HN Books @HNBooksMonth

Sad news. RIP, Ms. McCorduck.

The result was “Machines Who Think: A Personal Inquiry Into the History and Prospects of Artificial Intelligence” (1979), a chronicle of past attempts to mechanize thought.

I was first introduced to her name when I read a book she co-wrote along with Ed Feigenbaum - The Fifth Generation: Artificial Intelligence and Japan's Computer Challenge to the World[1]. It took me a while to later realize that she also wrote another book that was sitting on my shelf waiting to be read. That being Machines Who Think. I feel a bit sad now in acknowledging that the book is still sitting there waiting to be read. :-(

[1]: https://www.amazon.com/Fifth-Generation-Artificial-Intellige...

⬐ jhbadger

It was a fairly enthusiastic book about the Japanese Fifth Generation Project and the potential of logic programming and Prolog, combined with the sort of "oh, no, America has missed the boat on this technology and Japan will rule the future" lamenting that was so common in the 1980s. In retrospect of course, the 5th generation resulted in very little.

"Japan getting an edge with future technology!" When have I heard that before?

Oh, yeah, it was 1983:

https://www.amazon.com/Fifth-Generation-Artificial-Intellige...?

"Create and operate a National AI infrastructure" "National AI Research Centers" "National AI Laboratories" It's all about getting Government money.

Ed Feigenbaum was writing stuff like that in the 1980s. He was calling for a national AI lab, headed by him, to fight off the threat from Japan. See his book, "The Fifth Generation".[1] This was all for "expert systems", which didn't do much and led to the "AI Winter".

This time around, we don't need that, because machine learning already has profitable applications. Finding applications outside adtech, Big Brother, and finance might be useful.

Remember the "BRAIN initiative" from 2013 or so?[2] Whatever happened with that?

[1] https://www.amazon.com/Fifth-Generation-Artificial-Intellige...

[2] https://en.wikipedia.org/wiki/BRAIN_Initiative

⬐ fakename11

Looks like it's funding different projects: https://www.braininitiative.nih.gov/funding/funded-awards

⬐ mattkrause

The BRAIN Initiative is still happening.

We published some stuff funded by it (via DARPA's RAM/RAM Replay programs) earlier this month. I'm pretty sure there are other active grants and funding programs (RFAs, etc) out there too.

The program wasn't meant to completely solve the brain--that's going to take far longer than ten years. Instead, the goal, especially of DARPA's portion, was to give neuroscience a jumpstart and then let the promising results get picked up by more traditional mechanisms (NIH Institutes, industry, etc). That's worked pretty well: companies have come out of it (Nia Therapeutics was spun off from UPENN's work on RAM), along with a slew of new data, tools, and papers. On a very personal level, I'm working on something totally different than I planned on, and it's going well.

⬐ thatoneuser

This is very interesting to me. What are you working on?

⬐ mattkrause

Non-invasive brain stimulation, and specifically a family of techniques called transcranial electrical stimulation (tES). These use electrodes, placed right on the intact scalp, to generate weak electric fields around the head. The hope is that (some of) it passes through the skin, skull, and CSF and affects the electrical activity of the neurons beneath.

While people have gone crazy trying this for every sort of problem (basic science experiments, clinical conditions, performance enhancements), it's been unclear how tES works—or, indeed, if it works at all.

The human data is a mess: some of it reports dramatic effects, others find nothing at all, and many of the experiments are...not the best to begin with. Animal data has mostly come from isolated tissue or rodents, which have very different brains and heads (the mouse skull is paper-thin, for example, and the brain is smooth, which makes it easier to get an electric field in there).

To address this, we apply tES to awake monkeys playing "video games" (boring ones, but still). Monkey anatomy is pretty similar to humans, as is their behavior, but we can also monitor neural activity via tiny implanted electrodes.

To my immense surprise, tES appears to have effects on both brain activity and behavior. In our 2017 study, we used direct current and looked at effects in areas on the brain's surface. Someone wrote a nice blog post about it here: https://hackaday.com/2017/11/13/shockingly-darpas-brain-stim... (which also has a copy of the paper). More recently, we managed to find some effects on individual neurons and in deep brain areas that are often targeted for neurostimulation (twitter thread here: https://twitter.com/prokraustinator/status/11026751895638220...).

Never would have gone down this road without DARPA, but it's been fun (and exhausting). I do like that it's got some fairly obvious clinical applications, unlike my PhD stuff.

Ouch. That was a decade late to be doing that.

I finished a MSCS in 1985. Ed Feigenbaum was still influential then, but it was getting embarrassing. He'd been claiming that expert systems would yield strong AI Real Soon Now. He wrote a book, "The Fifth Generation", [1] which is a call to battle to win in AI. Against Japan, which at the time had a heavily funded effort to develop "fifth generation computers" that would run Prolog. (Anybody remember Prolog? Turbo Prolog?) He'd testified before Congress that the "US would become an agrarian nation" if Congress didn't fund a big AI lab headed by him.

I'd already been doing programming proof of correctness work (I went to Stanford grad school from industry, not right out of college), and so I was already using theorem provers and aware of what you could and couldn't do with inference engines. Some of the Stanford courses were just bad philosophy. (One exam question: "Does a rock have intentions?")

"Expert systems" turned out to just be another way of programming, and not a widely useful one. Today, we'd call it a domain-specific programming language. It's useful for some problems like troubleshooting and how-to guides, but you're mostly just encoding a flowchart. You get out what some human put in, no more.

One idea at the time was that if enough effort went into describing the real world in rules, AI would somehow emerge. The Cyc project[3] started to do this in 1984, struggling to encode common sense in predicate calculus. They're still at it, at some low level of effort.[4] They tried to make it relevant to the "semantic web", but that didn't seem to result in much.

Stanford at one time offered a 5-year "Knowledge Engineering" degree. This was to train people for the coming boom in expert systems, which would need people with both CS and psychology training. They would watch and learn how experts did things, as psychology researchers do, then manually codify that information into rules.[2] I wonder what happened to those people.

[1] http://www.amazon.com/The-Fifth-Generation-Artificial-Intell... [2] https://saltworks.stanford.edu/assets/gx753nb0607.pdf [3] https://en.wikipedia.org/wiki/Cyc [4] http://www.businessinsider.com/cycorp-ai-2014-7