A couple of weeks ago, Gary Marcus’s newsletter flagged a company (Inqwire) that had a statement on their frontpage, stating that they do not use LLMs*, and adding that they do not pretend to be using humans when they use chatbots.
Inqwire’s positioning is the complete opposite of pseudo-AI, in which companies sell certain services (e.g., 24 hour customer support) as being performed by AI but, then, use humans to deliver them. Sol I have been wondering: How did we go from companies paying humans to pretend that they were AI to, now, making a point of announcing, on the front page, that AI is not used (or, at least, not this specific type of AI)?
Sure, there are the performance issues, and the concerns over protecting valuable datasets. But, in Inqwire’s case, I think that there is also a positioning issue: the company may be saying that customers are too important to be handled by generative AI.
Traditionally, AI was seen as something largely inaccessible due to its cost and complexity. Firms were being encouraged to use it to perform very complicated calculations, spot patterns that are not easily discernible to the human eye or avoid human error. Generative AI, however, is in everyone’s hands. LLMs, in particular, are a form of AI that is very cheap to use, and the user-interface makes them very accessible, too. And, as a consequence, everybody and their dog is using LLMs to save time and cut corners. And that’s a very different proposition.
In my view, the accessibility and subsequent ubiquitous use of LLMs turned them into what economists call an inferior good. While for normal goods (e.g., fresh vegetables) we might expect people to buy more as their income increases, for inferior ones (e.g., tinned vegetables) demand falls when income increases. Tinned vegetables are very nutritious. But they are not aspirational. They have an image problem.
For generative AI companies to recover the huge costs of developing and running these models, they don’t just need better models. They also need to find a way to make generative AI users (namely, companies like Inqwire) look sophisticated. And cool. Like kale. So cool, that even Beyonce wears it on a sweatshirt.
As a consumer, what is the message that you get when a company says that they do not use AI? And what about when they say that they do not use LLMs, specifically?
*LLMs = Large-language models, one form of generative AI.
Ana Canhoto 