No surprise, Elon Musk announced a big launch on X. The notice was that Grok AI 2.0 is going to drop in August. He conceded that it appears hard to get out all the LLMs from the training data. Grok 2, though, will do a lot better in that regard.
Musk didn’t expand his tweet but doubled down on Grok 2’s expected superiority. In March, he said Grok 2 “should exceed current AI on all metrics.” This ambitious statement raised expectations.
Curiously, the life of Grok 2 seems to be very short with us. Musk has since tweeted that Grok 3 will release by the end of the year. It’s going to be big because it’s going to train on 100,000 Nvidia H100 GPUs. Musk promises that Grok 3 “should be really something special.”
Grok was announced as Musk’s answer to ChatGPT back in November, and then Grok 1.5 debuted in April and was only made available to Premium subscribers on x; this is as Musk’s xAI startup prepares to build a supercomputer in Memphis, Tennessee, where the facility will handle Grok’s training.
After all, Grok AI had a few reliability issues from the very beginning it was launched. Even though Musk tried to push back against the “propaganda machine” of ChatGPT, Grok has suffered in sharing fake news like attributing a false quote for Indian Prime Minister Narendra Modi about his election.
To counter this, X introduced a “fun mode” in Grok. One such mode will generate news in a silly and childish manner. This feature would be unsuitable during election years. The Grok experiences summarize the entire complexity developing sophisticated AI. Otherwise, Musk’s plans with Grok AI are extensive and of interest.
