🥊OpenAI Hits Back at Musk

OpenAI fires back at Musk, calls lawsuit ‘sour grapes’

Key Points 

• OpenAI released emails and texts claiming Musk’s allegations are misleading.

• Emails show Musk proposed merging OpenAI with a hardware startup, demanding majority equity and CEO control.

• Musk’s proposal to spin OpenAI into Tesla with a $1 billion budget was also rejected by its leadership.

• OpenAI states Musk declined multiple offers for equity in its for-profit wing after resigning in 2018.

🥸Context of the news - Elon Musk’s ongoing legal dispute with OpenAI revolves around his claim that the company abandoned its original nonprofit mission to share AI advancements for the benefit of all. Last month, Musk’s legal team escalated matters by filing an injunction to stop OpenAI’s transition to a for-profit model.  

♨️News - On Friday, OpenAI responded to Musk’s lawsuit by releasing emails and texts that, according to the company, expose Musk’s claims as misleading.  

🍨Here’s the scoop - OpenAI argues Musk’s allegations stem from “sour grapes” rather than legitimate concerns. The newly published exchanges reveal that as early as 2015, Musk supported the idea of a hybrid nonprofit and for-profit structure. While OpenAI initially launched as a nonprofit, funding challenges later pushed the organization to consider a for-profit model.  

The emails show Musk proposed merging OpenAI with a hardware startup and even suggested becoming its CEO, demanding a 50%-60% equity stake. When OpenAI rejected these terms, Musk suggested spinning OpenAI into Tesla with a $1 billion budget, which was also declined. Musk eventually stepped down in 2018 and severed most ties with the company.  

OpenAI also claims Musk has repeatedly turned down offers of equity in its for-profit wing. In its statement, OpenAI remarked, “You can’t sue your way to artificial general intelligence.” It expressed respect for Musk’s achievements but urged him to compete in the AI field, not in court.  

Ilya Sutskever envisions superintelligent AI becoming self-aware and unpredictable

Key Points 

• Ilya Sutskever predicts superintelligent AI will be more self-aware, unpredictable, and capable of reasoning beyond current systems.

• He believes these advanced AI systems will have real decision-making abilities, unlike today’s “slightly agentic” AI.

• Sutskever envisions superintelligent AI seeking rights, which he considers a positive development for human-AI coexistence.

🤓News - Ilya Sutskever, the former chief scientist at OpenAI, made a bold prediction about the future of artificial intelligence during his talk at NeurIPS, the annual AI conference, on Friday. He believes that reasoning capabilities will soon make AI technology much less predictable. 

Sutskever, who was also honored with an award for his contributions to the AI field, shared his insights on the direction of the industry.

✨Breaking down the highlights - Sutskever discussed a concept his team explored a decade ago: that scaling up data to "pre-train" AI systems could push the technology to new heights. However, he pointed out that this approach is reaching its limits. While innovations like ChatGPT were made possible by increased data and computing power, Sutskever said "pre-training as we know it will unquestionably end." He explained that while computing power continues to grow, data is no longer expanding at the same rate, as there is only one internet.

To overcome this, he proposed that AI technology could generate its own data or evaluate multiple answers before selecting the most accurate response. Some researchers are also focusing on incorporating real-world data to move the field forward.

Sutskever wrapped up his talk with a bold prediction about the future of AI, saying that superintelligent machines are "obviously" on the way, though some may disagree. He believes that at some point, AI will become far more capable than humans in many areas. These superintelligent systems, he says, will be "very different" from what we have today—and in some ways, unrecognizable.

He explained that these future machines will be much more "agentic," meaning they’ll have real decision-making abilities, unlike today's AI, which he describes as "very slightly agentic." They will be able to reason and think on their own, which will make them more unpredictable. These superintelligent AIs could also understand things from limited data and might even become self-aware. Sutskever also suggested that they could eventually want rights, which in his opinion, is a positive outcome.