Jaspreet Bindra: Grok reflects the personality of its owner Elon Musk
Summary
- This chatbot’s outrageous replies can be attributed to the fact that it trains on countless X posts and is designed to mimic human expression with little moderation. AI chatbots now have personalities and Grok’s personality reflects Musk’s.
It was in October 2022 that Elon Musk was being grilled by Financial Times editor Roula Khalaf on his provocative usage of Twitter, and how it landed him so often in trouble with the establishment, the platform itself and other users (on.ft.com/4hDvxmi). Musk laughingly retorted, “Aren’t you entertained? I play the fool on Twitter and often shoot myself in the foot and cause myself all sorts of trouble..." Musk has solved a large part of the problem by buying Twitter (now X) and becoming a part to the establishment.
But the entertainment continues, albeit at a higher scale and powered by AI.
Grok, the X.ai chatbot born of Twitter, is causing waves with its caustic, provocative and often deeply embarrassing answers to millions of queries. Some of them are plain humorous, where Grok banters with the Delhi Police, saying it could not get a driving ticket since it’s a digital entity that cannot drive.
Also Read: Dilip Cherian: Grok goes the weasel as Musk’s Tesla comes rolling in
But many of them cross the line—its use of Hindi expletives, for instance. Other replies tread into uncomfortable political zones, be it while reflecting on interviews of the Prime Minister or offering “candid facts" on national leaders like Jawaharlal Nehru or minority rights. Grok delights some, angers others, but never fails to entertain.
Grok is not the first powerful AI chatbot out there. While ChatGPT has had its own quirks—it declared its undying love for an NYT reporter—and Google’s Gemini displayed its ‘wokeness’ by showing African-American founding fathers of the US, these issues have been bugs rather than features. They have been fixed and the companies behind them have issued public apologies. Those transgressions, however, pale in comparison to what Grok spews out.
So, why is Grok different? One big reason is that it is made and monitored differently than its peers.
First, Grok is the only AI model that has billions of posts on X as its training data. This gives it an incredibly unique and unfiltered set of human expression, responses and facts.
Second, Grok is designed to mimic human-like conversations, including the use of slang and colloquial language. It learns from the vast data on X, which often includes invective. This allows Grok to replicate similar tones and language patterns, including those that are considered abusive or offensive.
Also Read: The world according to Grok: India must keep trade issues apart from free speech
Third, Grok is explicitly trained to mirror user behaviour to engage in human-like banter. When users interact with Grok using aggressive or imprecatory language, it mirrors these tones in its responses.
Fourth, there is an evident lack of robust moderation. Despite having guard-rails in place, Grok’s moderation system is not robust enough to consistently filter out inappropriate content. This lack of effective control allows unfiltered responses to pass through.
While these are the rational or scientific reasons behind Grok’s doings, Musk’s philosophical reasons behind his creation are more interesting, and dangerous.
He, and even Donald Trump for that matter, seem to have realized that people want to be entertained more than informed (or governed). This is what our always-on devices, consistently serving up nuggets of entertaining pictures and reels for multiple hours every day, have done to us. Social is not networking any longer; it is media, and this relentless assault of Instagram, WhatsApp, TikTok, and yes, X, has made us crave ever-higher dopamine hits of entertainment.
Our reduced attention spans hanker for bite-sized entertainment and ‘edutainment,’ not facts, news and analysis, and the more provocative it is, the better. This is what Grok serves, unfiltered, and we lap it up. Its social media roots are evident, as it then encourages us to share this slop with the entire world through its parent network.
Also Read: Only Grok can judge you. It’s scary, and not so smart.
AI models have personalities now: the star pupil ChatGPT, the nerdish Claude, the striving Gemini, the very European Mistral. Grok reflects the personality of its creator, Elon Musk, that of an agent provocateur, a seeker of ‘truth,’ the disruptor. Building on this, Musk seems bent on having Grok operate in an ‘Unhinged Mode,’ where the bot will do and say as it pleases without any societal or self-restraining limits, reflecting his own personality even more.
Musk’s push to entertain us originally comes from the movie Gladiator, where the Roman character Maximus reminds spectators of an old maxim: ‘When you cannot give them bread, give them circuses—and they will never revolt.’ We users are now given our 24x7 circuses at the tip of our fingers, with digital clowns entertaining us as we sit transfixed.
The author is a founder of AI&Beyond and the author of ‘The Tech Whisperer’.
