I've been off my science kick for some time now, but some news is just so exciting that it needs to be shouted from a mountaintop.
I feel like we get so much news nowadays that we really take things for granted, even when they're incredibly significant. When Alpha Go beat Lee Sedol, a World Champion of Go, people basically said: "Hunh. Neat." and then forgot about it. Those of us who are more technologically inclined noted the significance and realized that this was many orders of magnitude more difficult than making a chess bot that could beat a world champion, but we still said something along the lines of: "Hunh. Neat."
I'm no exception.
But now we're seeing something that is fully different, not even analogous to something we've seen before. A completely self-taught version of Alpha Go beat the previous one, which had been 'taught' with many human examples first. This means that in just one year the computer that can beat all humans at Go was itself trounced by another, more superior computer. And while Lee Sedol was able to eek in a few victories, this more superior computer beat the one that beat Lee Sedol 100 times without losing once.
I don't want to undermine the significance of this.
The thing to focus on here is that this computer taught itself from scratch, starting from no human examples and proceeding utter mastery.
The next thing to pay attention to is the shortness of time between the one accomplishment and the other. These machines are getting better and better in huge leaps from year to year, and it's just so incredibly exciting that I can barely even stand it.
With machines like IBM's Watson beginning to work on medicine, the future is looking incredible. Just imagine how much medical progress could be made in just a few years of a computer being put to the task. The main issue is that this would require simulations of humans to become meaningfully realistic, but you can bet your bottom dollar that big pharma is already hacking away at this on a small scale in it's secret labs.
Source: 1
