The breathlessness/excitement over AI is inversely proportional to people’s familiarity with computers and knowledge of productive processes.
That said, automation will continue, and if a job can be done better by a machine, it’s likely to happen.
@raneck:
I suggest you listen to the link I put up, and especially the last guy. What they are talking about with neural networks and deep learning goes beyond that, if what they are saying is true (and I suspect it is, I work with people who are heavily into machine learning and the like), and this goes beyond merely analyzing data, that a system like this can go beyond classifying and recognizing data and into the reality of being able to do things that at least parallel human attributes like intuition and other things that are supposedly things only a human being can do. And given the mentality that any labor cost is a bad labor cost, I wouldn’t bet against what these guys are saying coming true.
@raneck – exactly. NNs seem like a cargo cult way of simulating intelligence – we mimic the physical brain in hopes of getting a brain-like response. I don’t think that it will ever work properly except in very limited scenarios. In my field, which deals with real-time physical data, researchers went through an NN phase a decade or so ago. They’ve since backed away from it because they were never able to know exactly what the training was “latching onto” in the training data. Now it’s back to fuzzy logic and membership functions.
I attended a national engineering conference last month. One of the sessions was on 3D printing of structures. Uh, let’s just say they have a good ways to go before they have anything practical. The photos of concrete “structures” being printed were pretty funny. Looked like a five-year-old was squirting concrete everywhere. NASA is pushing the researchers to develop a material other than concrete that could be used for printing habitats on Mars, so we’ll see what they come up with.
One of the main speakers was a Polish guy I went to grad school with. The last time we saw each other was in 1984. I wouldn’t have recognized him - he’s old now, ha. Of course, he didn’t recognize me, either, when I approached him after the presentation…
AI scares the crap out of me. Humans don’t have a good track record with creating awesome technology that doesn’t become weaponized or otherwise destructive.
I’d very much like to keep machines dumber than us.
Right now, neural networks are doing machine translation (of Chinese to English, eg), face recognition, and interpreting X rays. Those “limited scenarios” the neural networks are good at include replacing radiologists.
@CardinalFang
Most machine translation, especially from Mandarin to English, is horrendously bad right now.
Robots taking over priests’ duties?! Yikes.
@mainelonghorn:
There was an article in the Times by Thomas Friedman talking about the whole small town america dying jobs disappearing issue and talked about the hurdles that were making the problem worse (most definitely, finding workers not messed up on drugs, it is that bad). Anyway, he highlighted a couple of companies using 3D printers and they are doing things like making car bodies or building parts on demand, it was interesting the kind of things they are doing with the technology, and it is still a pretty new technology, CNC once was a prototype, look what they do with that these days:).
As far as AI goes, it is a lot closer than people think, in the financial industry for example companies are putting more and more reliance on it for things like fund managing or trading strategies, whether as an assistant to human traders or on their own. We are far away from medical scanners that can likely replace the current costly and fragmented issue of human diagnosis, a computerized system can replace the lab that does the blood work and then based on the results, give a probably diagnosis. X prize just awarded a prize for a simplified version of the medical tricorder they used on Star Trek, it can diagnose something like 13 different human medical conditions in under a day, and that was a prototype the team built literally on one of the principal’s kitchen table. In reality, the obstacle is not necessarily the technology being ready, it is other barriers that are keeping it from being used. Not going to happen overnight, it has a ways to go, but a lot of this is likely inevitable unless we create some sort of society locked in time and forbid new technology and so forth, kind of like promoting a 19th century fuel in a world with much better options.