AI | The Oakland Futurist

Like many people I was shocked to learn that Ray Kurzweil had joined Google as some sort of employee. “Like, for a salary?” asked one incredulous acquaintance. Yep, I assume that directors of Engineering at Google receive a salary. That MIT tech review blog linked to above criticized his Singularity writings but lauded Kurzweil as an engineer. Some of my futurist friends have the opposite opinion of Kurzweil: his vision is interesting, but he hasn’t given us much in terms of engineering lately. Certainly Blio is nothing to write home about.

But I really want to talk about kool-aid. Has Ray drunk the Google kool-aid or has Google drunk the Singularity kool-aid or both? Maybe Ray was just as blown away by Norvig’s Deep Learning presentation at the 2012 Singularity Summit as I was. I previously commented that the way Google was able to pull in top academics like Ng and Hinton, team them with world-class parallelization engineer, Dean and eject a functioning face and cat classifier must take the wind out anyone competing with them in the AI space. Perhaps Ray said: if you can’t beat ’em, join ’em. He clearly wants to build a brain and he certainly doesn’t need a salary. So Ray seems to be drinking the Google kool-aid.

But what about the Google side? Are Brin and Page secret Singularitarians? Well maybe not too secret. Page did help setup the Singularity University in 2008. Foresight 2011 was held at the Google campus as well. So there is a lot of futurist synergy with Google. Some friends of mine think that Google’s senior executives just need to keep their interests quiet so they don’t spook investors too badly. Maybe that’s true. I just wonder how SIAI feels about this turn of events. I will have to ask someone over there. Do they think that Ray will inject caution and a concern for friendliness into Google’s pursuit of AI? Or will he help focus Google’s massive resources into an AGI project that will unwittingly create the search version of Skynet and perhaps convert all matter in the solar system into molecular adwords? It’s hard to tell.

UPDATE: I was further surprised to hear that some of my secret contacts at SIAI don’t actually think Ray will be working on AGI at all. AI maybe, but not AGI. Go figure.

I did break down and actually attend a couple of talks at the Singularity Summit this year: Vernor Vinge and Peter Norvig.

Peter Norvig gave a talk that would have satisfied any generic group of AI developers. Google is making some frightening progress. This Deep Learning project is the most interesting aspect of his presentation from an AI architecture point of view. It’s impressive that Google can pair two top-level researchers in the field (Andrew Ng and Geoffrey Hinton) with parallel processing expert Jeff Dean and scale up academic models onto a functional 1000 node cluster. Boom, you are identifying cats and faces from unlabeled YouTube videos. It must be sickening to anyone who wants to compete with Google in the AI space.

But he never really mentioned friendliness. I was hoping he would trot out some more theory behind this big data approach. He gave a similar talk to Monica Anderson’s AI meetup a couple of years ago. I was there for that and it was pretty cool to see him present to such a small crowd.

At the Singularity Summit this year, he also talked about Google’s translation service which basically derives translations by mapping many many identical documents written in multiple languages. I was hoping to ask him what happens when the algorithm starts consuming translations that were actually created by Google Translate. It’s bound to screw them up if that happens. But then I realized that Google probably saves every translated document and checks new documents checksums against previous translations before using them to build mappings. That’s hard to picture though. They manage: A. Mind. Crushingly. Large. Amount. Of. Data.

Vernor Vinge outlined some outcomes that he sees for the singularity. One crazy idea he puts forth is a digital gaia where the world is minutely ornamented with digital sensors coupled to processors and actuators. One day they all spontaneously “wake up” and all hell breaks loose. He describes a reality with all the stability and permanence of the financial markets. I had a vision of my SmartLivingRoom(tm) suddenly reconfiguring itself into a nightmare of yawning jaws and oozing orifices. But in reality, we might just see wild fluctuations in the functionality of computationally enhanced environments; from smart to dumb to incomprehensible.

Next up: Augmented intelligence, a neo-neo-cortex provided by technology. This is his preferred scenario. Crowdsourcing is cool, yada-yada. Vinge imagines a UI so extreme that accessing it would be as convenient as the supported cognitive features. I used to like this idea until I started thinking about the security implications. I don’t want my brain hacked.

He did make one amazingly succinct point about human computer synergy. Computers can give us instantaneous recall and amazing processing speed, humans can provide that which we are best at: wanting things.

Humans want things. For me this cuts to the very heart of the AI question. I always complain that none of these AI geniuses can show us an algorithm to define problems. (No, CEV doesn’t count.) Algorithmic problem definition is just another way to say algorithmic desire definition Good luck with that one.

All simple human desires seem to arise from biological imperatives. Maybe artificial life could give you that. More complex desires are interpersonal and might be impossible to reduce back to metabolic processes. You may want fame for the status, but the specific type of fame depends on which person or group you are trying to impress. And that changes throughout your life.

And if we do build Artificial Life, it may well be that it can only function with similar constraints as, uh, non-artificial life. In fact, Terrence Deacon may well be right and constraints are the key to everything. Ahh, the warm fuzzies of embodiment are seeping over me now.

But seriously, SingInst, where is this algorithmic desire going to come from? And once you get that, how the hell are you going to constrain the actions of GodLikeAI again? I know, I know, Gandi would never change himself into an anti-Gandi. But we may be like bacteria saying that our distant offspring would never neglect the all encompassing wisdom of nutrient gradients.