Are top AI labs on the verge of hitting the AGI benchmark? Your guess could be as good as mine. However, as AI progresses, it’s increasingly becoming difficult to determine what AGI means or constitutes. AGI (Artificial General Intelligence) refers to an AI-powered system surpassing human cognitive capabilities across various tasks.
Reports indicate scaling laws have begun to stop the progression of AI, leading top AI labs, including Google, OpenAI, and Anthropic, to struggle to develop advanced AI models due to a lack of high-quality training content. OpenAI CEO Sam Altman and Anthropic’s Amodei Dario predict AGI will be achievable in the next three years. Altman indicated that the benchmark is feasible with current hardware, further citing that it would simply whoosh by with “surprisingly little” societal impact.
As top AI labs race toward AGI, Sam Altman claims the term is increasingly becoming less useful. The executive further detailed that people use the term to mean different things, including a definition that falls in OpenAI’s o1 reasoning model alley as others refer to it as true superintelligence.
Sam Altman says the term AGI gets less useful as we approach it but the o1 model seems close and by the end of 2025 we will AI systems that “can do truly astonishing cognitive tasks” where it will seem smarter than humans at a lot of hard problems pic.twitter.com/DIHx3ktrANDecember 20, 2024
The OpenAI CEO disclosed that the firm uses a 5-level framework, indicating that the company has already unlocked the second level with its flagship reasoning model. Altman further alluded that the term AGI has seemingly become too “course” as OpenAI edges closer to the benchmark.
Interestingly, the executive foresees OpenAI unlocking new heights in AI progression, leading to AI systems that “can do truly astonishing cognitive tasks” by the end of 2025. The systems will seemingly be smarter and more capable at most tasks than humans.
AGI sooner than anticipated?
Earlier this year, Sam Altman detailed in a blog post that superintelligence might only be “a few thousand days away” from OpenAI’s doorstep. Perhaps more intriguing, an OpenAI employee recently suggested OpenAI 01’s release to general availability constitutes AGI. “We have already achieved AGI,” the employee indicated. “It’s even more clear with o1.”
The employee’s remarks closely align with Sam Altman’s comments during a recent interview, indicating AGI might be achieved sooner than anticipated. But what could the AGI benchmark mean for OpenAI’s partnership with Microsoft?
A former OpenAI employee corroborated claims the AI firm could be on the verge of hitting the AGI benchmark. However, he warned that the company could face difficulty handling all it entails. Microsoft CEO admitted that OpenAI had a 2-year lead in the AI landscape, allowing the firm to build ChatGPT and sophisticated AI platforms uncontested.
Microsoft is arguably OpenAI’s largest investor, betting $13 billion on the ChatGPT maker’s technology, which grants it access to advanced AI models. However, the partnership could be severed once OpenAI hits the coveted benchmark.
Even Microsoft CEO Satya Nadella indicated that would be the only natural cause of action, but the company’s focus is reaching superintelligence first. A recent report suggested OpenAI plans to scrap the clause that would void its partnership with Microsoft after hitting the AGI benchmark. This will ensure it secures future investment and cloud computing power from the Redmond giant.
