Baboon – real-time merge conflicts

I haven’t used this yet, but will try it out soon.

“The lightweight baboon client detects in realtime any file changes on your projects.”

It’s a daemon that runs on each system and requires voluntary subscription by each person. Seems straightforward, and also seems like it would be easy to bundle it together with git.

The github page has more details:


National Geographic on hard disk 1888-2009

Get the complete National Geographic (well, through to 2009) on a hard disk. This would make a great Christmas present.

Shopping link:

It says 160 GB, but the parallel “complete national geographic” on DVD ships on 7 DVDs, which is less than 60 GB even with dual-layer DVDs. So I wonder if the DVD version really isn’t complete, or if the hard disk version has extras.


Noam Chomsky on “Where AI went wrong”

Great article in The Atlantic.

Noam Chomsky on Where Artificial Intelligence Went Wrong

I have to admit that I’m on the side of “analytic theory” in the “big data versus analytic theory” debate going on in many fields of science right now. I subscribe to the belief that “you don’t understand something unless you can make it”. Also, there’s a fallacy embedded in the big data approach that’s very similar to the one that causes big problems in the financial industry – measuring the past doesn’t predict the future.

Yes, there’s lots for us to figure out in AI and biochemistry and psychology, and it very well may take us thousands of years to make meaningful progress. We’re impatient, so we are drawn to shortcuts. But the shortcuts are dead ends.

This is also why I don’t see behaviorism as being useful. I can’t use it to make new things, and I can only do relatively poor predictions with it. In physics, a prediction is pretty accurate, as in “so perfect it might as well be unbelievable”. I can aim a pebble from Earth and hit a spot on Mars within a hundred meters, without too much trouble. A smart high school kid or any college kid can do this (except for the actual mechanical part of building the apparatus). But accuracy in behaviorism is limited to 51%, 55%, maybe even a staggering 80%.

Look, what Google has done has been amazing, and useful. But I don’t see how you can build on it to go to another level.