[CraigD]

Ian Pearson, head of British Telecomunication’s futurology unit, seriously predicts that it will be possible to upload your mind into a computer by about 2050. (article)

Unlike robotics researcher Hans Moravec, who concludes that affordable computer hardware sufficient to match the intellectual performance of a human being will be available in the 2020s, Pearson goes a step further with his belief that the technology necessary to copy a human mind from its usual, organic host, into an artificial, cybernetic one, will exist not too long after.

I believe Moravec is correct in his prediction that sufficiently powerful hardware will exist. I’m much less certain that brain imaging technology will be able to copy a human mind by 2050. I don’t even want to consider the social and philosophical implications of such a technology.

[nkt]

Originally Posted by CraigD

upload your mind into a computer by about 2050. <snip>
affordable computer hardware sufficient to match the intellectual performance of a human being will be available in the 2020s,

I don't even think it will take that long. I think AI will be here in 5 years, at least to the level of no more call centers or human share dealers. (Then what will happen to the world of supply and demand? Most humans will be unable to keep up for even five minutes, and will be unable to earn a living!)

If you want a fairly good read that covers a lot of this ground, without too much science or moralising, try Accelerando, by Charles Stross (download from www.accelerando.org)

[CraigD]

I think AI will be here in 5 years, at least to the level of no more call centers or human share dealers

I agree, we’re pretty close or perhaps already there now. Reasonable good limited Turing test-capable programs were running on OTS hardware 10 years ago, and with the current state of voice recognition and synthesizing a computer “matching the intellectual performance of a human being” doing mind-numbingly repetitive tasks like call center work.

What Moravec is describing, though, is akin to an unlimited Turing test-capable machine, with general reasoning skills equal or superior to a normal human being. His argument, which he bases on some pretty impressive comparisons of computer programs an animal neurology, and which Pearson draws from, is that practical computer hardware is currently about 100 times too slow to fully emulate human thought.

The real barrier to achieving either Moravec or Pearson’s vision is, I believe, our inability to image the functioning human (or animal) brain in high enough detail to
copy it into a computerized simulation without understanding well what it is doing, or
understand what it is doing well enough to create a similar computer program.

Although I disagree with them, some very bright folk, such as Roger Penrose, argue that the necessary detail may never be achievable.

[alexander]

well, there are a few things that need to get done before that happens, lets start with what exactly you define as mind. No matter the answer to the first question, you need an interface to the mind, and you have to realize that your brain stores information in neural impulses, and just about any doctor is afraid of messing with the brain, so how do they propose will the brain be interfaced. This one kinda goes with the first one, Yes, you might have enough space to store the data and you might have a way of extracting it, but in the digital world everything is based on a binary system, and if mind implies feelings and emations, would you imagine how will it be possible to store those, what kind of a filesystem, format? 5 years just doesnt seem like the time frame to me. And last but not least, reading and writing in the mind impies different things and procedures, what is the use of storing your mind of there is no way of reproducing it or for that matter really studying it, well, i guess there is some, but still, maybe 50 years, wont argue it, we'll see what happens...

[CraigD]

lets start with what exactly you define as mind

I’ll follow the lead of Moravec and Pearson here, and give the lazy, near-tautological definition “the mind is the effect produced by a functioning brain”. The problem of “hosting a mind,” then becomes that of “emulating a brain”.

you need an interface to the mind

Here, I think, you’ve found the meat of the problem. Moravec doesn’t touch this, because he never claims that actual human minds will be uploaded into computers, just that programs will be written to run on sufficiently powerful hardware that are roughly equivalent in intellectual ability to a human being. Pearson, being a “high level” sort of futurist, doesn’t sweat the details, just assumes that some technique will emerge in the next 40 years that will allow a mind to be read into a computer.

This is where I suspect he’s over-optimistic. None of the current brain imaging schemes - the Wikipedia article “Brain imaging” is a good summary of them – are within a power of ten of the necessary resolution to “read” data from the brain.

In Rudy Rucker’s novel

“Software”
“Software”

, brain uploading is done destructively by robots that carefully cut up as analyze the living brain (disguised as a religious cult that murders people by eating their brains – this is, after all, a cyberpunk novel!). This isn’t far from how fine resolution brain analysis is done now, and might be an accurate prediction how a mind-uploader might someday actually work.

Personally, my greatest hope for a brain imaging with the necessary resolution is with a scheme I first encountered in the (regrettably) late Robert Forward’s non-fiction

“Indistinguishable from Magic”
“Indistinguishable from Magic”

. This scheme involves forming an image of the brain’s structure with atom-level resolution by measuring the annihilation by-products resulting from a minute stream of massive antiparticles (antiprotons, antineutrons, or protronium). Although you wouldn’t expect that beams of antimatter could be used for non-destructive tissue imaging, if you calculate the energies involved, you’ll realize this scheme is actually practical. What’s best, massive antiparticle annihilations produce very rich multi-particle annihilation events, which would allow individual atoms to be pinpointed in 3-d space.

your brain stores information in neural impulses

Your brain certainly processes information using neural impulses, but unlike dynamic RAM, this information doesn’t depend on a constant series of neural impulses to persist. Current theory holds that information is actually store in the brain via a complicated and far-from-understood combination of changes in the receptive and emitive potentials of the synapses at the end of neurons, and by the constant rearrangement of the physical connections of these neurons.

Though this distinction doesn’t make the problem of mind-uploading obviously easier or more difficult, it does imply that a scheme that images neural impulses (depolarization event) alone, such as MEG, won’t work.

how will it be possible to store those, what kind of a filesystem, format?

Compared to the “mind reading” problem, I think the “mind hosting” problem will be much easier. Folk have been experimenting with computer simulated neurons for some time. In everyone I know, interest peaked in the late 1980s – a websearch of the once-popular term for an idealized, simulated neuron, “neurode”, now yields several thousand hits of the term used in brand names, and only a hundred of so old ones like this referring to the once hot subject of “synthetic psychology”. Given the amount of processor and storage likely to be available in the near future, I don’t believe coding a “brain emulator” will be an insurmountable challenge.

[nkt]

£$%"£$% Google Adwords!

I just wrote a long and involved, well-referenced post on this topic. Then as a spot of light relief, I was going to quote

Originally Posted by adwords

Artificial intelligence
Buy Artificial intelligence now! Compare and save at Lycos UK.
Shopping.Lycos.co.uk

but it went to the page, and coming back, my work was gone... Grrr...



I'm surprised AI isn't on eBay yet, though... I've been offered Chinese girls and many more things that I'm sure eBay don't allow, via the adwords! I suppose that proves it isn't on Google yet, either?

[C1ay]

Have you read The Law of Accelerating Returns by Ray Kurzweil yet? It's a really long paper but worth the read. Mainly it's about the exponential growth of the exponential rate of growth of technology. (yes, that's what I meant to say) Some prominent dates from his analysis include the following:

* We achieve one Human Brain capability (2 * 10^16 cps) for $1,000 around the year 2023.
* We achieve one Human Brain capability (2 * 10^16 cps) for one cent around the year 2037.
* We achieve one Human Race capability (2 * 10^26 cps) for $1,000 around the year 2049.
* We achieve one Human Race capability (2 * 10^26 cps) for one cent around the year 2059.

Something to think about....

[alexander]

I know the law of accelerating returns Clay

[nkt]

Originally Posted by C1ay

Some prominent dates from his analysis include the following:

* We achieve one Human Brain capability (2 * 10^16 cps) for $1,000 around the year 2023.
* We achieve one Human Brain capability (2 * 10^16 cps) for one cent around the year 2037.
* We achieve one Human Race capability (2 * 10^26 cps) for $1,000 around the year 2049.
* We achieve one Human Race capability (2 * 10^26 cps) for one cent around the year 2059.

Something to think about....

Or just de-bunk. Once you reach the stage of having a human level AI for one cent (!) or even $1000 the entire concept of money is outdated, almost overnight.

How do we all earn money these days? I do some manual work, but mostly, it is via thought and creativity, isn't it? Hence, once I have 20 machines that are as smart as me, yet 100 times faster, I will be able to design 10000 web pages an hour in my sleep, for a near-zero cost (after capital investment). But I can't get rich, since so can everyone else. And if I offer my services for 1 penny a page, there will still be lots of people doing it for free!

At this stage, money is worth nothing any more, for any job that doesn't involve manipulation of actual things. Those jobs that do involve moving things, they will fall as soon as a robot is mass-produced to do that job (or, indeed, almost any job)

People with the money will be able to build an army (or anything they wish), while those without will have to bootstrap, by having the one AI play games to get more, then use that to build one fabricator, then two fabricators, etc. Accelerando by Charles Stross deals with this scenario, with the fate of humanity being rather bleak. Sadly, I can't fault a lot of what he says. It's a free novel, and worth a read. 7/10 for the story, 9/10 for the technical future content.

But what will this mean for the great unwashed masses? The people who demand wide screen TVs for pushing out babies at tax payers expense? They aren't even fit for cannon fodder now, let alone once Terminator style robots are running around the place!

[pgrmdave]

But...not all the mind is stored in neurons, we also have chemicals that influence our personality, and our feelings. In addition to which, most of our brain is dedicated to our body, which we would no longer have. Imagine the confusion a brain would experiance if it suddenly had no heart to regulate, no information about a stomach coming in, no senses providing more information. I think that the idea of uploading your brain to a computer is simply something that we can imagine but that there is no way of actually doing it.