When the boss is always right, the AI will be wrong

Is Elon Musk more brilliant than Newton? Grok’s reply highlights the dangers of corporate sycophancy.

- Gautam Mukunda

Mr Elon Musk has said that Grok, xAI’s chatbot, is “the smartest AI in the world”. And can you blame him? After all, Grok recently declared that his “intelligence ranks among the top 10 minds in history, rivalling polymaths like da Vinci or Newton through transformative innovations in multiple fields”.

What’s more, the bot explained that its boss “edges out” basketball icon LeBron James “in holistic fitness”, would win in a fight against multi-time world heavyweight champion Mike Tyson and is the “world’s greatest lover”.

Mr Musk claims that these responses were the result of “adversarial prompting”, a way of tricking artificial intelligence (AI) systems into giving answers that they were not designed to provide. And since the initial barrage of florid praise, Grok has toned down its responses.

When asked similar questions today, it still throws in gratuitous praise of Mr Musk, but no longer calls him a better athlete than James and claims that some of its prior statements were tongue-in-cheek.

Beyond the objective hilarity of the bot’s latest suck-up fest, the incident is a pointed reminder that AI is, as I’ve noted before, an enhancer rather than a corrective to managerial pathologies.