The same thing has been happening in music production for the last few decades.

Its gotten easier and easier to make music, so the market is flooded and the average quality is lower.

But at the same time I would say there is also more music in the good tiers than there used to be. Certainly the variety has exploded.

What id say is that in neither case is it really a problem if its in an ecosystem that already has some form of rating system.

Shit apps existed before AI, and they fell into oblivion with 1 star ratings. The same will continue to apply moving forward.

Most important step is to decide exactly where you want to draw the line.

You started by saying they produce the issues, but are now explaining what I said above, that they are a product of upstream social issues.

Is this worth discussing.

Usually id agree but im not sure that really applies here.

You’re confused, those things are the outputs out late-stage capitalism.

Arguing against something I didn’t say. I never claimed companies “by default make good decisions.” I said thousands of independent companies and individuals repeatedly choosing to pay for a tool is reasonable evidence that it provides some value to them. That’s not some corporate infallibility, it’s revealed preference.

Your personal observations are anecdotes. If your claim is that LLMs are far more often harmful than useful, support it?

Then we’ve narrowed the disagreement quite a bit. You accept that LLMs can be used effectively, you’re just asserting that they’re misused far more often than not.

But both that and “anyone with eyes can see thousands of companies are wasting their money” are still just restating your conclusion baselessly.

Widespread use alone isn’t conclusive proof of productivity, granted. But your alternative seems to require assuming thousands of companies are all independently making the same stupid decision, spending real money on tools that provide no value, while you’ve correctly identified what they’ve all somehow missed.

Corporate adoption can be wasteful, and some uses absolutely are poor. But that doesn’t establish that LLMs don’t improve productivity. At most, it shows the gains are task-dependent and easy to misuse.

That analogy only works if the AI literally stops functioning after a few uses.

A better comparison would be a tool that performs badly when used for the wrong job or by someone who doesn’t know its limitations.

“Sometimes produces bad output” isn’t the same thing as “shitty tool.”

Which just feeds back into what I’ve already said. Shall we go around again?

Explain why.

Agreed 🤷‍♂️

Conflating profitability with usefulness. AI companies are burning billions on R&D, infrastructure and expansion. That doesn’t mean the product isn’t already creating value for users.

The real question is whether that value eventually exceeds the cost of providing it.

These are just assertions.

If they were true, nobody would be using them.

Millions upon millions of people are using them every day?

What are you waiting for?

A product can be useful to its users while the companies providing it are still unprofitable because they’re spending enormous amounts on R&D, infrastructure and expansion.

This regurgitated, thought-terminating sentiment confuses utility with the providers current profit margin. Its fair enough to not understand how businesses operate but taking shots in the dark while not really knowing what you’re talking about isnt sensible.

Thats a take, but if you pair that sentiment with what I just said it paints a funny picture.

That’s possible, but I’m not convinced. At a certain point, a world dominated by services that most people can no longer afford stops making economic sense.

It’s easy to look at the current trajectory and assume it continues indefinitely, but systems are more adaptive than that. As the playing field changes, businesses, governments and consumers change their behaviour too. The incentives themselves shift as the consequences become clearer.

Imo, they’re just doing the same shit slightly faster. Whether or not it has value is separate and mostly in the eye of the beholder.

Slop is a thought-terminating buzzword, you’re relying on it too heavily to make your point and it undermines you.

You’re talking about the cost to the consumer, where I was talking about the energy cost of an individual unit of work by a given model, but they’re both relevant. The underlying energy cost of generating a token at a given level of capability has been falling as hardware and inference become more efficient. Those efficiency gains can feed through into lower token prices for users, but also a more capable model can often complete the same task with fewer tokens, fewer retries and less prompting.

So even if the headline price of a new frontier model looks similar or higher, the actual cost, both in energy and money, of getting a given piece of work done can still fall substantially. It’ll keep doing so too, its still in its infancy.

If you look at how exponentially cheaper individual units of work have become over the last few years there’s no real reason to worry about price gouging. Particularly when there’s still competition in the market, and open-source models.

Without those you’d have a point.

Im in a combined coding and marketing role so for me its a huge time saver. Far from perfect, but so am I.

What were they doing before?