Bonum Certa Men Certa

OpenAI and ChatGPT Could Very Well Collapse and Shut Down Later This Year (Huge Losses, Sagging Usage Levels, and Massive Debt)

posted by Roy Schestowitz on Jul 13, 2024

Empathy

OpenAI, which became (in)famous mostly for ChatGPT in its capacity as a Microsoft proxy, is perishing. Below we illuminate the suppressed observations that Microsoft-sponsored publishers and cheaply-made slop (LLM spew disguised as "news") try to distract from.

Background readings

June 2024: The Bubble Has Burst. OpenAI.com Traffic Plunges 60% in One Month, Tracker Shows.

September 2023: OpenAI Bankruptcy by Year's End? Earlier Today Similarweb Released August Figures, Showing That Microsoft's Chatbot Hype Was in Vain and Usage Continues Falling for Several Consecutive Months.

July 2023: Turns Out Microsoft Chatbots Are, Indeed, a Passing Fad (Userbase Down Sharply)

July 2023: Microsoft's Bing Falls to Lowest Share in 2 Years Despite Chatbot Hype and Amid Bing Layoffs, Executives' Departure

Reality strikes

A cofounder and many senior staff already leave. "Company insiders explain why safety-conscious employees are leaving," a recent report said and there are also severe financial issues. An associate reminds us: "There was a recent article in [Daily] Links about how each ChatGPT search burns as much electricity as running a lightbulb for 20 minutes (for some value of 'lightbulb')."

We'll revisit this later.

Media bites

Even less than a year ago (at "peak hype") the media said "OpenAI may have lost more than half a billion dollars over the past year", "ChatGPT costs $700,000 to run daily, OpenAI may go bankrupt in 2024", and "OpenAI could go bankrupt by 2024 as ChatGPT costs over Rs 5 crore to operate everyday" (it's probably more expensive now, for a variety of reasons including legal costs, staff costs, and scale).

Usage falls

In May the access to the "openai" domain had fallen by about 60% and last month it fell again by nearly 30%. Microsoft tried to fake "demand", seeing that usage levels of the chatbot had plunged, by interjecting/plugging it into everywhere in Windows, 'Webapps', and so on.

Does this seem like a real strategy or rearranging the seats on the deck of the Titanic in panic?

Nobody even asked for this. Many are so pissed off by it that they delete Windows completely and move to GNU/Linux.

Google not doomed

At the end of 2022 and start of 2023 Microsoft-sponsored media kept telling us all that Google was doomed due to chatbots. Nearly 2 years have passed and the market share of Bing is estimated to have actually fallen in Australia (since the LLM hype began). In some countries it is only up slightly, but in about half the countries it is either the same or lower. In fact, the same is true in Europe (Bing down), where Yandex is nearly more popular than Microsoft... it nearly caught up by now.

There's no sober person who nowadays insinuates ChatGPT will "kill" Google. It can kill the Web by filling it with slop (trash), but that just kills search in general, not Google. It pollutes the world with plagiarism, which is a moral problem and a legal problem (more lawsuits and regulatory actions).

It has gotten worse

Instead of improving over time, some people observe that the opposite is true. Having exhausted what they could train the LLMs on (e.g. Wikipedia), now they're chewing up their own slop and it results in worse performance. As someone said earlier this year in openai.com:

Then January came and it became ‘lazy’. It would just summarise in a short paragraph without much details. I had to give several prompts over and over again to get it to give me the answer in detail as it used to. However, all is well as my GPTs were doing okay.

Now my GPTs are NOT working too. It doesn’t listen to its custom instructions and my answer is short and vague. When I asked it to re-do, it f**king replies, “Unfortunately, I can’t fulfill this request.”

Even old prompts that used to work are no longer working. It either give short or vague responses, or it repeats whatever was generated beforehand.

Not just me, my friends who are using ChatGPT have said the same. ChatGPT has deteriorated. It is as if OpenAI decides that from 2024 onwards they will reduce token usage and make the answer more summarised, vague and ‘stubborn’. My friends are even speculating whether openAI is doing this on purpose to force us to use up our prompts until it hits the cap limit quickly.

I have never felt so frustrated with ChatGPT before. This is just horrible. At this point I feel it is even pointless to have a chatGPT Plus subscription.

It is a threat to human knowledge, but eventually it'll probably perish because not only training but also querying LLMs is far too expensive and people lose interest. openai.com will probably be offline within years or a decade. There's no business model there, just hype.

Other Recent Techrights' Posts

Censorship Versus Fact-Checking and Quality Control
It's not censorship but a matter of quality control
Gemini Links 07/07/2025: BaseLibre Numerical System and TUI Rant
Links for the day
Two Risks to Companies: The Microsoft Culture and the Microsoft Tools
Novell was killed by a form of "social engineering" by Microsoft
It's Hard to Trust People Who Worked - Not Only Those Who Still Work - at Microsoft
Bryan Lunduke is just what people would call an "arsehole of a person"
 
Large Language Models (LLMs) Externalise Their Cost to the Free Software Foundation (FSF)
"The forty-sixth Free Software Bulletin is now available online!"
Weeding Out Extremism in Our Community
To me it seems like Microsoft Lunduke is rapidly becoming like a "hate preacher" who operates online, breeding an extremist ideology or trying to soften its image
Reinforcing the Allegations Some More, Bryan Lunduke Digs His Own Grave
In his latest episodes he merely repeats his own lies, which I debunked using evidence right from his own mouth
Global Warming and Free Software as a Force of Mitigation
we'll need to think about Software Freedom, not just brands like "Linux"
Over at Tux Machines...
GNU/Linux news for the past day
IRC Proceedings: Sunday, July 06, 2025
IRC logs for Sunday, July 06, 2025
[Video] "Copyleft Isn't a Bug."
"Copyleft isn’t a bug. It’s a feature. GNU GPL forced the world to treat code like a public good."
Being in Social Control Media Means Exposing Oneself to Heckling
Richard Stallman does not (either himself or directly) post to any social control media
Links 06/07/2025: Airlines Perils, Scams, and Breaches
Links for the day
For the Second Time, Bryan Lunduke From Microsoft is Siccing Racist Trolls and Vandals at Me
You're only reinforcing the point we made yesterday
Links 06/07/2025: End to End Encryption at Risk, Reuters Twitter ("X") Account Withheld in India
Links for the day
Gemini Links 06/07/2025: Tinylog and Certification Rotation
Links for the day
Links 06/07/2025: Climate Change and "The Right to Criticise"
Links for the day
PCLinuxOS Sites Coming Back, Gradually
let's just be patient
Social Control Media, Even If Based on Free Software, Still Has Many Problems
a distraction from what actually mattered and still matters
IBM is Not Your Master
IBM makes friends with people who exclude the majority of the population: women
Help Fund the Free Software Foundation (FSF)
If you have some dollars to spare, go support the FSF
Over at Tux Machines...
GNU/Linux news for the past day
IRC Proceedings: Saturday, July 05, 2025
IRC logs for Saturday, July 05, 2025
A Short History of Attacks on Techrights (and Boycott Novell Before That)
good opportunity to tell again the story of several (not all) attempts to silence us
The Mainstream Media Took 4 Days to Realise Microsoft Shut Down Its Operations in Pakistan and Fired Everybody
We estimate that Microsoft has had about 29,000 layoffs since January
Leadership in Free Software
Don't let IBM lead. It's a terrible flag bearer.
“Twibel” Actions Against Comedians (and Why It's a Truly Low Blow)
they try to make up in quantities for a lack of merit or quality
Linux Foundation Apparently Flirting With Slop (Marketing by LLM-Generated SPAM)
The Web is in a really bad state!
COVID-19 Sped Up Site Improvements in Techrights
A few months later we created our very own IRC network
Gemini Links 05/07/2025: Negative Questions and 'Touching Grass' (Going Outside)
Links for the day
Links 05/07/2025: Dalai Lama Succession as 90th Birthday Approaches, 40 deg C in China
Links for the day
Links 05/07/2025: Hungary and US Defecting to Russia, "Google's Hotseat Hypocrisy"
Links for the day
Gemini Links 05/07/2025: 4th of July 2025 and "Zig Roadmap 2026"
Links for the day
How to Combat the Exploitation and Abuse by Microsoft GitHub
Not to mention corruption and crimes against women
Bryan Lunduke is Actually Sending His Audience to Attack People
"[Lunduke] is actually sending his audience to attack people."
Even The Right Wing is Rejecting Bryan Lunduke
no wonder he became so irrelevant and marginal
Microsoft's MSN Helps Microsoft Spread Lies About the Layoffs' Scale (Well Over 25,000 People Laid Off This Year)
There seem to be monopolies on lies and on truth
The Death of X Has Been Greatly Exaggerated (by Compromised Media)
X.Org Server is alive and well
Rewriting Things in Rust
How far would you go?
In 2025 Everything is "AI". Remember Blockchains?
Talk about what companies and things (services, products, software) actually do, not the labels they use
Julian Assange Has Been Free for a Year
Julian Assange and I disagreed on some things
Monopolies and Scalping
Monopolies gravitate towards price hikes
Over at Tux Machines...
GNU/Linux news for the past day
IRC Proceedings: Friday, July 04, 2025
IRC logs for Friday, July 04, 2025
Microsoft's August Layoffs Wave: "August is Confirmed for Additional Performance Based Cuts"
"August is confirmed for additional performance based cuts from the recent connects along with additional organizational cuts."
What Microsoft Reputation Laundering (With a Weaponised Law Degree) Looks Like in a Foreign Continent
You would expect this in uncivilised and primitive countries
Slopwatch: LLMs 'Write' Fake or Distorted 'News' About "Linux"
LLM slop disguised as news