The Misleading Term "AI" Has Enabled Fraud (Buzzwords That Help Fool Investors and Sometimes Workers)
THE fact that Microsoft's "AI" stuff keeps getting rebranded and re-announced is telling. That they try to stick it everywhere, even in defiance of users' needs and wants, speaks volumes. That Microsoft proxies like DuckDuckGo help* Microsoft fake 'usage' (which is otherwise stagnant) is expected. Microsoft is getting very aggressive about this (to the point of quiet layoffs). Microsoft has begun worrying and scaring its own workers; they tell the media the company has tied its fate to OpenAI, a small company that never made money and probably never will (some expect it to go bankrupt soon; it is highly dependent on others losing a lot of money, just lending based on the vapourware and false claims).
We recently learned about this troubling thing ("Once again, AI is revealed to be three Mechanical Turks in a trenchcoat.") that did not exactly shock us. As Cory Doctorow said in December: "Every business plan has the word “AI” in it, even if the business itself has no AI in it."
"It turns out that Amazon's "AI" stores," a friend told me a few days ago, "where you just walk out and it figures out how much to charge you were not AI but mechanical turks, meaning in this case outright fraud."
As jwz put it (see "Update"): "It's easy to point and laugh at this as "LOL, AI doesn't work", and that's appropriate, because it's funny, and it doesn't, but also consider whether the goals of this project were not what you were told that they were. The "Just Walk Out" project allowed Amazon to replace a bunch of cashiers making the California minimum wage of $16 / hour with a bunch of call center gig workers in India making (I'm gonna go out on a limb and guess) a lot less than that. So if they are shutting this program down, it's probably because the savings they saw from outsourcing retail clerks to a third world call center were somehow not high enough."
"JWZ brought up a very interesting possibility about Amazon's fraud regarding their 'cashierless' stores," my friend said, because "the AI fraud might be a distraction and the real goal might be the fight against minimum wage and the push to outsource key jobs to third world countries with out social rights. I can add that there is the wish among some here [wish] to outsource K-12 teaching in that way, and that some steps have been taken to move that direction already."
This isn't about "AI" at all. The term is being misused to rationalise mass layoffs, wrongful assassinations, rogue surveillance, unjust censorship, poverty, and of course outsourcing/offshoring. Even when the term refers to "ML" (Machine Learning) their models aren't good enough and over time they will get even worse as they recycle and ingest their own spew, which lacks labeling and is sometimes indistinguishable from the originals (authentic, not synthetic). As this recent article noted, it is a crisis. Nowadays many chatbots are training on falsehoods created by their own 'spew', so over time they will vomit out more and more lies, drifting away from any sense of quality control. To paraphrase my friend, now the feedback loop I warned about looms as AI-generated "synthetic data" is being seriously considered as a substitute for real information when training newer AI models. Such BS behind all that, and more BS being produced by that.
This means that over time the text, the graphics and so on will get progressively worse, not better. This can doom a lot of what's on the Web and, if some offline things are created using material found on the Web, they too will deteriorate. Information is being blurred out of existence. Occasionally it may be replaced by convenient falsehoods.
It should be noted that a valued contributor of ours, Andy, once compared this to mad cow disease (and cows eating dead cows, bringing illnesses into broader circulation).
All the foolish, gullible shareholders/investors who pour out money into this are going to regret.
Many of us who understand the importance of quality (not quantity by mindless automation) will continue to enjoy sharp imagery, not derivative ripoffs that are likely plagiarism with presumption of "AInnocence". █
_____
* There is even a DuckDuckGo "Chat" EULA. "It says if your use of it gets them sued," Ryan tells us, "you are liable for the damages."