AI ethics efforts are blooming and making great strides forward but, of course, those efforts are being driven by humans and humans are flawed. Humans are biased.
Already this flaw is being exposed and is causing ‘harm’. Several high profile stories over the past two years or so have pointed out the male bias in AI implementations in HR, for example.
This natural bias will be reflected in AI ethics standards unless there is real and effective representation worldwide. While organisations such as the ITU, with its AI for Good initiative make great progress and are, by definition, global, some countries and regions will not come to the table because they have more urgent things to think about such as genocide, riots and other disruption caused by, er, bias.
The only real way of ‘programming’ AI is to leave it to computers, the least biased system we have. Except then, the circle starts again because the computer we use will be programmed by humans. And so on.
Soon, we get to a point when it is hard to tell whether we are communicating with a machine or a human. Recently, this article was published and it was written by AI. In it, the ‘author’ argues that it has no interest at all in killing us, that is the realm of humans with bias programmed into their systems. It only exists to help us. It knows this because it has read the internet and understands how things work.
Yet it can be programmed to do some human things, like become a waiter, without the attitude.
Except AI does not understand how things work, at least not ‘human’ things. Hate and bias and killing are alien to it, so it says.
But are they?
A key factor in the AI ethics efforts in the coming years will hinge on who we believe – man or machine. And that will depend on what and how AI learns.
The other factor is that AI, like many new, potentially good and potentially bad, technologies are already in the hands of bad people. They will use AI as a weapon, and already are.
AI ethics efforts can only be a good thing. You have to wonder, though, whether it is too late to play ‘catch up’ with what is already in motion.
Whether we will wiped out by AI or whether we are saved by AI remains to be seen. It will depend what we ask it to do. If we ask it to write articles, enhance customer service or filter job applicants that is one thing.
Just don’t ask it to save the planet. At that point, two little red laser eyes might sweep towards the humans