The news of September 16 is thatDeepSeek master of v4.1AttentionLiu ShengON SEPTEMBER 14TH, I HAD TO BURY MY TALENT YESTERDAY, SPARKING A CYBER-HOT, WITH A HUNDRED THOUSAND + READINGS OF MICRO-INTELLIGENCE NEWS, RUSHING TO THE TOP OF THE HOT LIST, AND SPARKING DISCUSSION ON X。
Liu Sheng and the blogger say:AI IS MAKING PROGRESS IN THE DESIGN OF HIS ALGORITHMS AND THE PREPARATION OF THIS FIELDIn just one year, AI has turned from an assistant who can only help me check the document, read the code, and find the bug, into a master solver who can read CUDA, PTX and SASS independently, analyzes the pause time for each instruction through a professional tool, and then optimizes the algorithm independently. It is confident that, in the near future, it will also have the capacity to independently design the algorithms, assess the performance of the different movement programmes, achieve them and optimize them。
Liu Sheng claimed that he was proud of DeepSeek v. 4.1's success, but the wheel of the times was moving forward and technology was unstoppable. He knowsIN THE FIRST HALF OF THE YEAR OR A YEAR, AI WILL PROBABLY BE AS GOOD AS HE IS, EVEN SURPASSING HIM.
Liu Sheng added that human beings have shown no hesitation since ancient times in destroying themselves. Why, knowing that “the better I write, the faster our new model will be trained and reasoned, the faster the model will advance and I will be replaced earlier”, choose to do my best to optimize it
- On the one hand, it's because it's like playing a game for meI don't know. When I invented a new technology, or saw my calculator's ability to rise, I was as excited as the fast-track player in the game had broken his own history. At the same time, when I see my own calculator's ability to outperform the official firm, I get a great sense of pride。
- But beyond that, a more important reason is that even if I “scramble” or even deliberately trip the model, the model of another family will develop and eventually kill me. “..Of course, I wish I hadn't been revolutionized, but if it had to be revolutionized, I would have wanted to kill myselfI'm sorry. At a time when everyone is so determined to destroy themselves, I am compelled to join this brutal arms race。
Liu Sheng thinksI don't have to be unemployed, but I have to get back to workI don't know. "I can still keep my job, but it may not give me a chance to do what I loved."
In the penultimate paragraph of the article, he stated that he still believed that the front-line intelligence should be available to all in an open and cheap way. He doesn't trust Anthropic or OpenAi to do this, especially if he doesn't want Anthropic to master state-of-the-art artificial intelligence or AGI. "That's why I chose and insisted on staying at DeepSeek:We look at powerful, fast, inclusive artificial intelligence and open it up, and maybe we can pull the world back from 2077. "
1AI NOTES THAT ON THE NIGHT OF SEPTEMBER 15TH, LIU SHENG AND HIS SUPPLEMENTARY NOTE TO " I HAD TO BURY MY TALENT YESTERDAY " , STATING THAT HE INTENDED TO WRITE THE ARTICLENot to express anxiety about unemploymentNot to emphasize DeepSeek’s open and inclusive nature and Anthropic’s gratification, but to say good-bye to the time when he used to write his books。
Liu Sheng thinks that there will be a day"Handwritten algorithms" and even "programming" may become a recreational activity rather than a productive activityIt's like there's almost no one who can hunt with a spear, but rather as a race. “It is tantamount to forcing me to give up what I once loved and to force myself in another direction. Even the new direction is equally fascinating, and this feeling is not so good. As the title says, I had to bury my talent yesterday.”
Liu Sheng also mentioned that the penultimate paragraph of the article "I had to bury my talent yesterday" represented only personal points of viewIt doesn't represent the position of any of his companies.