Marissa Mayer of Google gave today’s Keynote conversation. It’s no wonder Google does such wonders when people like this are in charge. I did get a chance to ask about the intersection of search and AI and got a fantastic answer – she thinks they will intersect, and this could happen within about ten years. Also interesting was that she said they are now seeing things that “look like intelligence” emerging from the search algorithms. This is not thought, but she indictated that it’s possible to have thought like processes emerge in this fashion rather than with the massive computational approaches that were popular several years ago. This is consistent with Kurzweil’s notion that it’ll be massive parallel processing and not massive supercomputing that will probably bring the mechanical mind “to life” within the next decade or so. I’m glad Marissa Mayer seems to agree and I hope this is a focus for Google in the future (I got the idea it’s not a focus now).
I had a chance to ask Matt Cutts of Google engineering fame the same question yesterday and he was not as optimistic, thinking that it could take another 50 years to get conscious computing. But Matt correctly noted that Marissa would be more optimistic than he was because his Master’s program at University of North Carolina was lacking in much AI content due to the AI skepticism of the architect of that program.
Larry Page’s recent remarks suggesting that a viable human thinking algorithm may appear fairly soon are more in line with Marissa’s optimistic view that within a decade we could see mature, conscious, artificial intellects. The staggering implications of conscious computing are lost on many people in computing for reasons I simply don’t understand, but I think are related to the current focus on computing science as an engineering and calculation paradigm rather than a biological one. As the brain is reverse engineered and we begin to enhance neurons with forms of programming it seems reasonable to assume things are going to get … very interesting very fast.
Joe Duck 
Would you share the link to that Marissa Mayer’s conversation?
Hi Dennis –
Note that her talk was about search issues, the iPhone which she adores, and Google’s future:
http://www.toprankblog.com/2007/08/marissa-mayer/
After the talk I got to chat with her briefly about how search and AI will intersect in the coming years and was thrilled that she basically seems to agree with Kurzweil’s estimates of “about a decade” until conscious computing. I also spoke briefly with Jim Lanzone, ASK CEO, about this as well as Eytan Seidman of MS Search and Matt Cutts of Google search. Matt was the most pessimistic though even he felt that we’ll have conscious computing within about 45 years. He also noted this is not his specialty. I’m still not sure how much optimism about AI is clouding people’s estimates as it did back in the 70s, but I’m thinking “not much”. This is a new age for AI due to computational efficiencies and massive parallel processing and I’m looking forward to HAL 9000 wannabes though I do recommend they give them a less sinister voice…. “Dave, Dave?”….
Hey Joe, just a quick comment: I went to undergrad at Kentucky, but it was my grad school at Univ. of North Carolina at Chapel Hill that was more skeptical about artificial intelligence.
It was great talking to you in person! 🙂
Pingback: SES San Jose 2007 write-up
Hey, thx for clarification Matt. I corrected above post.
AI is dead, long live AL! 😉
I think the comment about the future being about massive parallel processing as opposed to massive supercomputing is very apropos although the mention of AI is almost a couple of decades behind, except in popular literature.
AL (Artificial Life) can more easily mimic actual thought processes because like the brain’s thought process broken down to its smallest component, the firing of neurons, AL Agents are very simple and coded with a very small set of rules to follow.
Also, just like the human brain is easily the most advanced as well as most massively parallel processor in the world AL Agents can each be given a small portion of a complex task, each perform their task individually and then as they each complete their respective tasks, contribute their output to the combined output and finally coming up with an answer.
Try to write a computer program to architect a termite mound using AI based question sequences and you’ll get nowhere. Try the same thing by programming individual Agents to perform the same simple processes that their natural counterparts perform and what before was impossible becomes almost trivial.
With AL though, as opposed to AI where the “intelligence” comes from the programmer, the best AL intelligence comes from having large data sets to train on and who arguably has some of the largest datasets in the world to work with?
Google and AL seem like they were literally made for each other.
Great comment Craig. I’d agree that the first big AI breakthroughs may come in AL realm – perhaps a reverse engineering of the human brain into AL style components formed from some type of organic material like neurons. There’s some really neat Israeli research now that is connecting processors to neurons. It’s interesting to think how we, as organic thinking objects, currently do a lot of interaction with silicon “thinking” objects like computers.
It’s obviously a very big deal but an eventual inevitability that this indirect interaction will eventually be direct – ie our assembly of human neurons will communicate more *directly* with the silicon assemblies rather than simply through our visual cortex. This will lead to a human intellectual enhancement that is hard to even imagine. Translation, math, and …. wow is my chess going to improve!
Pingback: Web 2.0 Conference and Expo « Joe Duck
Pingback: Are you biased? « Joe Duck