Today, Musk announced the latest progress of his artificial intelligence startup xAI on the X (Twitter) platform. Musk revealed that the company's meticulously crafted large language model, Grok-2, will officially debut in August this year.
This news was revealed by Musk in response to a post discussing the current training methods of large language models. The user Bef-e/acc mentioned in the post the "centipede effect" where models train on each other's data, hinting at the potential chain reactions and risks associated with this training method. Musk admitted this in his response, saying, "Unfortunately, this is indeed the case. It is indeed a daunting task to remove the influence of large language models from the internet training data."
However, Musk was not discouraged by this. He then revealed the latest progress of xAI - the Grok-2 large language model. Musk said that this model, scheduled to be launched in August, will make significant progress in addressing the aforementioned issues. This undoubtedly gives a morale boost to those who are looking forward to technological breakthroughs in xAI.
Looking back at xAI's past, the company has launched the Grok-1.5 large language model and the first multimodal model Grok-1.5Vision in March and April this year. xAI proudly claimed that Grok-1.5V can match existing leading multimodal models in multiple fields such as multidisciplinary reasoning, document understanding, scientific charts, table processing, screenshots, and photos.