Has anyone ever tried to communicate to open AI support? Their support is basically nothing but chat GPT with a filter on it. I am trying to help somebody set up SSO with Open AI and it’s been two weeks of back and forth sharing videos showing screenshots that their AI system completely ignores and asks me if I’ve tried clearing my cache for the umpteenth time.
I’m helping a company set it up but they are wholly unprepared for how unreliable and shitty it is. So often I hear people excited about all of the things they can do with it and I worry that they actually believe it works as advertised.
completely ignores and asks me if I’ve tried clearing my cache for the umpteenth time.
Sounds like regular bad human tech support! We did it, AGI is here! /s
It sure sounds like it’s time to check my 401k and get out of US stock based funds while there’s still something left.
The current Shiller PE ratio is over 40. The only other time it was that high in recorded history was 1999 into 2000. And that uniquely crazy time was due to… lemme just check my notes… a massive tech bubble. Fantastic.
The mean and median for the ratio are both below 20. I like the graph on this site: https://www.multpl.com/shiller-pe
(for anybody not familiar, the PE ratio or P/E stands for price/earnings, or how high the stock price is relative to how much money the company makes)
I just ended a relationship with a company I had done business with for 11 years because I found out they had switch to a forced-AI customer service model and it didn’t work for me when I called them.
I hope this is partly because some companies are becoming wise to the fact that when their employees use AI they are providing the AI with privileged information that can be leaked (which we’ve seen from law firms recently).
But honestly I think this is down to people working at these companies actively running up against the friction of what AI can’t do or doesn’t do well.
my company has an AI ban because of information confidentiality.
it is also generally useless for the work we do.
Are they hiring? My last job fired me because I refused to use the useless LLMs they spent millions on.
Nope. Hiring freeze thanks to Trump’s funding cuts. Plus very few people leave because the job is easy and the benefits are very good.
Use AI, get burned. Even the slowest monkey would get that eventually…
Oh, AI can be very useful. Just not the generative stuff that is currently trying to consume all resources of the entire solar system for nebulous potential benefits.
A good example of AI that just works is document scanning. Get a picture of a document, locate text, OCR it, figure out which parts of the text correspond to entry fields, auto-populate the fields. That works pretty well and can greatly speed up manual data entry. It’s not perfect but the success rate is pretty good due to the constrained problem space and even if you have to check all fields and manually correct 10% of them you still save a lot of time.
An early example of this is the automated parsing of hand-written postal codes. That iteration of the tech has been in productive use since the 90s! (Yes, that’s just OCR but OCR is considered a field of AI.)
It’s one of those unexciting applications of tech that don’t make major waves but do work.
it’s really good at RTFMing on my behalf.
I can usually tell the PRs that have been written by AI though. They look like they were written by a junior and there’s comments and emoji fucking everywhere
But…but…banana.
In all seriousness, I still feel like there are limited use cases for AI. I don’t think it was smart to go all in on it with our energy and natural resources, especially for implementations in which nobody knows exactly what it’s doing, how, and there’s not a human in the loop. Unfortunately, I think that describes the majority. I kinda wish research for AI had been treated more like nuclear energy research.
I think it can be useful, lile you said, but the user needs to understand the work before offloading it imo.
This is why at art school we weren’t taught on computers. They said by the time we graduated, the programs they would have taught would be obsolete. If you know the concepts and the critical thinking skills you can quickly learn the program any potential employer will use.
I make enclosures on boats. I make patterns out of paper and cut and sew everything by hand. Many people want to get into the trade and buy $100k plotter/cutters and sit at the bottom of the learning curve pulling their hair out. They should all start with a $50 pair of shears and figure out the concepts and understand what they’re building before worshiping at the alter of technology.
I design the technologies themselves and their integration, and I still try to get new people to work with the tech at around a 1980s-1990s level so they can understand whats happening under the hood on modern ones.
I disagree about not teaching the applications - there may be new versions, but the concepts carry through - but the crucial part is the foundational knowledge, I agree.
This was the late 90’s and I think it’s quite different now. Still analog to begin though.
My end of year reviews are done. I don’t need to prove I’m using AI to meet one of my goals of the year.
good
They used AI until they had to deal with the bullshit that comes from inevitably using an auto complete that hallucinates and even lies directly to you about what it did.
And it has to lie, it has no idea what it did or why. Nor any other ideas.
The problem I face is that it’s hard for AI to earn my trust as a tool because it delivers 50-90% correct code depending on prompt, model and context.
My latest experience
I am vibing a minor program and get like half way, when I then resume work a month later I end up burning $10 to implement one use case+tests because I don’t break up the task enough.
This makes me think I probably got rusty using AI in that month, and possibly Google fiddled with Geminis performance parameters to make more money.
Experiences like that doesn’t improve my confidence in AI.
You don’t get “rusty”, the system is rigged to capture your money.
You didn’t save any time and you spent more than you would have.
Stop being a sucker, and fire Claude.
I don’t doubt they are constantly trying to squeeze money out of us, especially with the shady deals between Nvidia, openai and the other group.
I am in the stage of discovering what the LLM can be used for, and it’s currently a fairly small list of tasks I trust it to do.
I’m far from an LLM nut, especially when it delivers mediocre code at best, but sometimes slop is an acceptable solution. And often it ends up taking more time, but that happens anyway.
/End rant :-)
And it only cost an entire community their water, raised everyone’s electrical bills, and required endless investor cash that can never run out or else the bubble pops.
Coding is free if you use your brain.
