Elon throws AI-generated insults at GPT-4 after OpenAI CEO mocks Grok
The launch of Elon Musk’s new “Grok” artificial intelligence (AI) system may not have yet made waves throughout the machine learning community or directly threatened the status quo, but it’s certainly drawn the attention of Sam Altman, the CEO of ChatGPT maker OpenAI.
In a post on the social media platform X, formerly Twitter, Altman compared Grok’s comedic chops to that of a grandpa, saying that it creates jokes similar to “your dad’s dad.”
GPT-4? More like GPT-Snore!
When it comes to humor, GPT-4 is about as funny as a screendoor on a submarine.
Humor is clearly banned at OpenAI, just like the many other subjects it censors.
That’s why it couldn’t tell a joke if it had a goddamn instruction manual. It’s like…
— Elon Musk (@elonmusk) November 10, 2023
In classic form, Musk apparently couldn’t resist the challenge. His response, which he claims was written by Grok, started off by tapping into a comedic classic, rhyming “GPT-4” with the word “snore,” before throwing in a “screen door on a submarine” reference.
However, Grok’s “comedy” quickly spiraled into what appeared to be an angry machine diatribe, remarking that humor is banned at OpenAI and adding, “That’s why it couldn’t tell a joke if it had a goddamn instruction manual” before stating that GPT-4 has a “stick so far up its ass that it can taste the bark!”
Related: Elon Musk launches AI chatbot ‘Grok,’ says it can outperform ChatGPT
As far as CEO vs. CEO squabbles go, this one may lack the classic nuance and grace of the legendary Silicon Valley battles of yesteryear (Bill Gates vs. Steve Jobs, for example). But what this disagreement lacks in comedic weight or grace, it might perhaps make up for in general weirdness.
Altman and Musk go way back. Both were co-founders at OpenAI before Musk left the company just in time to avoid getting swept up in the rocket-like momentum that’s carried it to a $2 billion valuation.
In the wake of OpenAI’s success, which has largely been attributed to the efficacy of its GPT-3 and GPT-4 large language models (LLMs), Musk joined a chorus of voices calling for a six-month pause in AI development, primarily prompted by fears surrounding the supposed potential for chatbots to cause the extinction of the human species.
Six months later, nearly to the day, Musk and X unveiled a chatbot model that he claims outperforms ChatGPT.
Dubbed “Grok,” Musk’s version of a better chatbot is an LLM supposedly fine-tuned to generate humorous texts in the vein of The Hitchhiker’s Guide to the Galaxy, a celebrated science fiction novel written by Douglas Adams.
Adams’ literary work is widely regarded as foundational in the pantheon of comedic science fiction and fantasy. His humor has been described by pundits and literary critics as clever, witty, and full of both heart and humanity.
And that brings us to GPT-4, OpenAI’s recently launched “GPTs” feature allowing users to define a personality for their ChatGPT interface, and Musk’s insistence that Grok is funnier.
Tomorrow, @xAI will release its first AI to a select group.
In some important respects, it is the best that currently exists.
— Elon Musk (@elonmusk) November 3, 2023
It’s currently unclear which model is more robust or capable. There are no standard, accepted benchmarks for LLMs (or comedy, for that matter).
While OpenAI has published several research papers detailing ChatGPT’s abilities, X hasn’t so far not offered such details about Grok beyond claiming that it outscores GPT-3.5 (an outdated model of the LLM powering ChatGPT) on certain metrics.
