should we fear Elon Musk’s generative AI which gets information from X (Twitter)?

Elon Musk has just unveiled Grok, his generative AI intended to compete with ChatGPT, which the billionaire considers too woke. The chatbot has the particularity of feeding exclusively on data from X. That’s promising!

After vigorously warning about the dangers of generative AI and trying to block its development, Elon Musk very quickly announced his intention to compete with ChatGPT and other generative AI with TruthGPT, a “artificial intelligence that tells the truth”. He subsequently founded a new company called X.AI and created a new team, called xAI, last July (see our article). On November 4, the billionaire finally presented its brand new chatbotrenamed Grok for the occasion — this name means “understand in depth and intuitively”, with reference to Stranger in a Strange Land (In a Foreign Land), a science fiction novel by Robert A. Heinlein. Similar to ChatGPT, Google Bard or Claude, the conversational robot aims to answer the questions of its interlocutors. Developed in — only! — four months, Elon Musk’s new baby will have the specificity of being the only AI connected in real time to data from X (formerly Twitter). Which leaves us fearing the worst for the future…

Grok: a sarcastic AI with real-time access to X

Elon Musk is not a novice in the field of AI since he had already co-founded OpenAI in 2015, before leaving the start-up in 2018, following a conflict of interest and a vision things are different – ​​he finds the company too “woke” and accuses it of making its AI lie with the restrictions imposed on the algorithm. Suffice to say that Grok will take a radically different path, for better and for worse! To answer its users’ questions, the AI ​​relies on real-time X data. For Elon Musk, it’s about“a major advantage over other models”, which are often limited to a restricted database – the free version of ChatGPT does not have access to information appearing on the Internet after September 2021 for example. To illustrate his point, Elon Musk asked Grok what outfit Joe Rogan wore during their recent interview. The chatbot was able to respond precisely, while ChatGPT was content with a general response on the American podcaster’s clothing tastes. According to xAI, the chatbot is also capable of solving mathematical problems, or even carrying out certain tasks in the field of coding. We can obviously question the impact of a database exclusively from X, given the titanic number of fake news and false information that abound on the platform.

Another point on which Grok stands out and which may pose a problem: his “humor” and his “open-mindedness”. Adept at sarcasm, the AI ​​is designed to respond to controversial and sensitive questions, where Bard and ChatGPT take precautions and refuse to take a position. On the account of the owner of Likewise, when the user asks him how to make cocaine, Grok advises him to get a chemist’s degree before “set up a clandestine laboratory” and hope not to get arrested. He concludes by recalling that “manufacturing cocaine is illegal and dangerous”. If the answer indeed makes you smile, we can nevertheless wonder how, in just four months, researchers were able to put effective filters on generative AI – we all remember the impressive road trips of ChatGPT, Bing Chat and of Bard in their beginnings, while the companies had nevertheless taken their precautions. This is all the more the case when we see the questionable information that the social network contains and when we know that Elon Musk uses the latter as a sounding board to share his own political vision. We fear the worst if Grok starts to give his opinion on the situation with Israel and Hamas, on the candidates in the Republican primary or even on the LGBT movements…

Note that xAI has great ambition regarding AI, since it indicates that, “Currently, Grok does not have other senses, such as vision and hearing. To better help users, we will equip Grok with these different senses which can enable broader applications, including interactions”. This explains the social network’s change in policy regarding the collection of personal data, the latter now indicating that it can collect users’ biometric data with their consent (see our article). At the moment, the chatbot is still in the alpha development phase. As soon as it enters beta, it will be accessible to X Premium+ subscribers, the platform’s paid offer of 16 euros per month. A way to encourage Internet users to take out their wallets… Note that the integration of Grok into X is part of Elon Musk’s project to transform the social network into a super application (see our article).