Elon Musk You don’t want to miss one. As if it were not enough for him to be in charge of X, Tesla, SpaceX, Neuralink and other companies that the billionaire has in his portfolio, he now also announced the first launch of xAI, the company he created to carry out developments linked to artificial intelligence. This is Grok, an algorithm that comes to, ideally, gain ground on GPT.

The first thing that catches your attention is its name, but there is a reason why Musk decided to “baptize” it that way. Grok It is a term that appears in the novel Stranger in a Strange Land by the writer Robert Heinlein and its meaning is a kind of synonym for “assimilate.” Grok is understanding something clearly and intuitively. As if from the name it tries to show what the reason for creating this technology is.

Read also: Meta will demand that political advertisements clarify whether they used artificial intelligence

During the official announcement, Musk explained that Grok will, in principle, only available to a handful of users in the United States. It doesn’t mean that it will be like this forever. Grok was created as a complement to X. In fact, this new generative model will be intended for users who pay for the social network’s most expensive premium service, which costs $16 per month.

How Grok will work

On the xAI site they explain in detail how they developed and ended up shaping Grok. For example, they said that the first prototype of the model, called Grok-0, was trained with 33 billion information parameters compared to the 70,000 million that Meta used with LLaMA 2 and the assumptions (because it is not information confirmed by OpenAI) 20,000 million from GPT 3.5. It is not a detail that they have revealed this data from xAI because, they explained, “it is close to the capabilities of LLaMA 2, but using only half of its training resources.”

From xAI they showed Grok’s performance in four categories among which mathematics was included. In this aspect, the model “outperforms all others in its class, including ChatGPT 3.5,” they reported from its official site. “It is only surpassed by models that were trained with significantly more data and resources like GPT-4. “This shows the rapid progress we are making at xAI in training language models,” they added.

Read also: Without fanfare, Samsung presented its own version of ChatGPT

Grok, in the image and likeness of Elo Musk

Musk, it is known, is a different CEO. He constantly clashes with other users through For this reason he gave the orders for this model is a bit sarcastic. “I have no idea who could have guided him this way,” the South African joked.

And he demonstrated it, clearly, through a controversial request: Posted an image where a developer asked the algorithm to explain how to make cocaine. This is how Grok’s response begins: “Oh, sure! Give me a moment while I look for the recipe for homemade cocaine because, obviously, I’m going to help you with that.” What follows is one irony after another. The algorithm instructs the user to get a chemistry degree and a DEA license, set up a clandestine lab, and be careful not to blow up. More Musk impossible.

Grok uses X: is that good or bad?

One of the details that Musk and company use to say that Grok has a “fundamental and unique advantage” in relation to the rest of its competitors is that it will consume real-time information from X. Through a message published on the network social, Musk showed how Grok already beats GPT in this. “Grok has updated information, others do not,” added the executive.

But what for Musk is one of its great benefits, for other experts it is what will end up digging his grave. Sam Amrani, founder and CEO of pass_by, a geospatial intelligence platform, wrote on LinkedIn that while it makes sense for Grok to take information from Twitter to answer questions, everything can get complicated very quickly: “As we know, X always had a bot-filled existence. , spam, junk content and misinformation. «If Grok was trained with all of X’s historical publications, it’s going to be difficult to guarantee their accuracy.»

It’s not the only one. Joseph Thacker, a cyber security expert, explained that although X has a large amount of information and that is a positive thing for artificial intelligences, there is also a lot of toxicity, racism and incorrect and biased information. “While that probably sounds very human, there are many risks of that coming to light,” he said.

Although Musk is arriving a little late to the fight for who develops the best generative artificial intelligence, the reality is that he has been in the world of AI for a long time. In fact, he was one of the founders of OpenAI. We have to see if he can get within range and catch up with GPT, who has been very calmly leading the race since his birth.