That is already a progress, thanks for the feedback. Appreciate it.
So you think you could help us here to sort of organizing the orientation for both ends of the problem:
- Open AI team to get more efficiency in ticket solutions.
- Forum users: use more of principle “search before you write” or any this suggestions?
Maybe a setup/tool/whatever else that would reduce redundant cognitive load on OpenAI by any form of self-orgsnized collaboration from users, especially new ones?
Maybe if we all just calm down and we talk about the problem instead of pointing at eachother we can find a constructive solution.
Some people complain because the water in their holiday pool was too wet and the customer support did not give them immediate attention (because they got a heart attack from laughing at the request).
And some people complain because something was broken like there was too much clorine in the pool and the kids have disappeared.
We don’t know which group you are part of unless you tell us what the problem was. And I tend to put people in the first one first - from experience with other complaints e.g. about a LLM not giving the full truth that exists in their very own bubble and they want to force their believes on others.
Frustrated person orientation assistant? Or no assistant?
@PaulBellow do we have access to forum data or ai to build an assistant (from your history recent posts) or should we look how to build one?
I don’t think you understand the need: the marketplace has rejected AI as customer service intercepting agents, for companies that want something even worse than outsourced third-world script readers or remote order takers. They’ll spend millions on sales and advertisement, have hundreds of people and openings in go-to-market teams and business promoters cold-calling and making presentations. Then throw up an AI that can take absolutely no action and does nothing but waste time, not realizing that support is your real sales tool.
OpenAI already has one of those AI chatbot time wasters offered as the only means of obtaining support. That’s the complaint. It also represents itself as human - a violation of terms if it was you. The one company that should be able to demonstrate an application and do AI right seemingly doing it the worst. OpenAI also has a bunch of forum support that will dump undisclosed AI output as replies, unable to understand the concern (although the “what ChatGPT said translated to English” support has been slightly better recently), but “support” still also unable to even make an API call themselves when following up a month later “still happening?”.
So an AI that cannot fix anything is not a solution to anything.
Hetzner has a very good implementation of a customer support bot. I was surprised how well it was done!
It was really asking the right questions.
On the other hand they are possibly not in the same situation where they would have to block accounts because people build browser extensions to save api cost (not to mention others installing the extension and getting their accounts blocked because the guy who made the extension is doing bad stuff) for their apps or distill the hell out of the models to make their own. Not to mention people using the service from countries that are not on the list of supported countries like Russia or China or perverted stuff or other things that are not allowed and people being surprised that the API costs money or that it takes weeks until they manually sort things out like special contracts etc…
And honestly I like the idea that the model is not capable of all that. Because it gives people time to get used to that stuff before they get overwhelmed with the stuff that is going to happen.
“Forbes Top 100”, read your original thread, lmao.
Resume flexing in some community forum.
The other two TAC threads are funny. I don’t wanna be that guy, but what if, instead of bugs, which is obviously possible, they simply revoked access for some users, where they smelled rotten eggs?
Given your comments it sounds more like the the other peoples issues, already been verified in the past, and likely monitored to some degree as expected and then got yeeted.
And now they’re getting mad they cannot social engineer their way into access to these models.
Not saying there cannot be false positives, just saying starting out your first thread resume flexing (not this one) like that didn’t really contribute to its believability, quite the opposite, because it implies to me that you already suspect subconsciously it’s got something to do with your use-case rather than merely a bug with KYC, since otherwise this info would be of zero value.
That’s my opinion.
Tobias, thanks for the uneducated opinion. But no, you’re completely wrong. I’ve never had TAC, and never been flagged for misuse. Keep trying to shill (flex?) for OpenAI though (talk about cringy).
Also, its not like defenders need OpenAI when models like Kimi K3 are at the same level or better for cheaper. OpenAI’s only edge over Anthropic was cost, but as of today, that edge is gone.
Also, they didn’t just “revoke access for some users”, they accidentally banned tons of people for no reason in an illegitimate ban wave.
There is way too much effort on trying to over-engineer a solution to the customer support problem.
Training, enablement (able to look into debug logs/troubleshooting tools), hiring more smart people, SLA reminders for responsiveness. Yes, it requires OpenAI to spend money and time training, enabling, developing better systems, but its pay off in terms of reputation and maintaining its customer base, is worth it.
I was reading your other posts now. It is pretty clear that you are upset because the algo and possible also some people decided that you are not trustworthy for TAC. You claim that it is a bug. Maybe it is - maybe it is not.
But why do you still stay:
I would only use Anthropic’s models again because their customer support bot gives you a refund when you test it for a couple minutes and you find out that it sucks in comparision.
I think it was a good way of dealing with it. I might test it again next year.
Google also has a great customer support. When you complain there then their human employees have the power to give you a refund or refuse it if you are not behaving well.
Treat the world like you want to be treated.
Ohh, “uneducated opinion”. Damn. I am stunned. How can I possibly recover from this devastating attack.
No buddy, I don’t shill for anyone. I just like to call out angry teenagers on the internet for fun sometimes.
And you clearly act like one.
Name literally any other big tech company with end users at this scale, that has good support at this point, due to the sheer scale and additionally bureaucracy forced upon them by not only themselves, but also governments.
Okay, we’re wandering into non-constructive territory again, imho.
We do want everyone’s valid concerns to be heard as they can be used to improve support for everyone. However, remember that if you dislike a user’s opinion, you can choose to not give them a heart reaction or even set them to ignore if you so chose.
There’s a few gems of posts in this thread, so let’s try to keep it fruitful, please?
Some of you are relatively new, so might I suggest dipping into this thread on the history of this place? Might give you some more perspective on how this community evolved since its inception a half decade ago. There’s a lot more to come, I think, but let’s be mindful of our guidelines here.
Appreciated!