Killer robots writing poetry. This week OpenAI began sharing access to GPT-3, its new language-generating AI, with a select handful of developers . Poetry, stories, essays, and bits of work generated by the program leaked to Twitter, and the general consensus was “holy shit, our replacements have arrived.” Core to the human being is our self-conception as fundamentally creative, which we contrast with every other lesser intelligence on the planet. Now here’s a robot in that game with us. Or… is it? Sam Altman, one of the cofounders of OpenAI, took a stab at explaining the awed reaction to GPT-3 here:

I agree the natural language link is probably key to the reaction, but the programming aspect seems a somewhat separate issue. Most people haven’t even used GPT-3, we’re all just judging the AI by its creative work. This got me thinking: is everything generative really creative? The relationship between language and sentience is something we all intuitively feel. What is desire — the state of wanting things — without expression of desire? Language is in some very real sense the window into who and what we are. Expression without desire feels empty because it is, and I did feel a kind of emptiness in GPT-3’s work. But I think this has less to do with the program than it does with people. Separate from the incredible technological achievement in GPT-3, I was reminded of two things: 1) we are still very far from general AI, which Sam and others freely admit (a stark departure from the enthusiastic era of The Singularity is Near, which these days feels further off than ever), and 2) most people aren’t especially creative.