Why we prefer EI to AI
As a co-founder of an AI company (Sentient Technologies) that used evolutionary learning to develop short term trading strategies, I am an expert on AI. OK, the first part of that sentence is true but not the last half. In truth, I was the fourth most important out of four founders, initially bringing subject matter expertise to the venture. But I did get to see under the hood of one flavor of AI and to see some of its weaknesses.
As one of our team described our tech, it was “a learning monster”. The problem was that the monster learned what it wanted. More precisely, it learned what we asked it to learn, but not necessarily what we thought we had asked it.
In one early case, the model developed what seemed to extremely profitable trading strategies by finding a few, very rare flaws in a huge data set. It was impressive that it found the flaws so quickly, but it was of no use in the real world.
So many AI projects are black boxes. We know what they do, but necessarily how. If the program is generating fun images or videos, that’s entirely acceptable. Should we hand over critical decisions to something we don’t completely understand? We hand over decisions and power to other people because, as fellow humans, we think we know enough about their thinking and motivation to judge them. I think we’d agree the results are mixed at best. And with machines, we don’t even know where they’ll go wrong.
This is why Elon Musk started Neuralink. He’d rather we merge with our tools than give them power over us.
At Outcome Trading, we agree. We’d rather see intelligence built into tools for human direction than see that intelligence developed to direct us. Of course, this is what humans having been doing forever. Tools have been growing more and more intelligent and useful.
And, that is what we have tried to do. We’ve built intelligence about trading and financial instruments into our tool. But, the human user is the ultimate controller. When you draw an outcome on our user interface, and tell the system “yes, I’d like to trade that”, you are using the embedded intelligence of the tool to achieve your outcome. A human is still in charge of the decision.