nuurdin

Smack-Fu Master, in training

Jesus, this is going to be like MS Outlook crossed with the smell of cat piss.

Jesus, this is going to be like MS Outlook crossed with the smell of cat piss.

And microwaved in the breakroom with fish.

OpenAI has started billing overages for our company as of this month. Microsoft added usage-based billing, too, though it requires setting up PAYGO and you can't go over base usage if you don't do that (we haven't).

Every development group at my company is reining in usage, wanting caps installed, demanding use justification. It's not stopping use of AI, but it has meant that there is almost zero interest in agents and even where there is interest, management isn't willing to pay for it. What looked like rapid expansion six months ago isn't coming to a screeching halt, but that growth curve is getting flattened hard.

I have a very dark feeling that the AI crash is coming soon, and it will arrive faster if Anthropic and OpenAI do their IPOs.

Why should I care about what OpenAI wants to do?

I expect economic value from AI and I am not seeing it. Im sure they want me to spend millions of dollars for nothing of value for my company in return.

I am fine with letting copilot exist to appease the managers that we have an AI deployment when its a small cost per head and I can control my budget. I will not sign off on token API's.

snafuu

Smack-Fu Master, in training

This is kind of like if an oil company was bragging about how much fuel a new engine could burn through. It would probably have convinced somewhat more people before prices got jacked up.

DrewW

Ars Tribunus Militum

Last edited: Yesterday at 8:20 PM

I’m confused as to what this means for general chat-only users. Sounds like there’s no real change for them?

General chat-only users are hugely expensive and the least likely to pay; the AI companies are desperately trying to get rid of them, hence all the stuff about loops and agentic workflows, they need people to be burning tokens continuously on an enterprise budget

General chat-only users are hugely expensive and the least likely to pay; the AI companies are desperately trying to get rid of them, hence all the stuff about loops and agentic workflows, they need people to be burning tokens continuously on an enterprise budget

Most banks are break even at best on checking accounts...

The desktop version of the new tool can also access and modify your desktop files or use a built-in browser to access online resources. OpenAI says an updated ChatGPT Chrome extension will also let you perform web-native tasks without switching out of the browser (OpenAI says its dedicated Atlas web browser is now being sunset, less than nine months after its launch).

Nice. I see the next attack surface. Rigging web pages with exploits that mess with AI agents when they come calling. (Why would I trust OpenAI's browser security?)

Despite all the carefully chosen words to give the appearance that these SALAMI (Systematic Approaches to Learning Algorithms and Machine Inferences) generators are very useful, they're probabilistic bullshit generators.

Not even the person that thinks they have the least important job in the world would benefit from this (except scammers).

You'd be better off spending time doing it by hand, whatever the hell it is, cheaper too!

timby

Ars Tribunus Militum

"Stay with a project for hours if needed" ... and then, since, y'know, hours, it'll slap you with a bill for twenty thousand dollars' worth of token usage.

Sarty

Ars Tribunus Angusticlavius

"Have you ever wished that outsourcing were something you could do to yourself? We have exciting news!"

I'm honestly a bit surprised that their customers aren't more nervous about that possibility. Not necessarily the rank-and-file ones, who are potentially looking at getting laid off one way or the other; but the ones actually cutting large checks for 'digital transformation'.

The end goal of running an "AI" company is not to run a temp agency with an API; so why would you expect them to keep contenting themselves with selling you temps should they ever actually get good enough that they can do better?

This was musk talking, so it was delivered with his characteristic taste, charm, incisive wit, and clear sense of what is actually practical; but the 'what if we just replaced software companies and called it macrohard because we're 10 lol" proposal is far from the most unserious idea in the glorious world of AI grifters; and actually more or less what you'd think people would be expecting if they consider the bot salesmen to be combining the old body shop/outsourced professional services business plans with a voracious hunger for literally any dataset they can get their hands on.

The sale is, presumably, simple enough when it is couched as a way for you to not employ people to do whatever it is you do; but when you are quite literally buying the capacity to do that from someone else why wouldn't you expect it to be them not employing people to do whatever it is you did tomorrow?

tRexx

Ars Scholae Palatinae

Why do I have a feeling that costs and price competition are going to be a problem with all those billions in datacenter costs that need to be paid back?

Sarty

Ars Tribunus Angusticlavius

It's pretty poor manners to suggest somebody is getting paid covertly to express an opinion when the much simpler answer is that people who mistrust or dislike the general AI field are likely under-represented in reporting on it, and those who think it's the new hotness and a bold, exciting new frontier are likely over-represented.

I don't think Beth Mole is secretly in the pocket of Big Gross Medical Weirdness. I think she finds gross medical weirdness fascinating, and her writing on the matter draws enough clicks that her bosses encourage her to keep it up.

“Never assume malice when naivité will suffice”

OpenAI has started billing overages for our company as of this month. Microsoft added usage-based billing, too, though it requires setting up PAYGO and you can't go over base usage if you don't do that (we haven't).

Every development group at my company is reining in usage, wanting caps installed, demanding use justification. It's not stopping use of AI, but it has meant that there is almost zero interest in agents and even where there is interest, management isn't willing to pay for it. What looked like rapid expansion six months ago isn't coming to a screeching halt, but that growth curve is getting flattened hard.

I have a very dark feeling that the AI crash is coming soon, and it will arrive faster if Anthropic and OpenAI do their IPOs.

This.

They HAVE to show a profit or a clear and obvious path to profitability, for an IPO to be successful. They're burning what's left of their VC funds to lure in and addict users to their product, while still mostly hiding the costs. Get them dependant on it, and then jack up the price.

It's literally the same exact method drug lords use on their victims.

Companies, always in a fucking hurry to cut costs, will fire people as AI takes over, thinking they're making a good fiscal decision, but the AI companies are still not charging what they HAVE TO in order to become profitable - even to a tiny degree, let alone enough to pay off the investors in a fucking human lifetime. As more companies sign up for the drug, the costs to AI rises - NO scales of economy there. The expenses rises proportionally, without any scaling.

So, the real question is will enough companies pay enough to keep AI's nose above water? And will JUST keeping AI's nose above water be enough to suit investors? After all, the investment in that alone is LIKELY approaching $2 trillion or potentially more. Repaying VC's who sunk THAT much capital in to this mess means almost none of them will be repaid in a human lifetime UNLESS businesses are willing to pay MORE for greater productivity. My guess is they won't, since more won't be as cost effective as what they do now in actual useful output.

AI STILL isn't perfect, and needs competent people to evaluate and test its output before pushing that to product. And they will probably have to pay more for that level of competence than before since there will be a lag between being fired from a job and being needed for babysitting the AI. By then, I'd expect most who got burned in the employee layoff AIpocalypse won't be very eager to come back to babysit what got them fired.

It's best for the tech sector if AI just collapsed and died before people are fired, rather than after.

It's best for the tech sector if AI just collapsed and died before people are fired, rather than after.

Solution? The 9999th edition of the tech bro funded "Learn to Code" campaign to wash it all away.

Scheduled Tasks, a souped-up version of cron jobs that can “take repetitive tasks off your plate”

Well of course, you'd want a souped-up version to use a bowl. Or at least a cup. /s

Jesus, this is going to be like MS Outlook crossed with the smell of cat piss.

One of my routine tasks is trying to get time stamps, not just date stamps, on the previous 12 months of Outlook emails when viewed in the 365 In Box view.

I surely hope the billions spent on ChatGPT can fix that little problem.

If your job can be done by an AI: Expect to be replaced by and AI.

If you are using an AI to do your job, you are helping improve the AI to replace you.

"can “stay with a project for hours if needed, and turn a goal into finished work.”

Built by billionaires, for billionaires.

So they invented openclaw... great. A bit late, but whatever.

Agentic tasks will blast through your token budget so fast it'll make your head spin. And no doubt it'll be locked into their models. I think I'll pass.

Well, hardly unexpected. General AI chat users are unprofitable.

The honest truth is AI is useless for most people. It’s not all that different from an iPhone or Android. Absent a company phone, most people just use it for games, videos, camera, social media, music, shopping, texting (if they don’t use a texting app like Signal or WhatsApp), web browsing, and email. Unless you’re trying to run 2K or 4K video or some game that maxes out your processor, few come close to utilizing the full capabilities of their $1,000+ mobile devices.

Forget even making phone calls if they’re younger. I had to tell my kids I will text or call as it pleases me, and absent it being at a truly bad time they WILL answer the call unless they feel like paying for their own phones and plans.