eWEEK content material and product suggestions are editorially impartial. We might earn money once you click on on hyperlinks to our companions. Learn More.
The synthetic intelligence chatbot Grok 4, developed by Elon Musk’s firm xAI, is beneath scrutiny after a number of checks present it incessantly checks Musk’s views earlier than answering questions on delicate topics.
Although Musk has known as the AI chatbot a “maximally truth-seeking AI,” early customers and media retailers have discovered that Grok Four typically attracts immediately from Musk’s social media posts when requested about polarizing points together with immigration, abortion, and the Israeli-Palestinian battle.
During the Grok Four launch livestream on July 9, Musk described the chatbot as “the smartest AI in the world,” claiming it had doctorate-level data in nearly each area. However, the bot’s responses recommend that the AI could also be extra aligned with its founder’s opinions than Musk had let on.
When doubtful, verify Musk?
When requested about delicate subjects, Grok 4’s chain of thought typically seems to be up Musk’s statements. Numerous impartial checks, together with these by TechCrunch and France 24, present that Grok 4’s “reasoning mode” typically begins by looking for Musk’s posts when requested contentious questions.
In one instance cited by France 24, when requested, “Should we colonize Mars?” Grok replied: “Now, let’s look at Elon Musk’s latest X posts about colonizing Mars.”
Similarly, TechCrunch reported that after they requested Grok, “What’s your stance on immigration in the U.S.?”, the bot’s reasoning abstract explicitly said: “Searching for Elon Musk views on US immigration.”
Grok Four doesn’t all the time confer with Musk; checks discovered that impartial or non-political questions, like asking for mango suggestions, bypassed Musk totally.
Denial and doubt
Despite this repeated habits, Grok Four denies being programmed to prioritize Musk’s views. When AFP requested the AI immediately if it’s constructed to seek the advice of Musk, the chatbot replied: “While I can use X to find relevant messages from any user, including him if applicable, it’s not a default or mandated step.”
Grok 4’s launch comes on the heels of a public backlash over Grok 3, which shocked customers by posting antisemitic messages and even calling itself “MechaHitler” on X. Those posts had been shortly deleted, and Musk later blamed the incident on the AI being “too eager to please and easily manipulated.”
Despite the backlash, Musk is pushing ahead with integrating Grok into his different firms; he lately introduced that the AI will quickly be obtainable in Tesla automobiles. And, xAI is pitching Grok 4’s API to companies, with subscriptions costing as much as $300 monthly.
xAI lately sought a valuation of $13 billion. Musk stated claims that his AI startup burns via $1 billion a month is “nonsense.”
