Why Nothing Is a ‘ChatGPT Killer’
Why Nothing Is a ‘ChatGPT Killer’
No, not even the "best AI model in the world," AKA Google DeepMind's Gemini
Gemini Ultra won’t kill ChatGPT because surpassing GPT-4 on a carefully designed language understanding benchmark using a rather convoluted prompt setting doesn’t mean anything in the real world. People are calling it a “ChatGPT killer” anyway.
Others are giving this ludicrous statement enough relevance to deserve any criticism at all. Others, among them some of my Medium readers (not my fellow TAB readers — believe me, there’s a significant quality gap between both comment sections), are conflating my assertion that Gemini Ultra’s superior benchmark results suggest it’s now the best AI system in the world, with saying that “it will kill ChatGPT,” putting both statements on the same hype bag. (I’m aware of this. I’ll comment on it later.)
This article is many things. It’s an explanation for why Gemini is not (and won’t be) a ChatGPT killer. It’s a defense of my article whose headline “The Best AI Model in the World: Google DeepMind’s Gemini Has Surpassed GPT-4” generated a negative reaction after the Gemini demo turned out to be highly edited. It’s an argument for why nothing really is a ChatGPT killer and those kinds of “predictions” are never intended as predictions but as attention hooks. It’s a critique of people using those unhinged clout-chasing epithets to write response pieces, implicitly ascribing to them a treatment of worthiness they didn’t earn in the first place.
Above all else, it’s unsolicited advice for how to navigate the world of AI news, which is filled with hidden incentives that rarely align with what you — the audience — want. The Gemini story is a perfect example to illustrate how this happens and what to do about it. (Know that this is my last article on Gemini until there’s something else truly valuable to say about it. I however consider it important to share this story to explain myself and especially to provide an unfortunately scarce perspective on how AI news writing often works against your interests.)