Neural Numbskulls

Adam Englebright -

Tobias Revell linked to this Verge article from 2019 about how 40% of EU startups classified as "AI companies" are either not at all meaningfully using artificial intelligence, or are using it in a peripheral capacity[1], but aren't correcting the record because investors are into it.

It reminded me of something that happened at an event I went to with my dad a couple of years before the article was published. The event was an evening of the talks on 'artificial life', and he was one of the speakers, billed as a 'robot enthusiast'[2]. (The other two speakers were a bit more artsy and theoretical and, biased though I am, generally less interesting[3].)

Dad bought along a load of robots and other robot-esque things he made, and spoke about our tendancy to ascribe consciousness to robots, like a little R2 unit he made from a bin. I drove it around during his talk, but because he was talking as though he were giving instructions to it, and I was off to the side of the room and the audience apparently weren't looking, it seemed a good number of them thought it was running without a human driver[4].

Afterwards, we were sat at a table chatting to some of the audience members, and had a bunch of little robots on display, one of which was a lad which switches itself off once it's been switched on, and then back again—a bit like a perpetual useless box, except the switch mechanism is on the robot's head. Turning it on starts a motor which makes his arms swing, which flicks the switch and turns it off, so left to its own devices it'll turn itself on and off until the battery runs out.

One guy came up to my dad and asked how it worked, and my dad (who was, at the time, doing a masters in computer science and is currently doing a Ph.D., specialising in machine learning) made a flippant joke about it running off a neural network. The guy pulled a slightly perplexed face, said "oh really? like..." and proceeded to try and articulate how he thought that might work. Dad's ability to keep a straight face and his interlocutor's credulity on hearing those magic words were able to sustain a conversation on that basis for, no joke, about ten minutes. I just watched in bewilderment.

To quote from the Verge article:

...the report’s findings show that there are clear incentives for companies not to speak up if they’re misclassified. The term “artificial intelligence” is evidently catnip to investors, and MMC found that startups that claim to work in AI attract between 15 and 50 percent more funding compared to other companies.

People will believe any old shit if you tell them there's a neural network involved. (I believe this lesson to be highly transferrable.)

  1. The article also highlights that a lot of uses of AI are extremely banal and straightforward, which chimes with my experience of other hyped fields: people's use of the phrase "data science" can be extremely pliant and very often doesn't venture beyond "linear regression". ↩︎

  2. He competed in Robot Wars a few times (can't believe I'm linking that video twice in consecutive posts), Techno Games once (I can't easily find a video for that, I don't think it ever really attracted quite the same level of obsessive fandom as Robot Wars did), designed some of the house robots too, and generally just likes spending time tinkering with this kind of thing. ↩︎

  3. Brighton has loads of events like this—three people talking about something vaguely interesting, free tickets, in the function room of a pub. They were more like the kind of folk you often get talking at these—immersed either too much in fine art or academia to be able to explain things in a manner that's either explicable or interesting. ↩︎

  4. which I think proves the point he was getting at. ↩︎