Slop Has no ROI, an Economy Built on False Assumptions of Slop is Doomed
Why trust this man?
A few hours ago we published thoughts from Akira Urushibata on "The Slop Ponzi Scheme" (our title, not his). In a nutshell, we're all going to suffer from this Ponzi scheme, even if we neither invest in nor use slop. If we have bank accounts, for instance, then our saved money is likely tied to the Ponzi scheme, which may also result in hyper-inflation, set aside rising energy prices, global warming, and water shortages.
This means that Scam Altman et al are a big problem for all of us. Ponzi schemes have systemic effects. And "investors appear to have already put in way more money than they can ever get back out."
So said an associate some moments ago.
"That is especially notable around the GPUs which are part of cards which only last 2 to 3 years when run full blast around the clock That means that in order to be profitable, the full investment has to happen long before that time is up so that there can be ROI [return on investment]. Yet speaking of ROI, the companies which have dumped money into LLMs and tried to integrate them into their work flows have not found any benefit."
There are some old Links (in Daily Links) covering each point above, e.g. lack of RIO. And that was one supplementary link, the associate added, "but the main point" is that slop will never be a viable business. Never.
"Yet apologists blame the victim," said the associate, "claiming they fail to see a ROI from LLM use due to 'bad strategy' rather than admitting that LLMs are useless." (The Register MS took bribes from Microsoft's Gartner to relay this victim-blaming lie).
The propaganda about this is astounding; imagine Tobacco Giants or Big Tobacco asserting that nicotine helps build up immunity from cancer, so if "smoking kills", then you're probably just not consuming enough cigarettes. Buy more! Consume!
"Also," the associate said, "the energy consumed per query makes the real costs for LLM companies much higher than they can find a market for. Mythos is so costly to operate the company simply cannot afford any more to allow free-of-charge access Those who waste money on a subscription find it underwhelming."
Mythoslop is not saving, it's costing. Many projects discovered (firsthand experience) that Mythoslop produces false positives which consume real developers' time. Does that save anybody's time?
More bug reports - like more code - aren't the answer. A firehose of false positive could be attained from fuzzing even decades ago, well before the LLM frenzy.
Speaking of quantity, this week Cory Doctorow repeats his older talking points about code as liability, not an asset. 2 days ago he said: "There's an analogy here to technology debt: technologically unsophisticated people think of software as a machine that never wears out and has no incremental usage costs (apart from electricity). In this framing, software is the perfect asset, one that never depreciates. But the reality is that software is a liability, not an asset" (good luck telling that to nontechnical managers). █