During the Fireside Chat with John Carmack announcing his AGI startup* he describes the reason now is a high leverage time for “well heeled individuals” to get into AGI if they are willing to invest in ground-breaking research:
“The data’s not that large, the compute’s probably not that large in the larger scheme of things. A point I make: A year of life is a billion frames at 30 frames per second; and that fits on a thumb drive. You can tell that a one year old is a conscious intelligent being, so it does not require all the data on the Internet to demonstrate artificial general intelligence if your algorithm is correct.”
Sara Hooker’s indictment of AGI research in “The Hardware Lottery” has its answer in The Hutter Prize’s restrictions on hardware. Of all people, Carmack should be able to relate to the path dependency of the emerging AI industry on GPUs. Hutter’s is a rigorously-founded direction for AGI research and it is from one of those most qualified to set it. This is as opposed to engineering/technology/development – all of which are utility driven which is where we get into the “ought” of intelligent decision making as opposed to the “is” of natural science.
We need a lot more focus on what “is” the case. The current hysteria over what “ought” to be the case “is” unhinged. Wikipedia’s data should be more than adequate for a decent start. Moreover, if Wikipedia is truly comprehended, then the latent points of view going into its content will be exposed, along with their biases, as a necessary byproduct. That will be a very interesting aspect of what “is” the case.
As I’ve point out here, data efficiency is what all The Big Boys are ignoring at their peril. There is a huge breakthrough waiting to be made in data efficiency. Carmack gets it.
Why aren’t there others?
Here’s what Carmack says:
“I’m honestly surprised that there aren’t more people … I mean there are hundreds of people, like me, that were technical people that, you know, that succeeded, sold companies that have the resources to go and apply this, and I’m kind of surprised that there aren’t more of them taking small stabs at it… the expected value of that if you have no fear of ruin is quite significant.”
* The host of this fireside chat starts out basically telling his audience he doesn’t care about them and he proceeded to demonstrate this is the case by making people wait and hour and a half into the chat before he allowed Carmack to make the above substantive remarks. Really strange behavior – and I’m probably one of the people most interested in what the host was interested in since Carmack got his start in 3D gaming – not only that, but 3D cloning a game that Silas Warner wrote (Castle Wolfenstein) and Silas actually asked me for my inspect code for Spasim before coming out weeks later with his “stab” at 3D online gaming. But even I didn’t care enough to listen for more than a few minutes before stopping the live stream. Really inconsiderate of Carmack’s host – so much so that I kind of have to wonder…