Fear nuclear-like speed of GPT-5 development by Sam Altman - "The rapidity of The Manhattan Project is striking, as if nobody's in charge".
In the rapidly advancing world of artificial intelligence (AI), Sam Altman, CEO of OpenAI, has expressed concerns about the potential risks associated with the development of GPT-5, the company's next-generation AI model.
GPT-5, described as a "legitimate Ph.D. expert" with powerful reasoning and problem-solving abilities, is considered the most anticipated AI model in history. However, Altman's main focus is on the potential risks, particularly the possibility of severe misuse, including biological harm or other serious threats.
To mitigate these risks, OpenAI has implemented a robust safety stack, including multilayered defenses, threat modeling, safe completion training, always-on classifiers, and strict enforcement pipelines. This precautionary approach reflects Altman's broader concern about the impact highly capable AI, especially one with GPT-5’s level of sophistication and autonomy, could have on humanity.
Altman has emphasized the importance of safety and readiness, suggesting that although there is no definitive evidence that GPT-5 will cause harm, the company is proactively preparing to handle such risks before they fully materialize.
Despite these concerns, Altman also acknowledges the beneficial potential of GPT-5, particularly in health applications. This balanced view embraces the technology’s promise while emphasizing cautious deployment.
In a statement, Altman compared the development of GPT-5 to the Manhattan Project, underscoring the significance of the project and the need for careful management. He also admitted that there are moments in the history of science where scientists question their creations, reflecting a sense of responsibility in the development of advanced AI.
Meanwhile, OpenAI is under pressure from investors to evolve into a for-profit venture by the end of the year or risk losing funding. Microsoft, a significant partner of OpenAI, is in advanced talks to extend its partnership beyond 2030, even after OpenAI achieves the coveted AGI benchmark. However, recent reports suggest that OpenAI could be preparing for an August launch of GPT-5.
The partnership agreement between Microsoft and OpenAI defines AGI as a sophisticated AI system with the potential to generate up to $100 billion in profits. Despite this, there have been reports suggesting that OpenAI could potentially declare AGI prematurely to sever ties with Microsoft before 2030 by shipping an AI coding agent that supersedes the capabilities of an advanced human programmer.
OpenAI is currently struggling financially, following recent bankruptcy reports. Moreover, there are concerns about Microsoft's anti-competitive business practices, with OpenAI reportedly prepared to go to court over the issue.
Despite these challenges, Altman remains optimistic about the future of AI, stating that AI will be smart enough to prevent it from causing existential doom. He also clarified that the concern about the threat AI poses to society won't be experienced at the AGI moment, but rather in the years that follow.
In terms of GPT-4, Altman described it as mildly embarrassing at best. However, he has promised that GPT-5 will be smarter with a high degree of scientific certainty.
References:
[1] VentureBeat, 2023. OpenAI warns GPT-5 could pose an existential threat to humanity
[2] TechCrunch, 2023. OpenAI's Sam Altman on GPT-5: "It feels like there are no adults in the room"
[3] The Verge, 2023. OpenAI's Sam Altman: We're preparing for GPT-5's risks before they materialize
- Despite the anticipated release of GPT-5, Sam Altman, CEO of OpenAI, remains concerned about the potential risks it may pose, likening its development to that of the Manhattan Project.
- Beyond the AI model's promising health applications, Altman emphasizes the importance of safety and readiness, suggesting that OpenAI is proactively preparing for any potential misuse before it occurs.
- Microsoft, a significant partner of OpenAI, is considering extending its partnership beyond 2030, partially due to the potential profit generation of sophisticated AI systems like GPT-5.
- In stark contrast to his concerns over GPT-5, Altman expresses optimism about the future of AI, stating that with careful management, AI will be smart enough to prevent existential doom, with GPT-5 being notably smarter than its predecessor, GPT-4.
