iPhone 7 Plus
ƒ/1.8
3.99 mm
1/20
40
My talk on AI from iterative algorithms on the USS Hornet just came out. Here’s the video of my talk.
6 responses to “Iterating to Superintelligence”
-
Speaker lineup
Backstage, just before I went on, I noticed that the cameraman had picked Genevieve out in the audience 
Such a cool venue, the USS Hornet was the aircraft carrier that recovered the Apollo 11 and Apollo 12 capsules and housed the astronauts in quarantine for fear of "moon germs". Here are more of my photos from an earlier visit
-
heh… thanks! And here is a more interesting conversation on superintelligence with Sam Harris
-
The next "better than Turing" test for AI will be when AI can find *and* prove interesting mathematical/ physical results. Maybe we’ll have to genetically engineer a few million scientists’ brains. I’d say it’s probably at least a couple decades away – having read Cedric Villani’s "Birth of a new theorem" refreshed my memory on how hard it is to do "real" stuff. Then again, I may be wrong, the advances of AI have been mind boggling even for me.
In the field of medicine, it might be much simpler, in some areas at least. For example take the very popular hernia operation, this is nowadays considered routine (using a patch). Turns out there is a better way, not using any patch, with better results. An AI machine with sufficient understanding of physiology could have come up with the same discovery (the Desarda method).
-
Yes… creativity and discovery may not be so uniquely human after all. About 10 years ago, a simple genetic program routinely discovered patentable inventions in analog circuit and antenna design.
From today’s WSJ:
"Drs. Pathak and Agrawal have designed a program to use curiosity in mastering videogames. It has two crucial features to do just that. First, instead of just getting rewards for a higher score, it’s also rewarded for being wrong. The program tries to predict what the screen will look like shortly after it makes a new move. If the prediction is right, the program won’t make that move again—it’s the same old same old. But if the prediction is wrong, the program will make the move again, trying to get more information. The machine is always driven to try new things and explore possibilities."
Leave a Reply