Here's what I sent them:

I don't use generative AI. I have a computer science degree so I understand how large language models work, and I don't believe that they have any value. They are just stochastic parrots. That they so beguile their users with vapid statistically-probable output is distressing.

But LLMs have still changed my life, because the training models are forever scraping my personal web site, costing me bandwidth and money, violating the copyright on my original content without my consent. The datacentres that house LLMs consume vast amounts of energy and fresh water, an environmental disaster in the making.

I expect that in the future, LLMs will once again change my life as I'm called to cover for an entire generation of workers who lack important life skills such as composition and critical thinking. I'm not exactly looking forward to it.

@MxVerda I've been a programmer and circuit designer most of my life -- and I've learned that even "binary computers" aren't binary either. Nothing in the universe is.

For the computers, they take some circuit value (typically a voltage) and say (for example) 0.5 volts will be '1' and 0 volts will be '0'. In reality, the voltages have noise, so they get a lot closer together than that -- but the main problem is that to switch from '0' to '1' it has to go through all the voltages between 0 and 0.5. And since the circuit can't be infinitely good at telling the differences between voltages, there will be some region in there where it can't decide what it's seeing. And since they can't be infinitely fast, there will be a period of time where it is undecided.

And there are some circumstances where the circuit must decide in that period of time.

So we make "synchronizer" circuits to minimize the probability of failure -- but it will always fail at some point, because we have assumed binary behavior from something that is analog.

[And even though you can prove rigorously that this problem is fundamental, there are people who get patents believing they have made the circuits "truly binary"]