Today, the latest news shows that the OpenAl Super Alignment team was disbanded on the spot. As soon as the news was issued, it immediately sparked heated discussions among netizens.
Some netizens ridiculed: This is the second season of OpenAI Gongdou has started again?
Soon, OpenAl confirmed that the company is no longer holding the so-called 'super-aligned' team as a separate entity, but is integrating the team more deeply into its research efforts to help the company achieve its security goals.
01
What does alignment team do?
It is understood that the super alignment team was established in July 2023, and its goal is to "solve (alignment) problems by 2027, supported by OpenAI's massive calculations."
Ilya Sutskever is one of the leaders of the project, and OpenAI says it will dedicate 20% of its computing power to research on the project.
In an interview, Ilya Sutskever boldly predicted that if a model can predict the next word well, it means that it will be able to understand the deep reality that led to the word. This means that if AI continues on its current path, perhaps in the near future, an AI system that surpasses humans will be born. But what is even more worrying is that "super AI" may bring some unintended negative consequences.
A few days ago, Ilya Sutskever, co-founder and chief scientist of OpenAI, announced that she was leaving OpenAI. This brings the end of the eight-year tie with OpenAI.
Many people may not have noticed that Ilya announced her departure at the same time as Jan Leike, the co-leader of the Super Alignment team.
Jan Leike is the Head of Security at OpenAl and the head of the Super Smart Alignment Team.
According to JanLeike, he and OpenAl executives have a long-standing disagreement over the company's core priorities, and the differences are so severe that they cannot be reconciled. The team encountered significant hurdles in advancing its research projects and securing computing resources, and the lack of resources seriously affected the progress and quality of the study.
02
What does a long-term disagreement mean?
Previously, during the OpenAI infighting, it was rumored that Ilya "saw something" that was powerful enough to make him worry about the future of AI and rethink the development of AI.
Since then, the American industry, including Musk, has been concerned about "what did Ilya see?" ”
Records show that Ilya has previously mentioned in a video that AI will not necessarily hate humans on the day AGI is implemented, but they will likely treat humans the way humans treat animals. People may not intend to harm animals, but if you want to build an intercity highway, you don't ask for the animals' opinions, you just do it. When faced with such a situation, AI may naturally make a similar choice.
Perhaps, this is what Ilya is worried about, and it is also the long-standing disagreement between the OpenAl alignment team and the higher-ups.
Last week, the outside world began to focus on "Where will Ilya go next?" ”
Some netizens speculated, "What would Ilya and Andrej look like if they started a truly open AI company?" ”
Andrej Karpathy, who is also one of the founding members of OpenAI, left on the eve of Sora's release, and if the two of them join forces, it could be another big event in the field of artificial intelligence.
To this, Ilya responded in a tweet, "I'm looking forward to the next plans – this project means a lot to me personally, and I'll share the details in due course." ”
Now, let's guess, "Where will Jan Leike go?" ”
If Ilya and Andrej did start an AI company, would Jan Leike join it?
[The author of this article is a dark horse, an entrepreneur who originally wrote.] If you need to reprint, please contact the WeChat public account (ID: iheima) for authorization. ］