Admittedly, this is very similar to what I wrote yesterday, but I contend that is because I really want to highlight (and underline) the degree to which naïve gullibility has grown to the point that it’s a social problem now. Whereas maybe ~20-30 years ago, a person could generally believe a company’s statements (maybe not a food company), hyperbole and straight-up lies have become so common that they’re considered normal.
I used to live in New York. Not long after I moved there, I went on the tenement tour (which was interesting cuz I lived in a tenement building so I had nearly the same floor plan). On the tenement tour, the guide said that in the late 19th century, street musicians commonly sang songs about things like dubious products used in food — like songs on the street had a role in consumer protections during a time when governments did not. Nowadays, ad hoc consumer safeguards are making sense again in the US, because the government appears to have no interest in concerning itself with the wellbeing of the citizenry.
Once again, I was just watching videos about the “AI” bubble, and I heard numerous statements from the CEOs of OpenAI and Anthropic, and they seem to attempting to deliberately deceive people. Set aside the extremely deceptive practices related to their accounting and valuation. In addition to that, they don’t honestly describe what their programs do and how (in the broadest terms). Like the milk company adding like talcum powder to milk to make it appear whiter, with no regard for folks drinking it, these LLM companies are behaving similarly as they describe a presumably better future by describing what “AI” might do someday rather than what it can do currently. And part of the reason is probably that the valuations of the big “AI” companies are so extreme that only extreme utility could possibly warrant such extreme valuations.
The frontier LLM companies are absolutely on the cutting edge of deceiving huge numbers of people so as to get their money. It’s really quite pathetic. Innovation and deception are separate things, and blending them does no good. Personally, I think the big “AI” companies are using a drug dealer business model: it’s free until you’re addicted. That is gross. What’s grosser is the way in which pretty much the entire economy is wrapped up in the “AI” companies’ quest for profits — talk about socializing the losses!
Once again, the fundamental problem is — perhaps ironically — a lack of skepticism. I, myself, have used multiple LLMs a lot. Because of how they work, they aren’t good at things that aren’t explicit. Like, if you say/type something to an LLM, it is received as literal. Whereas a person might interpret what you said as sarcastic, an LLM simply has no social skills because it is fundamentally a computer program. An ability to detect things like sarcasm can be programmed in certain narrow cases, but a human has a fundamental awareness of being human, which has aspects that are not nearly as rigid as the binary that is the basis of computers. Sure, software vendors can use LLMs that use percentages and probabilities to predict words, but that is not how humans work. Humans are capable of remaining skeptical even when other humans are attempting to mislead them. I highly recommend remaining skeptical, particularly during this era, when deception is seen as an intrinsic part of marketing.