This project is a bit ahead of schedule. Here's a graph by Hans Moravec from the 1990s estimating how soon a human-life intelligence could be achieved by computers:
As I listen to Dharmendra Modha, I hear too much general fuzzy stuff. This guy is very, very far from achieving what he claims. Going back to the mid 1950s computer scientists have been promising that a "brain in a box" will be achieved in a decade or two. He needs to think less about ubiquitous implementation and "productivity" and more about the hard problem of coming up with a wiring which can process inputs and "think" about them and give results. This is so far from being achievable as to be laughable.
Back in the 1950s linguists thought they could produce an automated natural language translator. If you have used Google Translate, you know that 40 years later we are very, very far from achieving even this limited goal.
If you look at Douglas Lenat's Cyc project, for 16 years it has been trying to build a base of knowledge rules that would let a computer take in information and reason about it as a human would. That project is nowhere near being ready to deal with real world communication. For limited domains I'm sure it is helpful, but that is many orders of magnitude from what was promised when the project started.
Here is a Wired Magazine article that looks at Dharmendra Modha's research and calls it a "scam" (by another IBM researcher trying to build his own brain in a box!). I wouldn't call it a scam. I wrote research proposals. The pressure is always on to hype your project to get funding. Modha is guilty of "excessive enthusiasm". But without that he wouldn't have gotten any funding. It is a hoax in that the research goals will not be met within the project timeframe or budget. In fact, that timeframe will be off by at least an order of magnitude and the budget will be off by several orders of magnitude, easily 3 maybe 5 or even 8.
2 comments:
Ry;
This is all very interesting research. Whether it is possible to create a thinking computer it will still benefit mankind to try. I remembered that there had been some efforts at building a biological computer using plant cells and looked for that, but found that some progress has been made with leeches and I saw some other nano type work but nothing on plants. But, I think a radically different approach to computing will be necessary to create a "brain in a box".
Thank you for the video.. very interesting thoughts and ideas. And, you never know how quickly people will learn; sudden leaps are made from one little idea or accidental discovery or from a dedicated person spending years or a lifetime to finally get the answers. Who know?
Thomas: The "leech-ulator" article is interesting. I remember the first experiments to integrate electronic chips and living tissue back 20 years ago. The wonderful thing about science is that it builds higher and higher on each discovery. The future will be very interesting.
I'm keen on progress in this area, but my posting took a tone of "realism". I've watched the field for well over 30 years and there has always been too much hype. I would rather that scientists keep their noses to the grindstone and spend less time regaling the wider public with illusory claims. I want to be surprised with real progress rather than keep hearing empty promises.
I have no doubt that great things will be done. What I've learned is that progress will come in smaller steps and take much longer than what the experts think.
Post a Comment