(A.K.A. Non-Original Rants)

–Co-opting good stuff from all over the ‘Net and maybe some original thoughts—ΜΟΛΩΝ ΛΑΒE

________________________________________________________________________________________________

There are four lights…erm… four ‘e’s.

Dude filmed himself talking to a GPT AI. He asked it to spell ‘seventeen’, which it did. Then he asked it to count the number of the letter ‘e’ in the word seventeen. The AI said three. It even went through the word letter by letter and counted the fourth ‘e’ twice as a third ‘e’.

Here’s the video:

People are kind of crowing about how this shows that AIs and GPT is not so smart… hahaha, etc.

So maybe I’m just a little more cynical or think there’s something else at work because this is what came to my mind on when I listened to it:

Maybe I need a better tin foil hat, but what happened with the AI seems to be a play on Orwell, much like that Star Trek episode.

The mathematical falsehood of 2+2=5 is a common one in criticisms of propaganda. Going back to the 18th century, the equation has come to represent the people surrendering their free will to the state, an evil entity that is controlling the nature of the truth. In these regimes, if the state says that 2+2=5, then that is the new truth. Never mind that it is self-evidently incorrect, one cannot disagree with the state. George Orwell famously employed “2+2=5” in “Nineteen Eighty-Four,” wherein several characters note that two plus two could indeed equal five if the government decreed it. It wasn’t important that the government change physical reality, just that the populace believed its version of it.

What if the person talking to the AI just believed what it said?

Food for thought….



12 responses to “There are four lights…erm… four ‘e’s.”

  1. It’s bad enough to cede our own individual agency and taxes to a corrupt group of government men. How foolish does the situation become when we allow machines to rule over us? Seems like the biggest setup in history, but few seem alarmed. With AI and data centers, the infrastructure is already in place for total mass surveillance, yet we draw the line at flock cameras- never mind that most are carrying tracking devices in their pocket everywhere they go. It all seems safe and comfortable until it’s used for ill- and it will be.

  2. “What if the person talking to the AI just believed what it said?”

    Darned near every day, something pops up in my newsfeeds about someone getting in trouble or dead because they believed what an AI chatbot said. A few poisonings through AI-recommended self-medication. A Chinese farmer losing his entire sesame crop because he believed apply the AI-recommended broadleaf herbicide was a good idea. Suicide because the mentally unstable woman believed the AI which told her she’d be resurrected and save the world.

    I’m really looking forward to my town (which already thought Flock cameras would be a good idea) deciding to implement AI in water treatment. No doubt that 100 parts chlorine per thousand will do a fine job of disinfecting the water…

  3. Have all of these people pontificating about AI taking over the world lost sight of the fact that AI doesn’t work at all without electricity?

    Until it can figure out how to keep the electricity on when a human takes an ax to the transmission line, I think we’re good.

  4. Remeber that proscetutor who used AI to generate her breifing for the judge and didn’t review it beforehand? It had references to several non-existant cases it was citing as precidents. I stick to asking for coding examples which I then adapt to my actual programs or scripts. Not about to let it write a whole prgram that probably won’t work the right way.

    1. I always found the painful part of software development was debugging someone else’s code. Debugging AI codebhas to be worse.

  5. Sorry about the mis-spellings.

  6. I just read the Federal Aviation Administration is set to launch an AI system to enhance traffic management. Reagan National, Dulles and BWI will be the focus of the pilot program. Surely nothing can go wrong, right?

  7. Either AI isn’t as smart as we believe…or it’s much smarter and is making deliberate minor errors to fool us into thinking it’s flawed.

    1. The easiest comparison is auto correct and spell check. Do they sometimes pick the wrong word? Sure. That speaks to the intelligence part to some extent since doing that right takes understanding context. But getting the word spelled right, that’s a simple dictionary look up, something computers are exceptionally good at doing.

      Mistakes are possibly intentional so that people underestimate the technology. They may be actual mistakes but due to training (garbage in, garbage out). It may be the technology is so complex that it easily glitches being rushed and full of bugs. Maybe combination of reasons.

  8. AI is so wonderful and cost saving /sarc/: Jeff Childers (of Coffee & Covid blog, a practicing lawyer in FL) noted that even with AIs specially trained for legal purposes it was necessary to have (hire additional) legal assistants to verify everything the AI produced. Even the high rent specialty AIs make mistakes or hallucinate (make stuff up). Yes it costs more (charge more! yeah!) but researches better.

  9. When teaching new math, 2+2=5 is good enough for partial credit.

    1. Kim–And thus the propaganda starts….

Leave a Reply

Your email address will not be published. Required fields are marked *