Over the past few months, speculations and rumors have been swirling the windmill, indicating top AI labs, including OpenAI and Anthropic might be on the verge of achieving the coveted AGI (artificial general intelligence) benchmark. OpenAI CEO Sam Altman indicated the benchmark would be met sooner than anticipated, narrowing the timeline to the next 5 years. However, contrary to popular belief, the executive claimed the milestone would whoosh by with surprisingly little societal impact.
Anthropic CEO Dario Amodei speculations suggested AGI could be achieved in 2026 or 2027 based on extrapolated curves of the progression of advanced AI models. Your guess on when the benchmark will be achieved could be as good as mine. The speculations mount at a critical time when several emerging reports suggest AI progression has hit a wall, making it difficult for AI firms to develop sophisticated AI systems due to a lack of high-quality content for model training.
But as it now seems, we might have gotten a roadmap on when AGI might be achieved, at least according to OpenAI CEO Sam Altman. The executive recently addressed recent AGI reports as Twitter hype on X:
"twitter hype is out of control again. we are not gonna deploy AGI next month, nor have we built it. we have some very cool stuff for you but pls chill and cut your expectations 100x!"
twitter hype is out of control again. we are not gonna deploy AGI next month, nor have we built it.we have some very cool stuff for you but pls chill and cut your expectations 100x!January 20, 2025
Per Altman's usual cryptic social media posts, we can at the very least not anticipate hitting the AGI benchmark next month and should lower our expectations 100x. X sleuths quickly responded to the executive's post, highlighting a message he'd posted on X at the beginning of the year citing, "near the singularity; unclear which side."
What we know about AGI so far
During OpenAI's 12 days of shipmas extravaganza, the ChatGPT maker finally shipped its o1 reasoning model to broad availability. A technical employee at the firm claimed the model's release constituted AGI:
"In my opinion we have already achieved AGI and it’s even more clear with O1. We have not achieved “better than any human at any task” but what we have is “better than most humans at most tasks”. Some say LLMs only know how to follow a recipe. Firstly, no one can really explain what a trillion parameter deep neural net can learn. But even if you believe that, the whole scientific method can be summarized as a recipe: observe, hypothesize, and verify."
However, a recent report revealed that AGI might have an entirely different meaning than previously thought, at least figuratively. According to Microsoft's deal with OpenAI, the ChatGPT maker will only hit the coveted benchmark when it develops an AI system to generate up to $100 billion in profit.
Altman claimed that the benchmark is achievable with current hardware and that his team at OpenAI knows how to build it as it seemingly shifts focus to superintelligence. Interestingly, popular AI safety researcher Roman Yampolskiy (well-known for his 99.999999% prediction that AI will end humanity) indicated that anyone can build AGI as long as " you have enough money to buy enough compute."
Related: Sam Altman predicts superintelligence will trigger a 10x surge in scientific AI breakthroughs
It's no secret that advanced AI requires an exorbitant amount of resources. Sam Altman was branded "a podcasting bro" after indicating that it would "take $7 trillion and many years to build 36 semiconductor plants and additional data centers" to fulfill his ambitious AI vision.
