One of the primary use cases for LLMs I've seen touted is automating customer support, the irony is hard to miss.

Also from a technical standpoint OpenAI seems to be failing at solved problems, I can understand model scaling issues, but basic infra and UI for a chat app (one way chat no less), when you're claiming to produce 10x developer tools, seems particularly weak. Especially given they have open tab at Azure and talking billions in funding.

The gap between what current AI tech promises to deliver vs. what it actually delivers is a chasm.

That was funny to me. I had to raise my spend limit twice and submitting the form results in a message - something like "we will review your request within 24-48 hours". It's a fully(?) manual queue for a company I'm trying to throw money at to use their AI to automate my customer service.

I know that's not their specialty but it made me laugh

Maybe being a 10x anything doesn’t necessarily lead to results?

i feel like we keep running into these problems repeatedly with a lot of our “disruptive” tech. scaling problems that come with usage in the real world. they’re “disruptive” until it gets to actual scale of the thing it was disrupting and reality smacks it around.

many of our complex systems are bloated because the problems consist of a lot of tiny problems that require a complex and often messy bloat to solve.

i already see this “let’s just rebuild it” attitude about some large software projects that are fairly modern. and repeatedly the people imagining a rebuild will fix the problem have little idea of complex systems and fractal depths of problems these systems must address as they scale in every possible direction. the people imagine a rebuild will somehow not require bug fixes, security patches, thousands of new layers of features, etc… etc… all done by dozens or hundreds of different people, each with different mental mappings of the overall system.

i’m obviously not suggesting that we don’t build or attempt to streamline, but omg the number of times i’ve seen people completely… just utterly fail to understand why something in the real world may be bloated or messy is unreal.

sure, it’s easy to be streamlined and agile as long as the product is new and not used in any kind of scale… until the product is used in reality, and it begins to scale, and the insane number of (very real) real world variety of complications smack it (and it’s cultists) into reality.

sometimes things are bloated, not because the previous generations were stupid — they weren’t, we’re standing on the shoulders of giants — but because scaling will naturally come with impossible-to-prediagnose fractal depths of difficulties.

What do you expect when you get a few 10s of millions of users basically overnight. Id take the more generous view which is what they have provided in a short time is very impressive.

If you paid for a product or service and didn't receive it or have it refunded then report it as fraud, because that is what it is.

It’s not a customer’s duty to chase a seller for their own money. OpenAI boasts about replacing customer support folks, they could use some of their “ai” to automate the process, yet either they do but its low quality or they dont and instead charge for services they dont provide. In my books thats fraud. They are well aware of what they are doing.

I think they also had problems before with the opposite issue. Once OpenAI introduced paid subscription, they didn't implement "back to basic" account conversion after the billing expiration, I think I used it for more than two months instead of a single one that I paid for. Felt really startup-y

They are doing a better job than Anthropic (etc); at least it is possible for the general public to use their product…

Why don't they simply use their wonderful AI products to fix this?