AI’s Going to Cost Us
Nothing from Silicon Valley is ever free
There are a lot of very smart people working on artificial intelligence right now, but I have a hard time taking them seriously for one petty reason.
It’s that they say “compute” instead of “computing power” or “computing capacity.” I guess that because they are busy, brilliant people with models to train, venture capital to collect, and hype to manufacture, they don’t want to say five syllables when two will do.
So these Silicon Valley guys run around all the time talking about “compute.” It’s one of their most significant concerns, because every time a middle schooler asks a large language model to write a five-paragraph essay on Lord of the Flies, a lazy middle manager prompts an AI to compose an email to her staff, or a lonely guy flirts with his new chatbot girlfriend, they’re using a lot of (ugh) compute.
Each of these requests gets sent to a server farm full of very expensive processors, which use a lot of power and have to be constantly cooled so they don’t overheat. The servers churn through the data on which the model has been “trained,” predicting a likely response to the prompt.
The model may not charge you for that uninspired lasagna recipe or image of a seven-fingered Joe Biden wrestling an alligator, but that doesn’t mean…
