DailyPost 3010
OPEN AI “WE’VE ACHIEVED AGI”
AGI has been made the current battle in the world of tech and by a direct corollary of the real world itself, by OpenAI, after establishing their credibility with the release and stupendous success of ChatGPT. OpenAI is considered to be at the cutting edge of the ongoing global AI research to achieve the cherished sci-fi dream of Artificial General Intelligence, AGI. The recent $6.6 billion funding of OpenAI solidifies the company’s commitment to pushing AI boundaries. In this backdrop the recent claim of having achieved AGI by a member of OpenAI’s technical team has set the world on fire.
What AGI would achieve even at a nascent level will be a paradigm in our existence itself and it for this reason, this claim needs to be dissected in our interest and fortunately it is being done. To start with the claim is controversial but there is no denying the fact there has been a rapid pace of advancements in artificial intelligence, increasing the capabilities of modern AI systems considerably. The model in question is OpenAI’s o3 models. OpenAI’s three month old o1 model, had demonstrated exceptional performance in various complex tasks, sparking critical discussions about the current stage of AGI.
Its latest o3 models performed exceptionally well on the ARC-AGI benchmark. This benchmark measures an AI system’s ability to generalize and acquire new skills. The o3 models represent advanced reasoning and problem solving skills, excelling in benchmarks. Whether AGI has been achieved or not is the moot question. The benchmark creator Francois Chollet, responded in an unequivocal no. He said that the o3 is very “impressive and represents a big milestone on the way towards AGI.” The challenge is that it still cannot solve a fair number of simple tasks.
Chollet added, “these capabilities are a new territory and they demand serious scientific consideration.” Core challenges are the ability to generalize and acquire new skills. There have been concerns over the so-called scaling laws, suggesting that the AI models performance improves predictably with the increase in model size, dataset size, and its computational power. Some critics say AI is slowing and talk about limits of current approaches. Semantic understanding; finding information based on deeper meanings and relationship between concepts still seems to be way off. OpenAI has still not confirmed a public release for the o3 models. The public release of o3 models would clarify where we are positioned in our journey to achieving AGI.
EVERYONE WANTS TO HOP ONTO THE AGI BANDWAGON MUCH BEFORE IT IS REALISED.
Sanjay Sahay