if the author is to be believed, they were pretty close to having something that would have been a breakthrough for AI. i wonder what happened to the source code, and why it wasn't pursued further.
Problems that seem easy in theory are so much harder in practise. Like the other day I thought I'd write a quick piece of software in java to automatically navigate from one spot to another and avoid obstructions. Sounded like a 15 minute job, lasted 3 days. It's the same problem all AI creators have, but with them, it sounds like 15 minutes, lasts 3 years and still doesn't work.
It's a great read, though I'm not sure how to take his words in the engineering domain (e.g. what does "3/4 complete" exactly mean?). I don't doubt he is a vary smart guy and also had amazing people; but many years of software engineering have made me quite conservative in terms of estimate.
If all the planned components are written, and you can feed the real data into the system and get the reasonable final result out of it, then I'd call it, say, 50% complete; there'd be tedious works after that to make things robust, or find users have unusual settings and you have to adapt to it, etc. Borrowing his rocket metaphor, the rocket without the nosecone is less than half way in terms of development...
Of course it really depends on the nature of the product; maybe things like AI engines you can convince the client to pay in the state of "it basically works" and then gradually adapt it to the client's needs afterwards. Also his writing is sometimes sarcastic (which makes the writing enjoyable), so I'm not sure he uses "3/4" or "almost finsihed" in ironic way (that he knows it wasn't "almost finished", but he thought so back then).
Anyway, it's so easy to criticize retrospectively. I still respect them for having tackled to such a big problem.
I have a friend with an AI investing system http://www.neuralinvesting.com/. He is a smart guy, too smart. I have found the smarter the person the harder it is to make money because they want to solve problems all of the time. If you are solving problems when do you stop and say "this is enough!" and just start pushing what you have made?
7 comments
[ 7.1 ms ] story [ 303 ms ] threadBetter to have tried and failed, than never to have tried at all.
if the author is to be believed, they were pretty close to having something that would have been a breakthrough for AI. i wonder what happened to the source code, and why it wasn't pursued further.
In reality, AI is like jumping across the Grand Canyon. You either jump all of the way, or Wile E. Coyote it.
If all the planned components are written, and you can feed the real data into the system and get the reasonable final result out of it, then I'd call it, say, 50% complete; there'd be tedious works after that to make things robust, or find users have unusual settings and you have to adapt to it, etc. Borrowing his rocket metaphor, the rocket without the nosecone is less than half way in terms of development...
Of course it really depends on the nature of the product; maybe things like AI engines you can convince the client to pay in the state of "it basically works" and then gradually adapt it to the client's needs afterwards. Also his writing is sometimes sarcastic (which makes the writing enjoyable), so I'm not sure he uses "3/4" or "almost finsihed" in ironic way (that he knows it wasn't "almost finished", but he thought so back then).
Anyway, it's so easy to criticize retrospectively. I still respect them for having tackled to such a big problem.