Using a Popular Ponzi Scheme to Sell Expensive Hardware Mislabelled as "Hey Hi" (AI)
Selling shovels in an age of "gold rush"
AS we pointed out earlier today in Daily Links, a "Former Enron Banker Points Out "Intriguing" Parallels With OpenAI". It says: "The ChatGPT maker was last valued at a whopping $157 billion after cinching another several billion in funding last October, making it one of the most valuable private startups, or unicorns, in the world. That's despite OpenAI never turning a profit, showing no clear road to profitability, and rapaciously burning through its mountains of money all the while."
Helped by corrupted and bribed media (usually paid by Microsoft), Scam Altman has been riding a hype wave. Who will stop this? Faking one's "value" like Microsoft does isn't that hard. The question is, how long for? The Microsoft mass layoffs - confirmed on the first week of the year - speak for themselves. No healthy company would announce layoffs so early in the year (many did not even return to work yet, they had been on holiday). As The Register put it, Microsoft "has made large-scale layoffs a pattern in the past several years". It's happening every month.
So far the main company to make lots of money (not raise "investment" money; that's not the same thing!) is NVIDIA. Many bought GPUs rebranded as "hey hi". Many businesses ordered these at high quantities (far higher than before) because of the hype, as they heard there's gold under the ground everywhere, all one needs is a shovel. Or a million shovels.
The market is overwhelmed now by an oversupply of GPUs. When the hype is over (bubble burst) people will find second-hand GPUs on the cheap; chips that may be branded "hey hi" but are otherwise quite ordinary.
The most ridiculous and resource-hungry/greedy "genAI" nonsense is videos (selling lots of hardware and burning the planet); those aren't even hard to detect and they're not worth watching. People have made videos like these for many years, predating by far the "AI" hype. The typical recipe is, prompt some LLM for dull and inaccurate text that "hallucinates" (perpetuates falsehoods), pass this text through a voice synthesis script, and then based on the words spewed out by the LLM fetch/synthesise some blurry frames sequence that roughly relates to the words. That's lame and hardly innovative, it just takes a lot of CPU cycles. It's the same nonsense with still images.
The bubble will explode. Not only is "open"AI the next Enron. Microsoft is too. The latter would be a much larger implosion. █