Hacker News Reader

Ragging on things that failed with the benefit of hindsight is a good way to sound smart. I have never liked that genre much.

5gen failed but Apollo, the Manhattan Project, DARPANet, the Bell Labs transistor project, and SpaceX reusable boosters (for a contemporary example) succeeded. This piece argues that the ambitious megaproject successes were more decomposable but I find this rationale to be an unconvincing post-hoc just-so story.

The conclusion I draw when I look over the history of innovation is that it's like a rare animal that almost never mates in captivity or one of those exotic fungi we can barely culture. We know how to set up conditions where innovation can happen: get smart people, provide capital, give them an ambitious but not impossible goal. Beyond this it's serendipity. Sometimes the muse pays a visit and sometimes it doesn't. The dynamics seem almost identical to what you see in art and music, and I don't think that's coincidence. Science and engineering are perhaps more like the arts than we tend to think.

5gen probably would never have achieved all its goals, but it might have achieved something really notable had the muse chosen to pay a visit.

This is my favorite comment in this whole thread. Looking back on big failures in history, it's too easy to point out the obvious mistakes or logical reasons why it failed, how unrealistic their dreams were.

That narrative doesn't take into account that notable successes in history, major innovations, were born from basically the same ingredients in a shared cultural context. The difference was (at least seemingly) often just a slight chance variation in conditions, being in the right place at the right time, working on an idea that happened to be ready to blossom.

I love the parallels you draw with exotic fungi, how sensitive these ambitious ideas are to its environment. There are so many ways it can fail, and the reason for success is often elusive and serendipitous.

> Science and engineering are perhaps more like the arts than we tend to think

This seems related to how great art and music are often born of a "cultural milieu", like the jazz scene, or Paris cafe culture in the 1920's that fostered poets, painters, musicians. I imagine Xerox PARC was like that, a vibrant intellectual subculture of creative people who worked in the medium of technology.

Speaking of the muse, I like the popular expression, "The elves have left Middle Earth." There's a kind of indescribable "magic" that can inhabit a place and time, pulling everyone around under its spell. Then, as mysteriously as it appeared, it's gone - leaving behind relics of wonders, art works or artifacts, that reason alone cannot explain.

> So I really don't understand where all the schadenfreude comes from, in the article. A big commercial project failed and took with it a whole branch of computer science. Who, exactly, benefited from this?

It wasn't a big commercial system; it was a big government system (you probably can't imagine the degree of shock and awe that greeted MITI people or even the mention of the minstry in the US). This was when Japan was going to destroy America and goodamnit, Something Must Be Done.

Of course as happens with such projects it all comes tumbling down, but usually with a whimper. Apollo is another good example: they hit their goal and then what? 5Gen was the more common case: just wasted away.

The schadenfreude comes not just from the central planning hubris but from the American freak-out response to what was really hype and a squib. The author says the same thing is happening today, with the hype-o-meter on 11, again in the AI field, and again with an "asian menace" on the horizon, this time China.

5Gen just rode a hype machine though; it didn't take the field out -- we could all see that coming anyway.

I think people today have a very hard time understanding what things were like during the Japan bubble from about 1987 to 1990. Coming out of the 1970s Japan ended up dominating certain parts of the electronics - and especially personal electronics, like radios or portable cassette players. They had terribly failed in _information technology_ however, but most people didn't know that, what they saw was walkmen, MITI coordinating wiping out American DRAM and television producers using dumping (among other things) and so on.

Another important event was that the US dollar was very high in the mid-1980s (Reagan deliberately de-valuing the dollar was frequent discussion on the morning news). A side effect of this was the transfer of investment to other shores, and Japan being the primary beneficiary. BOJ dropped interest rates a great deal and a speculative asset and property bubble was launched. This culminated in the purchase of many, many high-profile assets in the United States by the Japanese keiretsu and banks.

The Japanese were in a mania, and a lot of investors world-wide were as well, and the Americans, in particular, were experiencing a kind of crisis of faith in US-style capitalism.

Like most bubbles (try explaining the dotcom), it is very hard to make someone understand what happened. You can recount the facts but not the emotional context.

James Burke had a climate special (After the Warming) in about 1989-90 where his ideal future was the Japanese running things in a world government. Seems to have captured the zeitgeist and the present (creepy to hear the bit in the beginning though about "full-scale evacuation of New Orleans").

https://www.youtube.com/watch?v=RfE8wBReIxw

Wow. This is crazy.

His optimism is touching. The bit about carbon trading struck me. Lately I've been questioning the dogma of one man, one vote. What if the only way to democratically resolve climate change is to inversely weight your vote by the amount of carbon you emit?

In about a second flat the rich will find a way to make the poor take on a lot of CO2 responsibility.

I’m amazed that I’ve never seen this.

I think this deserves an HN post of its own. Absolutely fascinating to see so much basically correct (climate refugee crisis, massive efficiency gains, enacted carbon tax, free falling PV prices, my efficient smart home setup) with so much off target (widespread telecommuting! 9 billion population! A world Japanese world government!) That seems preposterous now.

I dunno, didn't look like it got anything right to me. 6 degrees of warming and two feet of ocean, though we're only a billion and change away from 9 billion.

Since when do we have a climate refugee crisis? People go to the tropics on vacation; it's fine there.

The current border crisis is a reasonable facsimile, same in the Mediterranean.

6 degree warming however, we are thankfully a ways away.

I think blaming the present day migrations on "climate" some how is "citations needed:" they're very obviously (to me) caused by political problems and wars the US fomented. Blowing up Libya and Honduras were shitty ideas.

>> It wasn't a big commercial system; it was a big government system (you probably can't imagine the degree of shock and awe that greeted MITI people or even the mention of the minstry in the US).

Sloppy turn of phrase on my part.

I do understand the reaction of the US, and not only. I did my CS degree in 2007 and I made heavy use of my university's library that had a solid selection of Prolog textbooks, which I devoured. At some point I noticed that a few of those textbooks (the more industry-oriented) had prefaces that stated Prolog was a very important language to know because it was poised for world domination thanks to the FGCP. Curious, I dug a bit further and found that book- the one from that American journalist woman and the man from the semiconductors business. If memory serves, it portended doom unless the US spent a few billions overtaking the Japanese. Later, I found out what happened after- but I got a pretty good idea of how people in industry in the US saw the whole thing, and it has helped me understand why logic programming was relegated to a mere niche in CS following the abandonment of the FGCP.

The similarities with the present hype cycle are not lost on me, either. However, this is what happens when industry takes over science. We lose the science. There is nothing here to celebrate, or laugh at. Like I say above, missing an opportunity to discover new knowledge because some idiots want to make a quick buck is a tragedy.

I wonder if the hangover from this is behind Will Byrd’s difficulties in getting tenure despite his truly groundbreaking work with miniKANREN. Maybe people just identify KANREN with Prolog and dismiss it?

(I know it's bad form to complain about downvotes, but this one is really puzzling me. Does someone think Will Byrd doesn't deserve tenure? How could anyone think that?)

Is there a good summary somewhere of what miniKANREN is and of the difficulties finding academic support for those of us who work in industry and don't know as much about the tenure system?

(FWIW I tried to help correct the downvotes. I personally believe very very few comments should be in the gray.)

As for what miniKanren is, there are a couple of books (one being Byrd's dissertation) and a lot of papers. I think maybe the best summary I've read is actually the interview transcript at https://www.infoq.com/interviews/byrd-relational-programming.... But also, there's a summary on http://minikanren.org/, with links to the papers and talks, and an older summary on https://kanren.sf.net/. Most of these are a bit hard to understand the significance of if you aren't already familiar with logic programming. The talk on https://www.youtube.com/watch?v=OyfBQmvr2Hc is a popular description of one of the most astonishing results; I haven't watched it but people tell me it's good.

As for understanding academia, yeah, I don't know.

Most such questions come down to the fact that Academic politics is inscrutable.

One of the hilarious things about 5th generation computers is how certain people were about all this.

More like painful. I went through Stanford in the mid-1980s, about when it was clear that expert systems were only marginally useful and strong AI was not around the corner. Many faculty were in serious denial about this. You can write rules for an expert system, and what you get out is pretty much what you put in. Too many rules, and you're likely to have contradictions and the inference system gets confused. The logicians were trying hard to fix that, which generated a lot of papers with obscure notations.

Eventually we did get massive parallelism, but it doesn't look anything like the Thinking Machines or LISP machine approach.

>This was still the main AI textbook in 2014 when I did my MSc in the subject

FYI seminal books don't rotate in and out over a five year timespan, it is still the seminal book and will likely continue to be given all the fuel being burned on neural networks that generate more heat than light.

I had an intro AI class for which Leslie Kaelbling got use of a draft of Russell&Norvig (it was 2 thick stacks of photocopies, with comb bindings), before it became famous. Good call, LPK. Sadly, I donated&sold all my books when I did an extreme-minimalism housing move after grad school, or I'd still own this collector's item.

Wow that is cool, and I feel bad for you because that would be a cool collector's item! Though to be practical, I do prefer nicely bound and polished publications rather than the photocopied and combly bound.

By the way, though signifying nothing but apropos to this topic: Stuart worked with us at MCC, Peter did not.

I think Locklin is saying that AIMA is a good source for understanding GOFAI, and that GOFAI was touted as on the verge of achieving AGI, not that AIMA touted GOFAI as being on the verge of achieving AGI.

I agree with your distaste for the self-congratulatory tone.