[Dprglist] The dead end of Artificial Intelligence

Sat Aug 1 15:44:10 PDT 2020

On 2/08/20 9:03 am, Karim Virani via DPRGlist wrote:
> This not a technology assessment, but an interesting read:
> 
> https://www.freshconsulting.com/wp-content/uploads/2020/06/Why-Robotics-Fail_Fresh-Consulting.pdf
Hi Karim,

Thanks for that link -- it is as you say quite interesting reading. While
robotics and AI aren't the same subject, this does remind me of a story...

It's a quiet Sunday here so I'll tell it. Please pardon the length.

----

In 1979 I'd just finished high school and was looking for a university
to attend. Our family had just moved out to California and I went on a
tour of SRI since I'd been admitted to Stanford (but in the end I
didn't go as we didn't want to pick up the tab).

That tour of SRI included a demonstration of a system that had been
populated with some US Navy data, had a natural language interface,
and you could ask it questions, typing directly on the command line.
The demo question I remember was "What direction are Admiral Fitzgerald's
shoes facing right now?". The answer from the system was a clarifying
question: "Can we assume he is facing the same direction as his ship?",
to which we answered "yes" and it then gave us the bearing of the ship
he was commanding.

This feat of 1979 magic was the product of the Cyc ontology project, the
brainchild of Doug Lenat, who would later take that project back to Texas
and start Cycorp.

Fast forward to 2000. I'm having lunch with Doug in Austin, Texas, us
trying to figure out if there might be a fit within Cycorp for the likes
of me (there wasn't). I asked him about that 1979 command line question
and he confirmed, yes, that's exactly what they were doing back then.
Now, rather than me go on much further for those not-so-interested, I'll
provide a link to Roger Sperberg's summary of the XML Conference that
year [1].

At that conference Doug Lenat announced that he would "open source" the
Cyc upper ontology and much (but not all) of the Cyc ontology engine as
"OpenCyc". He did, and I played around with it a bit in the early 2000s.
OpenCyc lasted until 2017 when Doug retracted that promise because, as
he stated, it "proved to be more confusing than it was helpful." [2]
Some may disagree, but it's his baby and he can kill it if he likes.

If you were to read both of my last two references in their entirety
you'd see the disconnect between the "early" enthusiasm for Cyc (despite
by 2000 it already been a two decades-old project) and the withdrawal
of it from public sphere 17 years later. But Cyc was hardly alone in its
failure. As Marvin Minsky is quoted as saying, "Unfortunately, the
strategies most popular among AI researchers in the 1980s have come to a
dead end." [3]

Earlier this year I was reading various papers by Rodney Brooks' [4],
who provided a much-needed corrective on that somewhat silly enthusiasm
for AI. And while Behaviour-Based Systems (BBS) are an entirely different
rabbit hole, it's one that's actually turned up some interesting and
productive outcomes, possibly because it was never so ambitious in scale
nor so clever in its marketing. A few days ago the NASA Curiosity Rover
took off for Mars, and onboard it'll be using more BBS technology than
AI/Expert Systems technology. Curiously.

And this reminded me of just a few days ago reading Maureen Dowd's
recent interview with Elon Musk in the New York Times [5], and reflecting
on how truly silly is Musk's fear of "dangerous AI", and that anyone that
disagrees with him isn't in his estimation as smart as him. Well, he may
be an intelligent guy, but clearly, he and Kanye West are both simply
nuts. Their main talent, like Steve Jobs, is marketing.

And this gets us back, full-circle, to robotics. Robots have always been
an intriguing field, the stuff of science fiction. Clever people like
Doug Lenat have spent their entire careers going down very deep rabbit
holes, but have found no actual rabbits. Things that look a bit like
rabbits, but not rabbits. We're nowhere near the film fantasy of Ex
Machina, and I frankly don't think we will get there. Not before our
species implodes. Or goes to Mars. Or Venus [I once met a guy in Fair
Oaks, California, who thought he was from Venus, but that's a different
story...      ...lots of people in Fair Oaks are probably from Venus].

That's not stopped people from selling the rabbit hole journey and
gaining billions of dollars in government grant money, but I'd have to
say that I'm about the opposite of Elon Musk: I have no fear whatsoever
that humanity is doomed by some future plague of AI robots that become
our overlords. For that to happen, we'd have to be a lot better at
building it than we are at marketing it, as even the marketing has in
the end turned up as a failure. I'm a lot more afraid of rogue
governments mounting weapons onto a Boston Dynamics robot dog (battery
life: 90 minutes!) and remotely controlling them for crowd control. Or
sending thousands of them out as warriors. But even though Boston
Dynamics robots are pretty amazing as hardware, they don't really
qualify as what I'd call "AI", not at least in the sense of what Elon
Musk fears as "AI". They're only nominally autonomous and I wonder
how much of a world view they actually embody. They don't on the
other hand seem to be a BBS either.

But *very* clever marketing [6]. Maybe that's actually the problem with
all of these projects: people begin to believe the hype, which is
certainly a lot more fun than the more-limited reality. Who wouldn't
like a self-driving car? Uh, actually, anyone who understands how
dangerous that would be, like, uh, Elon Musk...

sigh

Cheers,

Murray

[1] https://www.xml.com/pub/a/2005/08/03/extreme.html
[2] https://www.mkbergman.com/2034/fare-thee-well-opencyc/
[3] https://en.wikipedia.org/wiki/Cyc
[4] https://service.robots.org.nz/wiki/Wiki.jsp?page=RodneyBrooks
[5] https://www.nytimes.com/2020/07/25/style/elon-musk-maureen-dowd.html
[6] https://www.youtube.com/watch?v=wlkCQXHEgjA
...........................................................................
Murray Altheim <murray18 at altheim dot com>                       = =  ===
http://www.altheim.com/murray/                                     ===  ===
                                                                    = =  ===
     In the evening
     The rice leaves in the garden
     Rustle in the autumn wind
     That blows through my reed hut.
            -- Minamoto no Tsunenobu