On November 24, Musk posted on X, saying that the original design and engineering of all the original Tesla Roadster is now fully open-source. Not only that, but its AI modelGrok will be officially launched next week. The launch of GROK is a very important milestone. Because this happened exactly one year after the release of ChatGPT, in just one year, generative AI has grown considerably and has become a wave of deterministic technology.
Trained on billions of pieces of data to achieve high efficiency and time-saving.
It is understood that there are two modes for grok to choose from, namely normal mode and fun mode. In fun mode, Grok gives the AI more intelligence and humor, making it a unique and fun chatbot.
Compared with other AI models, Grok has a unique advantage:It can be done by getting real-time news and tweetsGet the latest information and trends quickly。For this reason, Grok provides its users with the latest news summaries in an extremely concise manner, which saves users a lot of time, which is a core competitiveness among similar AI products.
In addition, Grok has been trained on millions of tweets and threads and has excellent writing skills, and its tweets and threads will probably write better than most.
Grok is growing quite fast. In July of this year, XAI quickly trained the basic model of GROK-0, which was then tuned and evolved into GROK-1. Although the parameters of this model are only about 33 billion, the capabilities of this model have exceeded those of LLAMA2 70B and GPT-35。
Musk's backstab Open AI, is about to become"Data feudalism
Many people feel that Musk's launch of Grok at this time will pose a greater threat to OpenAI. The reason for this is not difficult to understand, because Grok has important human data. Today's large model data is quite tight, even the originator of generative AIopenaiThey all have to train directly with their own users' data.
In this case, backed by the huge social platform X, Grok has a huge and constantly updated database, which is a trump card with a lot of weight.
However, just as coins have positive aspects, the massive amount of data that X is sitting on is also a double-edged sword. As a large social platform, most of the data on X is unverified, and the quality is certainly uneven. The data is directly used to train large models without filtering, and it is difficult to ensure the quality of the content provided by grok.
Therefore, according to the analysis of industry professionals, although Musk has a lot of money, on the track of AI, he has limited room for improvement, and at most he can only become a "data feudal lord" of one side.
So, what is a data feudal?
To put it simply, it is based on unique dataFoundation, in certain verticalswith"Enclosure is king" in the fieldand strive to be the best. The concept was first coined by former Greek Finance Minister Yannis Varubax, who came to this conclusion after witnessing European countries being repeatedly exploited and exploited by American tech companies. Nowadays, the road to the application layer is gradually blocked by GPTS, and this way of relying on unique data enclosure may become the only way for other enterprises to fight their way out of the encirclement. Based on the data of the X platform, Grok seems to be developing in the direction of a new generation of data feudalists.
Such an approach will yield results in the short term, but it will not be a long-term solution. Because the so-called "unique data" seems to be an impenetrable barrier, but it is bound to be weakened with the development of technology, so that it will be eaten away by stronger general models.
This is something that we ordinary people want to get, and Musk will definitely not ignore it. Compared with the pure data feudalism, Musk's grok has shown the potential and ability to solve this problem.
Musk once praised China's WeChat in an interviewHe thinks that WeChat is a good sample with great interfaces and apps, because in China you can almost "live in WeChat".
It is worth mentioning that so far, the United States does not have a comprehensive platform like China's WeChat, which integrates social networking, payment, entertainment, and information. This is not because the United States does not have the capacity to develop such an integrated platform, but because there is a competing relationship between the major financial institutions and the tech giants in the United States, and it is difficult to achieve a cooperative relationship. Musk's acquisition of X not only ended such a "separation" to a certain extent, but also provided the foundation for it to build a super app similar to WeChat.
In time, ifThe X platform can really be onesetAudioMessagePaymentAll-in-one integrated platform, Grok will become the center of this data hubcan be usedDifferent industries, differentregion, differentmodalityMassive amounts of data. In this way, the positioning of Grok is no longer thereLimitationsOn the social platform, it has become a comprehensive mutual entrance connecting various fields.
Musk's layout of the AI game of chess is very large.
Assemble a luxury technical team to create artificial intelligence that can "understand the true nature of the universe".
Talent is the primary productivity of the AI track, and it is also one of the most money-burning conditions. And Musk, who has the "ability to make money", has also invested heavily in building a technical team. Most of the members of Musk's technical team are from Deepmind, OpenAI, Google Research, Microsoft Research, Tesla, and the University of Toronto, and they have led several AI basic algorithm research and well-known AI projects. It is worth mentioning that in Musk's teamThe proportion of Chinese is extremely high, and two of the researchers have more than 200,000 citations.
For example, Greg Yang, the co-founder of XAI, received his bachelor's degree in mathematics and master's degree in computer science from Harvard, where he studied under Yau. Yau took Yang to attend the event, met doctoral students and mathematicians from all directions, and recommended him to apply for the Morgan Prize, the highest honor that undergraduate students can achieve in mathematics.
In addition to Yang Ge, there are also Guodong Zhang, Zihang Dai, Yuhuai Tony Wu and others in the core team, who have made great achievements in the underlying technology.
It can be seen that XAI has no shortage of talents, let alone computing power and funds. While there is still a clear gap between the GRAK-1 and GPT-4 that is being launched today, the gap between the two is bound to get smaller if a version with larger parameters becomes available.