AI experts v. Bookies AI programs
October 21st, 2017 at 2:26:16 PM
permalink
Over at Diversity Tomorrow I just posted a /. reference to Technology Review article wherein AI experts used historical data to make 3.5 percent profit and used random bets that resulted in a 3.3 percent loss.
Once they started betting for real, the bookies promptly restricted their accounts as having some "edge".
Should have used some of that 3.3 percent as "cover".
Once they started betting for real, the bookies promptly restricted their accounts as having some "edge".
Should have used some of that 3.3 percent as "cover".
October 22nd, 2017 at 3:05:59 PM
permalink
It seems the bookies had hired better AI experts and hired them sooner.
Deep Learning Neural Networks is the only field where you make a program intentionally work at far less than optimal settings (to avoid over fitting) and then wind up with an optimal program anyway.
Deep Learning Neural Networks is the only field where you make a program intentionally work at far less than optimal settings (to avoid over fitting) and then wind up with an optimal program anyway.
October 22nd, 2017 at 3:15:39 PM
permalink
My favorite CS joke is that Artifical Intelligence is just what computer scientists like to call statistics.
