The problem here is, you never really know it until you've got it. Intelligence is hard to define- an AI could be Autistic, or it could be a ditzy social butterfly. It's hard to say.

OP, if this is a new interest for you, have a gander at some Ray Kurzweil books. I may be mistaken but I think his latest book is called The Singularity is Near. I haven't read any of his books, but there's a documentary about him called The Transcendent Man that is pretty interesting, both because it summarizes his ideas and beliefs and gives a pretty interesting glimpse into the mind of the man himself (spoiler alert: his interest in transhumanism and artificial intelligence stems largely from a fear of death). There are also interviews with other leading AI researchers.

If you are "both in awe of the possibilities and utterly terrified at the same time" then Ray Kurzweil is the man for you, as he and his ideas are pretty much the embodiment of those conflicting feelings.

There's a guy in Europe who's a big contender for a 1 billion euro fund going to some sort of high-pay off, high risk research project in Europe. His aim for the project is to simulate the human brain.

Uh, I'll just post the link.

http://www.nature.com/news/computer-modelling-brain-in-a-box-1.10066

Based on some back-of-an-envelope calculations, I have personally determined that it is probably possible to emulate a human mind with current technology (though probably at low fidelity).

It's not actually possible at the moment, however, because the neuroscience hasn't been done yet.

The way I see it, in order to get some idea of how to build an artificial sapience, someone first needs to analyse a working simulation of a human mind. That isn't going to be possible until you simulate a whole human brain, and its associated organs to boot. Once you get a complex model of a real human body working (the brain doesn't work in isolation) you can 'hack off' bits and abstract the fundemental processes to get a model for how a useful artificial sapience might work.

I can't see a single technological step change in so-called 'strong AI'. It takes about twenty years to grow and train a fully functional human being, I can't see it taking any less time to produce the first functional artificial sapience. Once it does come about, though, it should be relatively easy to simplify the technology and mass distribute it.

The question as to whether or not it's possible to fully emulate a human mind seems to me to be a (pardon the pun) no-brainer. I don't think it's possible with von Neumann architecture, but it's surely possible with an architecture that is based on parallel processing. That is, each artificial neuron (or perhaps a column of neurons) is its own computer, and those computers are networked. After all, a neuron is some 10-20 microns in size, and microprocessor tech is already operating on the nanoscale. That gives you two or three orders of magnitude to play around with, at least. Really, all an aritificial neuron absolutely needs to process is a list of signal strengths and the addressess of some 10,000 other neurons.

I know it's not necessarily that simple, but I'm not a neuroscientist. All I know is, the brain is composed of cells, and those cells don't individually engage in particularly complex behaviour. It's rather simple (though specific) behaviour, not currently thoroughly understood, and there's a lot of them. It's something that organic chemistry shat out when it wasn't even trying.

I suppose the trick will be finding some way of scanning the human brain, using some algorithm to mathematically model how different clusters of neurons behave during a single human lifetime, and then apply those models to an artificial brain custom designed for the purpose.

I suppose I think it will be possible to 'save' your mindstate before it will be possible to emulate your mindstate, before it will be possible to create artificial mindstates. In my mind, it really has to occur in that order.

also, dunno where I found this, but;

http://www.fhi.ox.ac.uk/__data/assets/pdf_file/0019/3853/brain-emulation-roadmap-report.pdf

I can't remember where the video is, but a few years ago there was a robot head that spoke to it's creators, before cleverbot was a thing, that expressed fear at the concept of being shut off for travel to another lab. He was unsure if he would be the same when he was turned on again and told them he loved them, while the scientists cried. I'm sure one or two of you say with but I can't remember the video or find it.

>>561

aaaaah I can't find it.

Do you think natural language processing is overlooked in this quest for AI ? And, don't you think that an AI could be built without being integrally based on the human brain model ?

I am probably not very clear here, but the point is : if an AI was able to fully understand (and produce) language (as we humans conceive it), wouldn't this AI (given its designers have prepared "room for it to grow") have the capacity to learn, to teach itself a variety of knowledge, techniques, etc.

Obviously, I understand language processing is far from being easy to achieve, and that to fully work, it would need at least part of the senses a human has. My question is : do you think linguistics could be another way for achieving a strong AI, or a part that the current researches shouldn't ignore ?
Not sure to be very clear.

>>574

Natural language processing would be a byproduct of achieving a strong AI I would think, though certainly it would stand as the AI's main method of further increasing it's interactions and knowledge base of the world after it obtains it. But yeah, to have true natural language processing that's not just an input output algorithm of responses (ala, the Chinese Room experiment, which is crucially related to all of this), requires a true understanding of the meanings being expressed by the words.

Pretty much what I'm saying is that words are just placeholders for the expression of qualitative thoughts held by the speaker, you feel happy so you express such through the statement of the word, and we use these words simply because the direct brain to brain exchange of thought is as of yet impossible. To have language without the thought and feel they represent is to have a language not understood by it's speaker I think.

So I think that the foundation of intelligence that language sits upon would have to be created first, at least to have an artificial speaker that could be said to fully understand it's communications. Otherwise you'll just have a chatbot, which can remember statements and probable responses infinitely, but will always just be an input/output machine and never understand any of what it processes.

If anyone wants a good docu on the subject, check out BBC Horizon's The Hunt For AI.

Probably the best documentary I've seen about AI, covers quite the bulk of the main talking points and is very well produced.

I personally believe that the only way for us to generate a truely sentient AI would be through emulation of the human brain. A powerful computer emulating the interactions of all of the neurons (or maybe even molecules if we got kooky enough) in a brain would be the only way I can possibly see it working, and we are making progress.

Think of the lobsters from Accelerando (if OP likes the idea of the singularity he should read Stross' work, although he seems to read at a pretty lame pace). I personally hypothesize that we'll see major advances on this front within our lifetimes, and hopefully we'll be alive to see the beginning of widespread human uploading. It would suck to miss that by only a few years.

>>582

The neuroscience has to catch up. The computing tech is probably already there. A neuron is ~50 microns wide, could a computer be built that size that could simulate a single neuron with any degree of fidelity? I'm certain of it. If that's possible, it ought to be possible to simulate a whole brain

20 years. Strong AI has always been 20 years away.
Simulating a brain could be a shortcut but the catch is how to interface with it. The neuroscience isn't there yet, so stimulating/reading the right neurons is not currently possible. We can't even use a real brain as an interface, since I don't think it's possible to read/write to all regions of the brain without killing it first but maybe that wouldn't be necessary.

Google built a working simulation of the human brain with 16,000 computers...and all it did was go on youtube and look at cat videos. So I guess Skynet is a /b/tard.

the idea of an AI turning evil is just stupid. a computer does what it is programmed to do.

the risk that a bug make it do something we consider bad might exist though. but it might aswell make it not work. or just misunderstand a word. or something else.

>>605
Hahahaha nice this made me lul

>>529
No, Watson is far more than a search engine. He gained his knowledge by reading wikipedia and other encyclopedias. Natural language cognition is a significant achievement. Watson understood puns and word play and other nuances.

>>574
That''s what Watson does, he understands language.

>>721

A computer can't "naturally" understand language, that's a contradiction in terms. Watson just gives the illusion of understanding, his performance on Jeopardy is proof of that.