>but the AIs slowly start to drift.

It's essentially the tower of babel. Each person will devolve to speak their own internal language only they understand. Each language will need to be encoded down to its meaning to be reinterpreted. None of us will know if the transformers are accurately decoding, or if the other person is accurately interpreting the decoding (which is arguably already a feature of human language without the computers in-between.)

a business can be sold to a younger generation.

I dont have to be an expert to recognize when something is said with authority and confidence. Any time a model says something with conviction, my instinct is to double check.

Now if they taught them to express uncertainty and speak in terms of probability, I might be more likely to be blindly fooled by some kind of uncertain conviction.

youve got to get rid of the list with the icons as bullets. vertical or 3-4 of them horizontally is a dead giveaway.

your posters also dont suffer it, but most of the ones in the wild have 3-7 extra catchphrases or taglines thrown all over the poster. same but different. this but that. not this that. chatgpt just really likes to add phases and annotations all over posters.

“Good enough to replace humans” isn’t necessarily the benchmark.

The question is is a computer with a human stronger than a computer without a human. At what point does the hybrid go from being stronger, to the human getting in the way, or steering the computer in more wrong directions that right ones, or the human not being able to keep up. Does the human add enough extra randomness to be of value for a while, even as a minor co-processor.

Underappreciated point. The point of inflection is where humans switch from being a driver to a liability.

But I don't think it's randomness, because that would be easy to add. It's more like a different perspective on the training data, a different set of perception categories, and a different set of skills used to work with all of the above.

Those skills aren't very efficient, but they're the best we can do. We're used to their strengths but we don't like to think about their limitations.

It's completely plausible that AI will replace some of them, and not implausible it could replace and improve on all of them.

It's also not necessarily implausible that AI/we decide that performance is better with humans in the loop somewhere, even if its reduced to something like mechanical turk.

I have little doubt that for many years, AI + human will be better, and then eventually AI will be so good that humans mostly won't offer anything. That latter state will probably take at least 10 more years, but will likely happen within our lifetime.

When was the last time you tried to log into something like Google, Apple, Microsoft etc without your phone nearby, on a fresh computer?

What if my phone is dead? Stolen?

In the same vein as 2fa, going up to a fresh computer and trying to log into anything is now a nightmare. Every service has a 2fa that somehow loops into another provider that also has 2fa.

And some 2fa, if not many, make accounts weaker. Apple's solution to get around 2fa is to put in SOMEBODY ELSES phone number that I trust, as a backdoor. It's an insane solution. And its normalized, and nobody questions it.

While we're at it, the insanity of those "recovery questions", made even worse when many companies demand the same ones.

but you can type anything in the recovery box.

its just three additional passwords.

should you type the same brother name in each time? should you even answer with a name?

don't get me wrong, they were insane, but they can be repurposed.

Good luck remembering what you put there when asked 10 years later.

Yeah, many do not even allow you to enter in a custom question, which would be marginally better--at least for users who care about security.

Then you'd have the option to craft a question which is (A) personally memorable (B) specifically phrased to avoid ambiguity, (C) not public or guessable, and (D) unique to a service.

* "Where is your first pet buried?"

* "While on vacation, The Noodle Incident happened in which city?"

* In what year did you first see Example Band live in concert along with Alice and Bob?"

put it in your password manager?

It’s literally translation technology told to guess what’s most likely next instead of translate what it was given.

It makes total sense it’s good at regurgitating the edge cases of language.

Am I looking at the academic theater of sycophantic LLM fed cranky prompts?

"fractal recursive residual coding" = 45% size reduction per iteration after 13 passes of the output of LZMA2 piped back into LZMA2? If true, that alone is crazy.

it reads like Adam Curtis, although a bit more repetative - https://www.bbc.co.uk/webarchive/https%3A%2F%2Fwww.bbc.co.uk...