Andrew Smart, AI, Consciousness, Turing Test |519|

#23
Unless the brain only modulates consciousness arrising from an external source (a simpler model because it doesn't require consciousness to be uprooted at birth/death).

In that case the primary obstacle of any AI aiming to replicate brain bevavior is one of recieving/decoding (and willingness of an external consciousness come to think of it). Acid trip will be a breeze if they can nail that (not holding my breath).
I deeper thought would be how it is that most people are thinking with their asses, yet shitting out of their mouths
 
#24
agreed. it might also be that he wrote this book before the full google indoctrination took effect :)
It is more like this "you know you know you know" sack of shit guy doesn't give a damn about the truth. This is one of the most insincere guests that you have ever had on your podcast. I hope he reads this, would love to hear his opinion of how he waffles worse than a breakfast at a midnight diner. This guy is a fucking joke. If he has a problem with it, please email me at [email protected]. Total fucking joke, man.
 

Alex

Administrator
#25
Assuming that human consciousness does survive death, and\or that it can access some sort of global consciousness, then it seems to me that any AI equivalent would have to have whatever we have in our brains that does this.
I think this depends on how you define " human consciousness." I don't think andrew is willing to assume that human consciousness survives death
 
#28
I think the AI folly was there right from the start of computers. Computers that filled a room but ran incredibly slowly with tiny memories were extolled as "giant electronic brains" because they could add up a column of figures much faster than a human could.

With enormous effort (human intelligence) computers can now recognise speech to a limited degree, and produce verbal output - so again that can wow the masses so that they believe the computers are artificially intelligent.

The bottom line is, whether this is a question of what piece of technology can wow enough people, or whether there is a deeper question.

Thinking that a computer that can do X,Y, Z must be conscious and intelligent, is rather similar to the reverse fallacy of B.F. Skinner - that you could ignore the mind completely and just look at the brain in terms of its inputs and outputs.

David
 
#29
Assuming that human consciousness does survive death, and\or that it can access some sort of global consciousness, then it seems to me that any AI equivalent would have to have whatever we have in our brains that does this.
Insane assumption, no? I mean, we have no evidence for any of these assumptions and no means to test for them. While we shouldn't delude ourselves to implying untestable things don't exist, why do we feel anythimg is outside?



Given this, there could be the equivalent of an acid trip - though it would be unlikely to use the same molecule - LSD!
A robot could probably be constructed that appeared to react to LSD. However, I don't feel it would be sentient.

I don't believe in algorithmic consciousness, and I think this boot is very relevant to this discussion:
An algorithm certainly could simulate any aspect of consciousness you know how to measure without satisfying us.
Note that it is written by an AI expert.
Does an expert have magic insight into something we do not know how to measure? I like that it looks like he comes our in the negative. Maybe he could create the LSD simulation though.
 
#30
I think the AI folly was there right from the start of computers. Computers that filled a room but ran incredibly slowly with tiny memories were extolled as "giant electronic brains" because they could add up a column of figures much faster than a human could.
Why was it a folly? Turing simply defined an approach allowing ppl to move forward. A simpler well defined problem that was close to solvable after 75 years of effort.

With enormous effort (human intelligence) computers can now recognise speech to a limited degree, and produce verbal output - so again that can wow the masses so that they believe the computers are artificially intelligent.
We made progress towards Turing's goal, yes.

The bottom line is, whether this is a question of what piece of technology can wow enough people, or whether there is a deeper question.
The question we can't answer remains the same because we can't directly measure what we mean by
consciousness. We have no means of moving forward. I cannot directly measure your qualia, how you see red but I can try to define markers. Perhaps when you see red, a certain part of your brain lights up.
Insane assumption, no? I mean, we have no evidence for any of these assumptions and no means to test for them. While we shouldn't delude ourselves to implying untestable things don't exist, why do we feel anythimg is outside?

How can I know whst you perceive,
 
#31
Why was it a folly? Turing simply defined an approach allowing ppl to move forward. A simpler well defined problem that was close to solvable after 75 years of effort.
Unfortuntely Touring did not live to update his test in the light of the unbelievable growth of computer capabilities and the equally unbelievable growth of the internet.

When I was a university, the mainframe computer serving the university had 2 megabytes of main memory and ran with a cycle time of something like 1 microsecond!

If by 'AI' we are referring to an intelligence like ours - or at least on that scale - then I doubt we have made progress. This is most clearly demonstrated when we look at a task that really does require human intelligence - such as driving a car. For years we were promised driverless cars - reams and reams of newsprint extolled how this would change society - and it would have if that goal had been feasible. Now hardly anyone mentions that goal. The claim that this technology was feasible was ultimately based on the mistaken idea that anything that a human could do, and which simply required processing data and sending out control signals could be done reliably by a computer.

I'm old enough to have been an adult in the 1980's when the previous AI hype happened. It was exactly the same, mountains of hype (which I believed for a bit) ..... and then nothing.
 
#32
Unfortuntely Touring did not live to update his test in the light of the unbelievable growth of computer capabilities and the equally unbelievable growth of the internet.
I think, if Turing had lived out his natural life, he might well have discovered Cook's theorem. Cook's theorem basically says if an algorithm can be found to solve the 3-satisfiability problem in polynomial time, then P=NP where P is the set of problems solvable in polynomial time, NP the set of optimization problems where candidates can be verified in polynomial time. The theorem was proven using Turing machines. It is the fundamental theorem of pre-quantum computing. The hardest problems are a whole family of problems like the traveling salesman problem that are polynomial reduceable to 3-SAT.

I think, he'd probably have lost interest in his test. If he did maintain interest, he'd likely have focused
on a mathematically deep subproblem but I suspect he'd get more curious into attempting to prove NP isn't P.

When I was a university, the mainframe computer serving the university had 2 megabytes of main memory and ran with a cycle time of something like 1 microsecond!
True but problems exist that don't scale. This is what NP-completeness is all about. You also have unstable systems (choas systems) like weather prediction modeling the real world.

If by 'AI' we are referring to an intelligence like ours - or at least on that scale - then I doubt we have made progress. This is most clearly demonstrated when we look at a task that really does require human intelligence - such as driving a car.
Yes, we have reached a point where we can almost pass the Turing test, in the last decade, long after Turing's natural life might have passed. But alive today, might he venture into Quantum computing? What limits lie therein?

Yes, we can already use computers to do things we can't; e.g, computer PI to a billion digits, predict weather for a week, etc. Yes, perhaps they can have a new intelligence beyond ours at some point but tyey already see things we can't. . Where are Asmov and Philip K Dick when we need them?

I'm old enough to have been an adult in the 1980's when the previous AI hype happened. It was exactly the same, mountains of hype (which I believed for a bit) ..... and then nothing.
It wasn't hype. All that meaningless research lead to the advances you allude to.

Specifically : neural networks, expert systems, speech recognition, image interpretation and the like.

A new frontier is, we are starting to understand the human brain. Neural science is advancing like crazy.nI can't see blue like you do perhaps but maybe I can look at your brain and know how close my blue is to your blue?
 
Top