Deep machine learning and intelligence

Stockholm, Gärdet 2014-12-29

Andreas sent me this link, where Jeremy Howard talks about machine learning and an emerging IT revolution were computers replace humans for a bit more advanced intellectual work than today’s ERP, CAD and etc. systems are capable of. Very interesting TED speech, if I understand mr. Howard right he implies the technological singularity is imminent with the help of an algorithm called deep learning, earlier this year I found a prediction on Wikipedia this will happen around 2040, I doubt. According to mr. Howard the necessary building blocks for super human intelligence are already here, we just have to put the pieces together better, and this will go fast. What I am absolutely certain about, the pace of computer ability is accelerating exponentially, every new prediction tend to be closer in time, but for superhuman intelligence. 
While predictions for computer ability in general is similar to the predictions of the mappings of the (human) genomes, it was thought to take an excruciatingly long time until Craig Venter came along. I think the predictions for the singularity more will resemble of predictions of the commercial breakthrough of the fuel cell, it is always twenty years in future.

This doesn't mean computers are not getting smarter. Computers or rather computer programmes are transcending from aiding decision making to make decisions, this has been ongoing for some time now and this pace is accelerating, we are in the dawning of a new computer revolution where intellectual workers will be replaced by computers. And this is pretty important, in five to ten years we will have a new world, with new opportunities, not all good.
Some weeks ago I participated in a seminar at KTH discussing digitalisation of business opportunities. The participants were top IT managers from some of Sweden's biggest corporations plus me. One participant told us in near future no one will let humans take care of their financial investments ‘Do you let a human do your investments? Are you stupid!’ We were all in agreement that this new IT revolution has already begun and there are great (business-) opportunities for us all.

But for computers being smarter than us I think we have a long way to go. Mr Howard is most probably right, machine learning is important for computer development. He says ‘Machine learning revolution never settles down’, how does he know? There are capping limits to all things under the sun (and beyond), ultimately physical limits like the thermodynamical laws. Quantum mechanics most likely has limitations too. In my somewhat pessimistic post I state man has the ultimate intellect, we can figure out anything logical just given time. But it might so happen speed is what smartness is all about, if so then computer programmes will prevail over man. Shortly after man will most likely be annihilated, fortunately for me this will only happen twenty years from now.

No comments:

Post a Comment