Thursday, October 20, 2011

#SS11 Ray Kurzweil on Watson, consciousness, And the Paul Allen rebuttal

Leading futurist, Ray Kurzweil rang in this year's Singularity Summit like a boss. The "rightful heir to Thomas Edison" -Forbes.

Ray stood firm in his predictions, such as human level artificial intelligence by the year 2029. At the summit Ray laid claim to continued evidence, and cited US government predictions of continued exponential growth in price/performance of hardware, and even more so in software. The 2011 Singularity Summit was intended to be inpart a commemoration of IBM's Watson victory over Ken Jennings on Jeopardy. However, the calm before the 2011 Singularity Summit was interrupted in what appeared to be a pre-emptive strike from Microsoft co-founder Paul Allen. Mr Allen's guest blog entry on Technology Review titled The Singularity Isn't Near went viral in the futurist community just days before this year's summit.

In the "spirit of point and counter-point", this would not go unmentioned.

Ray Kurzweil started off the summit brilliantly with his talk "From Eliza to Watson to Passing the Turing Test". As always, he is a man comfortable and confident in his projections of the future. Watson's victory, A[G]I, and consciousness are all brilliant topics. - All a perfect fit for Ray Kurzweil's upcoming book How the Mind Works -- and How to Build One
Indented quotes are excerpts from Ray Kurzweil's speech - Singularity Summit 2011 @ 92 Y NYC Oct 15, 2011
"I'm writing a book now - How the Mind Works-- and How to Build One. I talk about this issue and I refer to the mind as a brain, because the mind is a brain that's conscious. So it brings in this philosophical issue.
My seclusion is that you've got to have faith. Otherwise, we couldn't get up in the morning without some notion of who and what is conscious, and where our conscious comes from. It's actually a very illusive subject, even the question of what I'm conscious of. I mean, I came over this morning from a hotel. How much do I remember of that voyage over here? I remember that I made the voyage, at least I can logically deduce that I must have made it over here, but what do I remember about it? Almost nothing. ...
We do associate memory with consciousness, that is why we think we are not conscious during anesthesia, because we wake up and we don't remember anything, so we assume we weren't conscious, but that's actually not proved at all. I don't remember this trip over here, or very many details about it. Does that mean I wasn't conscious?
It's actually a very good question. I'm not sure of the answer. So, there continues to be debates, which is something we encourage at venues like this; Singularity Summit. ... In terms of talking and thinking about the singularity, theres no one path in terms of the right way of thinking about these things. I think the underlying observation is the ongoing exponential growth of information technology, and were that will lead us."
The bulk of much criticism today still attacks Ray Kurzweil's exponential arrival time projections. Ray, a master of his craft, professionally presents up to date evidence supporting his extrapolations based on the Law of Accelerating Returns. The growing professional consensus is catching up with Ray's predictions. Within most of our lifetimes, the consensus of when a computer could pass the Turing Test went from 200 years into the future, to today's projection which is within 25 years. Which is very close to Ray's year; 2029.

Ray started writing an essay to address the criticisms of his 1999 book The Age of Spiritual Machines. His essay turned into the book The Singularity Is Near (2005) . These same criticisms, which he has answered already, still haunt him today. Those with locked linear intuitive viewpoints, to neo-Luddites and religious fundamentalists; all often don't agree with various predictions presented by Ray Kurzweil. Much of the mainstream criticism in particular Ray says includes; consciousness, the end of Moore's law, and that exponentials can't go on forever.
"One of the objections in Paul Allen's article is that exponential growth continues, until it doesn't. Well, thats true depending on what it is your measuring. If you were measuring the power of vacuum tubes; the price performance of vacuum tubes, ... That continued and continued until it hit a wall, and it stopped. So if you were looking at just that one paradigm it continued until it didn't, but what happened was it created research pressure to create the next paradigm, when transistors came along."
"We have already started the 6th paradigm,which is 3 dimensional computing. ... There are arguments of what the 7th paradigm might be, but you don't even have to go there, because the ultimate computing at the 6th paradigm, basically computing in 3 dimensions at the molecular level; would give you a two pound computer that's trillions of times more powerful than the human brain."
"It's typical of a lot of commentary around these issues, that Mr Allen brings in sort of what I would call de novo arguments about the singularity, and not really having familiarized himself with the literature. ... he alludes to my book The Singularity Is Near in the title [The Singularity Isn't Near]... but acknowledges in the article that he hasn't read the book."
WATSON - IBM Super Computer and Jeopardy! Winner
Ray Kurzweil, was apparently satisfied with the IBM Watson performance, but also talked with an optimist's caution about the machine's existence; it being both impressive, yet limited. He had said it was not to be viewed as a nonthinking emulation of the human brain, rather to be viewed as a powerful computer today, with amazing future potentials.
"...we are already translating these conceptual insights into biologically inspired paradigms, now Watson is not, by any means, a sort of un-thinking simulation of the brain, but these biologically inspired paradigms have seeped into Watson."
"[Watson] got its knowledge from reading natural language documents. ... It was not hand coded by scientists."

Watson is an impressive AI milestone. Watson has advanced (yet not human) control of the English language. Had it been human control of the english language, well then Kurzweil would have been too conservative on his predicted year of 2029. Watson would have passed the Turing Test.
"One criticism of Watson is correct. Which is it doesn't have a human level command of language. Its level of command is below humans. If it were at human levels we would be at the Turing Test now, but its able to take a level of understand that is non the less cogent, and apply it to a vast amount of material, ... so its taking a natural strength of computers and then applying it with a human like language ability. Not yet at human levels, but its a very powerful combination, so with a lower than human level of understanding it was able to get a higher score than the two best Jeopardy players in the world put together."
Ray Kurzweil closed his speech with a dramatic graph showing life expectancy rising globally. It was great to see such a positive closure, and with his new upcoming book How the Mind Works-- and How to Build One, I think next years Singularity Summit with just announced Stephen Pinker (How the Mind Works) will make for another amazing summit next year. The two of them don't agree on AI, yet both great cerebral optimists.

Here is one final excerpt from the speech, it includes a dazzling example of how imaging, and brain emulation have come. Both brain imaging and emulation are important tools in the growth of AI.
"I'm writing a book called How the Mind Works-- and How to Build One, as I mentioned, and we are making exponential gains at every level. We're turning that data into working models and simulations. ... I had debated John Horgan. He had written this article that you need trillions of lines of code. In his article was this picture saying this is just a small slice of the cerebral cortex, look at how complicated it is. We can never hope to understand this level of complexity. So I actually looked up this picture before the debate, and it turns out not to be a picture of the cerebral cortex at all. It's a picture of a computer simulation of the cerebral cortex. ... but clearly its a simulation. We understand the complexity of it."
We got a chance later in the day to ask Ray Kurzweil whether merging with technology would result in the eventual merger of all consciousness. Ray Kurzweil responded with the example of how we network computers today, to handle larger tasks, and then separate and function as individual units again. He predicts this will continue; man/machine intelligence merging on mass scales to tackle large problems and then dividing again into individuals. Ray, being the optimist he is, made sure to clarify - it won't be forced.
"Millions of PCs today can become one. ... and then those millions of computers can become individual computers again. So we can have it actually both ways. We can be individuals. We can also then merge into complex and greater intelligence. Certainly when we talk about the wisdom of crowds were doing a little bit of that. ... Things do that very readily today, and as our intelligence migrates to more digital form we will maintain that ability to do both; being one collective mind, and can also be individuals.
The collective will not be a forced collective."
Today we live in a world of always on, always connected. By the time we can merge our thoughts/consciousness we will be dealing in super intelligence, in a super smart world. I can't help but think; there is a clear difference here between being "forced" and being "convinced." We are growing more intimate with our technologies, we are merging with them. It's nothing like the world has ever witnessed.
Cheers to the future of our man/machine civilization!

-Stay tuned for further blog entries of the 2011 Singularity Summit-

-Jeremiah Bilas

-Personal entry-
To see Ray Kurzweil live was an emotional experience. It was Ray's work which introduced me to such profound, yet reasoned life-changing SL4 concepts. He is an idol, but to also realize he is just a man containing such a large idea, with such insane depth was moving. His methodical calculations and successful predictions made a Singularitarian out of me, and a supplement popper too! -- I currently take about 12 pills a day. -- Ray and Terry
I may be thinking a little too post-singularity to reason out a single super-consciousness, but to finally ask the man [Ray Kurzweil] in person, it was the largest single question I could think to ask him.

--technology is a cocoon -- from grub to butterfly in 30steps-- #singularity @JeremiahBilas

No comments: