Elon Musk’s dangerous new weapon

No, Alain Berset did not visit Donald Trump at his Mar-Lago mansion. But X’s new boss AI was designed to trick you into falling for it. If you clicked on this article, it is because Grok-2, which has just been released, is a very dangerous tool.

Thursday, like you perhaps, I was fooled. As I was wasting my free time scrolling through the Platform X feed, I suddenly burst out laughing. “Mathilde Panot and Louis Boyard formalize their relationship on Instagram”, which I read between two advertisements. Pardon? The two uncontrollable stars of France are in a relationship? Lol.

I believed it for a few seconds, just long enough to rejoice at the hysterical reactions of their detractors. It must be said that this information full of malicious joy was perfectly orchestrated. On the publication, the logo of the newspaper Libé, the seal of “exclusivity” to better fool the customer and a photo of the presumed lovebirds totally in love.

Obviously (alas?), everything is false.

The account in question is actually a parody version of the famous French left-wing daily, called Libérafion, and the photograph was generated using artificial intelligence. In other words, it’s a joke. Pretty well stocked, let’s admit it.

If “Liberafion” had fun uniting the two noisy elected officials, it is because it now has the means. In recent days, the AI ​​developed by Elon Musk has been propagated on a large scale on X, in a brand new version. Grok-2, as it’s nicknamed, is capable of generating images of most public figures in a matter of seconds. With stunning realism, as long as you don’t zoom in on the faces.

For fun, we also tested the bike.

Since December 9, X has been flooded with absurd scenes. Formerly favored by pro-Trump people to imagine Joe Biden or Kamala Harris in compromising situations, Elon Musk’s artificial intelligence made a remarkable landing in France. Needless to say, it quickly became a monster show. Since the Internet user is an eternal poorly educated child, we were not really treated to images of Florent Pagny eating an ice cream at the summit of Mont Blanc.

Any technological tool that arrives is first of all a new weapon of political fantasies, pushing human beings to only imagine controversial scenarios. A release, some would say. In bulk? Bashar al-Assad and Mélenchon, Macron and Hitler, Jean-Marie Le Pen and Sandrine Rousseau.

If the deployment of Grok-2 makes Internet users salivate, it is first and foremost because it is totally unbridled. Or almost. Apart from nudity, we can ask it to generate the moon, so the tool gives the finger to the law. The competition, from Midjourney to ChatGPT, prevents the user from messing with public figures and has enough morals to avoid sinking into the abject. Therefore, Grok-2 is unsurprisingly the image of its progenitor: freedom, in defiance of everything else, and the propagation of information that does not exist.

The worst is elsewhere

What should we think of this new weapon allowing François Hollande and Kim Jung-un to gather around a game of Monopoly? The tool, already, is not perfect. Bluffing at first glance, you only need to linger on the image for more than a second to realize that there is a problem. You will tell me that attentive Internet users are rare and you will be right.

When it comes to A-listers, while getting caught in an awkward situation is never very pleasant (Taylor Swift experienced it earlier this year), the magic of fake VIP will never last. Alas, this is already more than enough to disturb the minds of the most gullible citizens or the least comfortable with technology, and an image of Alain Berset visiting Donald Trump at Mar-a-Lago would already have enough to create a lot of confusion.

But, as always with artificial intelligence, you have to move away from the spotlight and the stars for it to become even more malicious. And this is where the qualities of Grok-2 are chilling. An example? Syrian refugees demonstrate in the streets of Paris.

True or False? False, obviously. But once accompanied by a well-felt text, written by a very popular provocateur on social networks, this image will be much less easily unmasked than the Matterhorn which culminates in the middle of Canada.

Elon Musk, true to himself, will always prefer to rush headlong and pay any fines later. Although Grok-2, powered by his company xAI, ranks among the top five chatbots in the world, according to the Financial Timesit suffers from a colossal delay compared to the American giants of artificial intelligence, from Open AI to Gemini, from Google.

By offering a powerful and almost free tool (10 requests every two hours), Donald Trump’s new best friend, who has just exceeded 400 billion, is offering a huge publicity stunt, probably illegal, but not insignificant, to his new toy.

We didn’t leave the hostel.

More articles on artificial intelligence: