First step in building a successful ANYTHING: make sure your product works! They are claiming to possess algorithms that are able to predict the success of a given startup. I'm curious to know how many startups they have successfully invested in that they have developed algorithms against.
I would imagine PG would be in a better position on MANY levels to write something like this, if possible. Indeed if such a thing were possible, I would imagine there would be some serious PhD research in this area.
I find it funny that this got as much initial positive press as it did whereas programs that can predict the stock market reliably (a much simpler problem given the relatively controlled/closed system of any given market compared to the virtual chaotic environment of a startup in the wild) are often passed off as snake oil in many circles. Taking this problem alone (automated trading programs) there are multi-billion dollar investment firms that spend millions on teams of developers and PhD working out formulas and algorithms to crack this problem; I know from personal experience that these systems are only good for a few months and are constantly changing. This fact alone diminishes the predictive value of whatever algorithms they have developed over time...
If two 20 somethings have something that groundbreaking that does NOT require constant tweaking (unlikely) then they must have discovered some new mathematical property governing human decision making at a large scale (even more unlikely).
"..there would be some serious PhD research in this area."
PhD research? I think you mean just 'research'. PhDs aren't the only ones doing research in this world. Neither are they the ones doing the 'hardest' or 'highest' research. Academia is just a different community from that of founder/VCs, or writers, or movie makers, or chess players. It has its own language, its own jargon, and its own objective functions for assigning value to work. But it's not more 'research-y'.*
'PhD research' means 'research work for which the audience is in academia'. The way it's normally used, one can just drop the first word without loss of meaning. Just like in the phrase 'personally believe' that William Zinsser picks on. (http://www.amazon.com/Writing-Well-25th-Anniversary-Nonficti...)
</rant>
* - Nor is it more open anymore, with all the emphasis on open software in industry, and the increasing commercialization of academic projects.
I would imagine PG would be in a better position on MANY levels to write something like this, if possible. Indeed if such a thing were possible, I would imagine there would be some serious PhD research in this area.
I find it funny that this got as much initial positive press as it did whereas programs that can predict the stock market reliably (a much simpler problem given the relatively controlled/closed system of any given market compared to the virtual chaotic environment of a startup in the wild) are often passed off as snake oil in many circles. Taking this problem alone (automated trading programs) there are multi-billion dollar investment firms that spend millions on teams of developers and PhD working out formulas and algorithms to crack this problem; I know from personal experience that these systems are only good for a few months and are constantly changing. This fact alone diminishes the predictive value of whatever algorithms they have developed over time...
If two 20 somethings have something that groundbreaking that does NOT require constant tweaking (unlikely) then they must have discovered some new mathematical property governing human decision making at a large scale (even more unlikely).