“The Model”

In watching the video announcing GPT-4o, I was struck most not by any particular demonstration or new feature, but by a single word: model. Everyone at OpenAI, led by CTO Mira Murati, is spot on in referring to their creation in noun form merely as, “the model.” Surely, this is intentional.

For the previously generation of conversational AI, which mostly sucked, companies went out of their way to personify their AI chatbots with fun or human-like names, like Jarvis. I suppose the intent there was to try to make something that wasn’t very human-like at all seem more human-like. OpenAI’s challenge is essentially the opposite.

Each update to GPT makes it almost uncomfortably more human-like. Referring to it merely as, “the model,” dehumanizes it presumably for comfort and to allay AI fears. I’m sure leadership gave serious consideration to exactly what wording to use. “The model” is about the most anodyne term they could without veering into a degree of vagueness so great as to be mundane, like saying “our technology.” What they have built is indeed a model. But it’s also clearly so much more. It’s not unusual for us to resort to euphemisms when the magnitude of something is more than we can handle.

For tech startups we have not layoffs but “RIFs,” or reductions in force. Our government was not torturing detainees in recent decades, but rather using “enhanced interrogation techniques.” Sometimes when we can’t handle the truth we simply call it something else that sounds, well, softer. That’s my feeling of what “the model” is. Alternatively I might call GPT-4o, “Her.”