The answer to the ultimate question about life, the universe and everything is 42. At least according to the Hitchhiker’s Guide to the Galaxy, in which it takes a supercomputer 7¼ million years to compute the answer.
What does this have to do with AI? AI could probably do this faster, much faster, in just 7 ¼ years perhaps. Not better, but faster.
We are witnessing a completely new way of problem solving: Computation at a speed so fast that the amount of data almost doesn’t matter anymore. This impresses me very much and that is why I am naming AI the hottest technology today.
It is incredible how AI is enabling the finding of answers to previously unanswerable questions because of the high speed of doing analysis: Not better but faster. NVIDIA has a demo running at their headquarters in Santa Clara where flowers on photographs are being identified. They can speed up the process by adding more and more processors until the human eye cannot even recognize the images anymore: Not better than a botanist, but millions of times faster.
I understand that this is the current state of the art. And I am fine with this because it allows the automation of solving many problems: Think cancer images instead of flowers. AI can automatically detect the majority of the images where there the patient has no cancer. And a real doctor only has to look at the no doubt low percentage of scans that are dubious. This would increase the quality of work for the expensively trained doctors by a lot! Think inspections of weldings or elevator cables or protective walls of nuclear power plants – the same thing applies. Great news for both productivity and for workers’ health.
The pattern is always the same: Train algorithms on data and then let it do the analysis fast and faster. It’s a tool. And I am fine with this.
It’s not the algorithms that pose the problem: It’s what problem they are being trained that may become a problem. It’s an ethical question at the end of the day. I for instance completely disagree with the application of AI by the Chinese government to identify faces captured by video cameras and to track persons or to publicly shame persons for undesirable behavior such as jaywalking – I am not fine with this. It is unethical because the people were never asked for permission to share their identities in public. However, it will allow the Chinese to gather more data for algorithm training on that anybody else. And as a consequence, China might develop the best AI in the world. But millions of Chinese will pay for this AI with their involuntary disclosure of their personal identity. I am not fine with this and I hope many buyers of AI will think the same way.
We must be careful. We must be critical. Autocratic regimes and free-ranging corporations have no scruples. AI can become a terrible tool in their hands. Society has yet to hold a discourse about AI, what it may be used for and what rules and regulations it should be subject to.
In the meantime, there are plenty of large problems out there for AI to solve, those that involve the automating of analysis at a million times the speed of today. Let’s solve these.