Where do you see any "name calling"?

You've got nothing. Pathetic.

I don't get the the hate for FORTRAN either. Though it's worth pointing out that the algorithms in the paper are very obviously not in FORTRAN.

LOL! You think you were "discussing" something?! What a joke!

Here in reality, you were saying a bunch of nonsense and ignoring easily verifiable facts. That's not a discussion. That's you denying reality.

Some people aren't very bright. I guess that's why they think LLMs are magic.

Are you sure about that?

Because it does not reason, consider, analyze, or even understand your prompt ... or the response. It just generates next-token probabilities, one at a time, until a stop token is emitted. That you see this as producing responses that are 'valid', 'insightful', 'incorrect', or whatever is what makes the illusion work.

That's simply not true. AI Companies are losing money on inference. That's reality. You can continue to deny it if you like, but it won't change that simple fact.

Again, the price you are paying is not even close to the actual cost. This is why AI companies are burning money. Companies are not going to lose money forever so that you can enjoy cheap access to shitty chatbots.

This isn't difficult.

LOL! Don't you freaks have a single other example? I didn't think so. Do you know why?

Diversity, Equity, and Inclusion in advertising works. The almighty market has spoken.

Get woke or go broke.

Cheaper != cheap. New technology has made inference less expensive, sure, but not nearly enough to matter. Again, these things are expensive. OpenAI, for example, was forced to raise prices just to slow the bleeding. They are absolutely hemorrhaging money, even on their $200/month "pro" plan.

That was far from universal, as anyone with a red box 30 years ago could tell you.

Later payphones were surprisingly 'smart'. Not only could handle payments, they could even be scheduled to call home to report. Others you had to initial the call,and hope no one answered! Sometime around the turn of the century, I worked on a project that used that data to create dynamic collection routes.

I hate to break it to you, but most people aren't bigoted freaks who can't handle seeing non-white people in ads without crying.

Corporations don't care about social issues. They only care about money. They put "DEI" in ads because it works. Get woke or go broke.

Only stupid people who don't know the difference between profit and revenue.

Don't let the fact that these things are inescapable at the moment fool you. These things are absurdly expensive. Everyone is banking on the tech improving rapidly and the cost falling dramatically. Enjoy the affordable access while it lasts because it won't last long.

So what? They all produce nonsense the same way.

It's astonishingly bad at summarizing text. It will ignore important details and 'hallucinate' others. Oh, and if the thing you want it to summarize isn't accessible or doesn't exist, it will still provide 'summary'.

The output looks really good if you don't bother to check it for accuracy.

They suck at summarizing text because they're not actually summarizing text. All these things do, all they can do, is next-token prediction. That's why it doesn't matter if there isn't any text to summarize. Next-token probabilities are produced the exact same way, no matter what the context happens to be.

To be fair, I think we're all guilty of that. If not when it comes to AI generated nonsense, then a book or some other media. It's impossible for us to be experts in everything, so we all lean on expert opinion. We also tend to associate confidence with certainty, which is fine most of the time, provided we don't also mistake certainty for accuracy!