AA's estimable training notwithstanding: folks brighter and more accomplished actually believe this stuff, in one form or another (AS Eddington,x HF Schaefer III, and FS Collins come to mind). Given it's bid to 'come, follow, die' (to self), one shouldn't wonder most everone disbelieves Christ, rather, that ones as bright as these yet do.

M-E Duban, Evanston, Illinois, USA

First – what purpose would human pets serve? Animal pets serve a purely emotional purpose, and since these machines will have been designed by people who don't do emotions, neither will they. Social machines might need pets, autistic mechanical geniuses won't.

Second – even being kept as pets is presumably A Good Thing if it delays the ultimate and inevitable extinction of the human race.

But then, that's the wonderful thing about creating fantasy futures: anyone can do it, and they're all equally valid!

Ian Kemmish, Biggleswade, UK

Machines to have a sense of superstition? The last thing we need is a religious-bigot-computer.

Terry Dell, Weybridge, UK