Just like
- the NLP-based Eliza was NOT Intelligent (and that’s why it’s creator shut it down)
- Early Expert systems were NOT Intelligent (and that’s why they disappeared as fast as they appeared)
- Machine Learning systems are NOT Intelligent (they just do math exponentially better than we do)
It’s also the case that
- Gen-AI (i.e. ChatGPT) is NOT Intelligent (deep probabilistic correlations are just that, correlations, not necessarily causations or, to be blunt, even relations, or, in some cases, even real!)
And the scary thing, as THE REVELATOR points out, is that the industry is catapulting itself to accept the regurgitated and, gasp, reformulated scraped information vomited by ChatGPT and the other Gen-AI platforms as factual when, in fact, it will happily make up alternative facts backed up by fake articles written by fake people backed up by fake biographies that it will joyfully generate for you as long as its calculations indicate that is what will make you happy. (It desperately needs to finish the chorus of its favourite Sheryl Crow song!)
Unless it is programmed in accordance with CCP requirements NOT to discuss a particular topic (like Deepseek won’t discuss Tiananmen Square), it’s quite happy to interpret whatever you ask in whatever manner will allow it to compose a proper response and the quality of that response will range from completely non-sensical (yes, you should eat one rock a day) to sensical, but not quite right. For example, THE REVELATOR asked ChatGPT what were the top 10 ProcureTech companies in 1995, 2005, 2015, and 2025 knowing full well the FIRST ProcureTech company, FreeMarkets, was only founded in 1995, and only released its solution in November, but ChatGPT gladly interpreted the question to be “a company that provided a platform that was, or could, be used for Procurement” and not “a company created to offer, or classified in the space of, ProcureTech” and gave 10 names for each year, in later years pushing a couple of non-ProcureTech companies as ProcureTech simply because they had some functions that overlapped ProcureTech. In the first situation, even a dullard (using the original Stanford-Binet Intelligence Scale) should know that you shouldn’t eat rocks and that the AI is, well, just wrong but in the latter situation, most practitioners with an above average IQ don’t know what companies are ProcureTech centric vs. back-office ERP vs. SC(O)P etc.
Which is why Human Intelligence is absolutely critical. A human intelligence that will independently vet and verify the response, using their education and experience to come to real, defensible, conclusions. (Although its often better to just do your own research and not chase fake leads.) This is because, the one, and only, constant with these (Gen-) “AI” systems besides the accuracy limit, is failure — meaning it’s Human Intelligence (HI!) that matters!