Artificial Intelligence and Horse Racing
In 1996, I came across the following question from someone on the Internet, who wanted to improve an AI algortihm for horse racing. I suggested him to contact Charles Babbage, and got this reply:
>Joerg Heitkoetter wrote:
> |> I am looking for some one who might have experience in this field
> |> dealing with horse racing and AI. I have the horse racing program,
> |> ROI's at 20-35%...just want to make it better...
> |>
> |> Any input?
> Ask Charles Babbage.
who is charles babbage? where do i find him?
Hi there, the short answer to your question is:
Charles Babbage (1792-1871) the "father of the computer" started building something in 1821 he called "differential engine"--then abandoned his plans when in 1834 he came across a steam powered loom, hat was programmable with punch cards, invented by Jean-Marie Jaquard (1752-1834); he went on to build an "analytical engine" but never finished it.
However, the concepts of the analytcal engine is mostly the same the concept used in todays PC. (In 1991 it was rebuilt at the London Science museum; it really works--you can visit it.)
At Babbages time, he had some funds raised for his differential engine (actually 17,470 pound sterling; at that time a locomotive cost only about 1,000 pounds sterling) and the English research society didn't wanted to grant more money for a second project, since Babbage still didn't finish the first engine;
So Babbage needed money, and that's where the horses come into play: Only one person trusted Babbage at that time: Lady Ada Lovelace, daughter of the famous Lord Byron, she was she first programmer (that's why the programming language ADA is named after her), and had a fortune that she handed over to Babbage. Babbage needed to enlarge this fortune and promptly invented a failsafe algorithm for horse races. Needless to say that he lost Ada's fortune, who died early afterward.
So, what does this have to do with ai? Easy--computers can only efficiently compute things that you can already compute/handle in the real world; if you cannot give a real world algorithm (as for horse racing) and implement a model of the algorithm on a computer; you cannot expect ai to work for you. And once you have a real world algorithm to beat the odds of horse races, you don't want it on computers, since it would be easy to copy to everyone, which would render the whole system of horse races useless.
So yes, it *is* impossible. And that’s a good thing.
