Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[PTTG??]

The first AI's with any real depth will start on the internet.

I guarentee it.

No other place is more suited to interaction, for a machine.

Spam-writing software will beat the Turing test first.

[zchris13]

Now this is more like it.

[Toady Two]

I haven't read the whole thread and many of the hugenormous posts so I don't know if this has been said already.

I think the recent craze with the potential for advanced AI is not a thing without precedence. You say this "singularity" or at least the invention of humanlike artificial intelligence will happen within the lifetime of some of us.

Remember like, in the 1970's everyone was sure we would be a space-faring race by the 2000's. Its 2010 now .

In reality its very hard to make predictions especially about things on such a grand scale as space conquest and technological singularities. I believe that more likely than not the advancement of computers will hit a financial or technical limit. Lets say that in 2020 it becomes feasible to build a computer that fully simulates a dogs brain. The machine is would be the size of a ten story building and would cost as much as a manned mission to Mars. It would be deemed a waste of resources and likely advancement would grind to a halt with no one willing to pay the exorbitant price for these machines.

I'm not saying Transhumanism and a Technological Singularity aren't interesting concepts worth discussing. I'm just afraid most people are jumping the shark way too soon.

[smigenboger]

If it's not too late to talk about Corbald's first big post, I'd say the moral argument isn't true in the sense that two entities will always judge something the same way, but it certainly is true that two similar entities would judge two similar events the same way, because morality is a logical choice, and if nothing has changed, nothing will. I think the thousands of fish argument shouldn't be about 'what would AI do compared to me?', since it's essentially, 'If me and me had to decide on a situation, me would do exactly what me would do', which is redundant, but true.

I bring up the vats because we still have no clue what's going on around us. If Joseph has absolutely no idea Rosanne is cheating on him, what makes you think he understands what really makes the sky blue? Perhaps the sky is really a shade of something we can only perceive as 'blueness'. The AI will inevitably become more objectively aware of the world around it, but in the long run, it won't be much more. Consider this, neither human or AI could possibly be aware of all it's atoms at one time, and either are just one tiny part of this universe, so even if it could keep track of itself completely, it still has a long, long way to go to understanding what's going on.

[Sir Pseudonymous]

I haven't read the whole thread and many of the hugenormous posts so I don't know if this has been said already.

I think the recent craze with the potential for advanced AI is not a thing without precedence. You say this "singularity" or at least the invention of humanlike artificial intelligence will happen within the lifetime of some of us.

Remember like, in the 1970's everyone was sure we would be a space-faring race by the 2000's. Its 2010 now .

In reality its very hard to make predictions especially about things on such a grand scale as space conquest and technological singularities. I believe that more likely than not the advancement of computers will hit a financial or technical limit. Lets say that in 2020 it becomes feasible to build a computer that fully simulates a dogs brain. The machine is would be the size of a ten story building and would cost as much as a manned mission to Mars. It would be deemed a waste of resources and likely advancement would grind to a halt with no one willing to pay the exorbitant price for these machines.

I'm not saying Transhumanism and a Technological Singularity aren't interesting concepts worth discussing. I'm just afraid most people are jumping the shark way too soon.

Modern supercomputers already match the human brain in raw processing power, and it's estimated that within a decade a $1000 processor would match that. You're also confusing pop culture sci-fi with actual engineering projections...

[zchris13]

Except most people here are on the pop-science [fiction] bandwagon, dude.  Get real.

[Sir Pseudonymous]

Except most people here are on the pop-science [fiction] bandwagon, dude.  Get real.

Remember like, in the 1970's everyone was sure we would be a space-faring race by the 2000's. Its 2010 now .

Pop-culture sci-fi.

I think the recent craze with the potential for advanced AI is not a thing without precedence. You say this "singularity" or at least the invention of humanlike artificial intelligence will happen within the lifetime of some of us.

Engineering projection.

[zchris13]

Spoiler: SCIFI (click to show/hide)

I have noticed there are a few tranhumanists/singularitarians on these forums, and since most don't have any idea about what those are, I have decided to create a thread to sum up the general concepts of said ideas. Its a bit long (800+ words), and you may actually learn something; YOU HAVE BEEN WARNED! 

   The main concept behind the Technological Singularity is that when artificial intelligence surpasses human intelligence, they will be able to improve upon their own code at an ever faster rate. This will lead to an ever more rapid increase in the intelligence of these AI.

Spoiler: B.S. (click to show/hide)

Assumption 1: The human mind can be emulated within computer software. This occurs either through outright simulation (much less likely) or using programming techniques which are inspired by the brain's structure (very likely).

Evidence: There is, as of yet, no theoretical reason as to why it would not be possible to emulate the human brain. In fact, many AI have been created which emulate various parts of the brain to do specific tasks.

Assumption 2: Technology will continue to advance at a rate which is, at the very least, slightly below linear for the forseeable future.
Evidence: While the exact curve of the growth of technology can be debated (with estimates ranging from slightly diminishing returns all the way up to above exponential growth, depending on the estimator), there is no reason to think technological progess will slow considerably for at least another century.

Assumption 3: The most advanced general AI will be relatively similar to humans in their thought patterns, ethics, morals, and emotions.
Evidence: If we create AI based on reverse engineering of the structure of the human brain, the AI itself would have a mind similar to that of humans.

   These three assumptions are really all it takes for a Singularity event to occur. The first assumption ensures the ability to create general AI. The variations within the second determine the timing of the Singularity event. If, as some suggest, technological growth in exponential, we may see a Singularity event within the next 30 to 40 years. If it is linear, it would likely occur before the end of this century. This is something to keep in mind, the Singularity, if in deed one is to occur, will happen within the lifetimes of some of us who are alive today. The third ensures a Singularity event would actually take off. An advanced, general AI alone will not spark a Singularity. That AI must also have the curiosity about discovery and intelligence humans do in order for it to be motivated to improve upon its own code. For example, a general AI meant for military purposes as a drone would almost certainly not start a Singularity. As such, Skynet scenarios are relatively unlikely due to the vastly different goals of such an AI.

Spoiler: Good (click to show/hide)

The Hardware:
   Moore's Law pretty much sums up this section. The power of computers doubles aproximately once every 2 years; even less for some components. Estimates for the computational power of the human brain vary, but even anti-singularity websites give estimates which are well within amounts achievable through Moore's Law within 2 decades on a supercomputer. Singularitarian websites usually have lower estimates which predict that supercomputers at the current day are capable of running human-level AI and that within a decade, so will $1000 consumer computers. In either case, it is a difference of relatively small quantities of time.

Spoiler: ??? (click to show/hide)

The Software:
   The hardware required for a Singularity event will be around for several decades before the neurological basis on which to create a general AI becomes available. The brain is somewhat mysterious because it is incredibly difficuly to study; especially if one is attempting to study it when it is still running. However, the technology to study the brain on the individual neuron level is becoming available; projects to map brains have already begun. The time frame in which these efforts are completed is probably going to be one of the largest determinants of when a Singularity occurs.

That's how I feel.

[Corbald]

Remember like, in the 1970's everyone was sure we would be a space-faring race by the 2000's. Its 2010 now .

Erm... We actually... ARE... You know... A space faring race... Already... Have been since the '60s. We also have flying cars, and fully automated kitchens. Robots that serve us. And researching on AI isn't grinding to a halt, even though our closest attempts so far ARE the size of office buildings AND cost the same as the gross national output of several small countries.

Just because everyone doesn't have access to these things, doesn't mean the engineers weren't right, they just aren't sociologists. Or economists. Ya know... 'cause they are Engineers...

[Sir Pseudonymous]

Remember like, in the 1970's everyone was sure we would be a space-faring race by the 2000's. Its 2010 now .

Erm... We actually... ARE... You know... A space faring race... Already... Have been since the '60s. We also have flying cars, and fully automated kitchens. Robots that serve us. And researching on AI isn't grinding to a halt, even though our closest attempts so far ARE the size of office buildings AND cost the same as the gross national output of several small countries.

Just because everyone doesn't have access to these things, doesn't mean the engineers weren't right, they just aren't sociologists. Or economists. Ya know... 'cause they are Engineers...

I can't tell if you're being sarcastic or not, because technically, what you're saying is more or less true (although the 6000 ft², $100 million supercomputers aren't being used for AI work, but rather various simulations; AI research is done on high-end consumer grade equipment, from what I've heard).

[andrea]

hardware is probably the easiest to overcome difficulty in AI research probably.
I would imagine coding a software complex enough to imitate life would be a much harder task.

[Corbald]

I can't tell if you're being sarcastic or not, because technically, what you're saying is more or less true (although the 6000 ft², $100 million supercomputers aren't being used for AI work, but rather various simulations; AI research is done on high-end consumer grade equipment, from what I've heard).
[/quote]

It's a tongue in cheek, funny-because-it's-true kind of sarcasm. And http://www.engineeringservicesoutsourcing.com/b/fe/2008/03/can-supercomputer-think-like-brain.html Technically a 'Supercomputer,' though not the office building scale type.

[Eagleon]

Regarding the space exploration argument, it's pretty easy to explain why things haven't progressed past basic research, long-range probes, and hugely wasteful propulsion methods - NASA, and fusion. Fusion is always 20 years away, and NASA is perpetually stuck 20 years in the past. Many government agencies (including NASA) to this day use tape storage, placing the blame of their inadequacies on Sun/IBM/etc. failing to improve the format and readers rather than their own unwillingness to dedicate some time to upgrade to more reasonable standards. Because of the fact that we've seen NASA and other government agencies as the pinnicle of innovation in space research, that the public has the opinion that private space exploration is somehow too difficult to finance or develop (bizzare considering the low opinion many US citizens now have of their various government agencies), and that NASA is a primarily research-oriented organization (not that that's a bad thing, but it doesn't make for fast commercial and practical development), we haven't really made as much progress as we could have.

Computers, on the other hand, have vastly outpaced our estimates on the requirements for AI. As the latter has jumped up repeatedly as we better understand the brain, the former has hypertrophied with a voracious demand for better, faster, more energy-efficient devices. In addition to the incredible leaps of power from conventional and stream-processing, there have been some impressive developments in neuromorphic architectures recently, the most exciting in my opinion being the practical exploration of memristors.

Also interesting is Stanford's Neurogrid. It relies on the fact that you can do useful NN simulations with a very, very high noise level - on the scale of 1 mistake in 10 operations, versus modern CPUs which are engineered to avoid mistakes at all costs - 1 in a trillion, perhaps. The big reason is energy consumption - in any system, if you want to reduce noise, you have to increase signal or shield. The latter is nigh impossible at the scales that CPUs use, while the former requires pumping more energy into the system. So these will likely provide us useful, complex brain prosthetics, instead of ones that will cook you from the inside out Pretty cool.

[PTTG??]

Wait, brain prothestics?
 
The word you're looking for is an everythingextomy.

[Toady Two]

Except most people here are on the pop-science [fiction] bandwagon, dude.  Get real.

Remember like, in the 1970's everyone was sure we would be a space-faring race by the 2000's. Its 2010 now .

Pop-culture sci-fi.

I think the recent craze with the potential for advanced AI is not a thing without precedence. You say this "singularity" or at least the invention of humanlike artificial intelligence will happen within the lifetime of some of us.

Engineering projection.

Its not really a feasible engineering projection because you have to make the assumptions mentioned in the OP. What I mean by space-faring wasn't pop culture sci-fi. I mean what was once the logical next step after the lunar landings. Many subsequent lunar mission, establishment of a lunar base and a manned mission to Mars by the year 2000. None of these things happened. Mainly due to geopolitical changes and a sudden halt in progress and interest.

By comparing the singularity predictions to old space voyage prediction I was trying to prove that it's very hard to predict the course of progress. There might be a thick technical wall that processing power will hit just around the corner.