INDEPENDENT
Less than eight months after discouraging companies from developing advanced artificial intelligence, Elon Musk has unveiled his answer to “woke” AI chatbots like ChatGPT.
The tech billionaire claims his new Grok AI is both smarter and funnier than its rivals, offering paid users of X (formerly Twitter) the chance to ask it “spicy questions that are rejected by most other AI systems”.
The X boss offered an example of how it will answer “almost anything”, sharing a screenshot of a user asking it how to make cocaine.
“Grok is designed to answer questions with a bit of wit and has a rebellious streak,” a blog post announcing its launch noted. “Please don’t use it if you hate humour!”
What differentiates it from OpenAI’s ChatGPT and Google’s Bard is that it has access to real-time data from X, which Mr Musk took over almost exactly a year ago. Before the takeover, AI firms were using Twitter as a data set to train its models, however the tech billionaire shut this down following the release of ChatGPT last November.
Initially labelled “TruthGPT”, Grok takes its name from Robert A. Heinlein’s novel Stranger in a Strange Land, meaning understanding something thoroughly and intuitively, while the tone of its responses are modelled on the same writing style as Douglas Adams’s Hitchhiker’s Guide to the Galaxy.
But with X as its training set, Grok risks mimicking the same misinformation and toxic discourse that has plagued the platform since before Musk’s takeover.
Not only does it adopt a more informal tone to its rivals, Grok also appears to have less safety filters preventing it from answering questions about sensitive topics.
Despite claiming that Grok outperforms ChatGPT, which is freely available, xAI did acknowledge that it does not yet match the capabilities of OpenAI’s more powerful GPT-4 model – which carries a similar monthly fee to Grok. In its own in-house tests, xAI graded Grok against GPT-4 on the 2023 Hungarian national high school finals in mathematics. Grok passed the exam with a mark of 59 per cent, while GPT-4 scored 68 per cent.
In March this year, Mr Musk was among hundreds of leading tech figures to add their name to an open letter calling on all AI labs to pause the training of AI systems. The letter warned that artificial intelligence with “human-competitive intelligence” could pose “profound risks to society and humanity”, potentially leading to the loss of control of human civilisation and even its extinction.
This letter of discouragement looks increasingly like a plea to allow his own companies – which include the newly formed xAI – to catch up.
