OpenAI Declares the “AGI Era”

 

OpenAI Declares the “AGI Era”

AGI (Artificial General Intelligence) refers to AI that can perform a wide range of intellectual tasks rather than specializing in a single field.

These systems have reached a level where they can discover new security vulnerabilities in web browsers or operating systems and develop ways to exploit them without step-by-step human instructions. OpenAI is gradually making advanced cybersecurity capabilities available only to verified users while strengthening safeguards against dangerous requests and unauthorized actions.

We have not reached a consensus that “AGI has fully arrived,” but we have entered a stage where the question “Could this be early AGI?” deserves serious debate.

I believe the most accurate description of the present moment is the threshold of general-purpose AI or early AGI.

There is still no universally accepted definition of AGI.

  • OpenAI: Highly autonomous systems that outperform humans at most economically valuable work.

  • Google DeepMind: A framework that evaluates not only performance but also the breadth of capabilities, generality, and autonomy across different levels.

  • ARC Prize: The ability to efficiently acquire new skills in unfamiliar environments, as humans do.

Because these standards differ, it is difficult to imagine a single official declaration suddenly establishing that “AGI begins today.” OpenAI’s Definition, DeepMind’s Levels of AGI Framework

Nevertheless, recent systems are qualitatively different from earlier AI.

They handle writing, coding, images, search, and tool use within a single system, carry out extended plans, and adapt considerably well to unfamiliar problems.

In particular, in September 2026, GPT‑6 Astra scored 99.9% on ARC‑AGI‑3 when using a dedicated interface. This benchmark tests the ability to independently discover rules in unfamiliar environments.

It also completed 96% of stages using fewer actions than humans.

This is clearly a major milestone. ARC Prize Evaluation Results




However, this result alone does not prove that AGI has been achieved.

  • The same model’s score dropped substantially to 62.7% when using a provider-neutral interface.

  • The evaluation took place in a closed environment with limited rules and objectives.

  • Supervision is still necessary when dealing with ambiguous real-world goals, long-term responsibilities, interpersonal relationships, organizational management, and exceptional situations.

  • Plausible-sounding errors, overconfidence, and problems with memory and consistency have not been fully resolved.

  • Passing a benchmark and reliably performing most real-world jobs are two different claims.

ARC Prize also described the result as a major advance in generalization while explicitly stating that it was not claiming AGI had been achieved.

My conclusion is therefore:

Technologically, we are approaching the end of the pre-AGI era. But it is still too early to say that an era of AGI validated through real-world use has begun.

What matters most, however, is the impact rather than the label.

Even before AGI is officially recognized, the ways we perform office work, develop software, teach, and create content could change dramatically.

Just as the gradual unfolding of the Industrial Revolution mattered more than the precise date the steam engine was invented, AGI is likely to arrive through a gradual process of becoming integrated into the economy and everyday life, rather than through a single announcement.

It is reasonable to see the present moment as the middle of that transition.

OpenAI and Anthropic Are Preparing to Go Public

Another reason the competition between new models is attracting attention is that both companies are pursuing listings on U.S. stock exchanges.

Anthropic and OpenAI confidentially submitted their IPO filings in June.

Developing and operating AI models requires enormous investment in advanced chips and data centers. As the two companies prepare to go public, the performance and commercial potential of their new models will provide important support for their valuations.

For AI models to perform more complex tasks, they need to move large amounts of data quickly.

The semiconductor technology that supports this is high-bandwidth memory (HBM). As AI competition intensifies, attention is turning to Samsung Electronics and SK hynix, two leading HBM manufacturers.

Samsung Electronics, in particular, is expanding its collaboration with OpenAI on next-generation semiconductor development. Competition among AI models is therefore also driving semiconductor investment and demand.




#AGI #AGIEra #ArtificialGeneralIntelligence #OpenAI #AIRevolution #FutureTechnology #AIOutlook

Comments

Post a Comment

Popular posts from this blog

Could I Be Losing My Hair? If You’re Worried About Low Hair Density

Image
Could I Be Losing My Hair? What to Check First If You’re Worried About Low Hair Density “My hair has always been sparse, but lately it seems even thinner.” Have you ever had this thought while looking in the mirror? When you feel less hair in your hand while tying it back, or your scalp shows unusually clearly at the crown in photos, you may worry that hair loss has begun. However, having sparse hair alone does not mean you have hair loss. Everyone has a different natural hair density and thickness, and how much of the scalp is visible can also vary depending on your part and the lighting. What matters is not whether you have less hair than other people, but whether your hair has changed compared with how it used to look. Check whether your part is gradually widening, whether the thickness of the hair at the front is changing, or whether hair density is decreasing in a particular area. Wet hair or oily strands that clump together expose more of the scalp, so it is best to compare your ...
Read more

모발 영양제, 누구에게 필요할까? 복용 전 확인할 점

Image
  모발 영양제, 누구에게 필요할까? 복용 전 확인할 점 머리카락이 많이 빠지거나 가늘어진 것 같으면 모발 영양제부터 찾아보게 됩니다.  제품 설명에는 비오틴, 아연, 철분처럼 모발 건강과 관련된 영양소가 적혀 있어, 여러 성분을 함께 먹으면 머리숱이 늘어날 것처럼 느껴질 수 있습니다. 하지만 모발 영양제가 모든 탈모의 해결책은 아닙니다. 머리카락이 빠지는 이유에는 영양 부족뿐 아니라 유전, 출산, 질환, 약물 복용 등 여러 가지가 있습니다.  원인을 확인하지 않은 채 영양제만 먹으면 필요한 관리를 놓칠 수 있습니다.  반대로 실제로 부족한 영양소가 확인된 사람에게는 의료진이 그 영양소의 보충을 권할 수 있습니다.  중요한 기준은 광고에 나온 전후 사진이나 후기보다 내게 부족한 영양소가 있는지, 복용하는 제품에 무엇이 얼마나 들어 있는지 입니다.  영양제는 식사를 대신하지 않으며, 많이 먹을수록 머리카락이 빨리 자라는 것도 아닙니다.  복용을 결정하기 전에는 빠지는 양과 기간, 식사 상태, 함께 나타나는 증상을 살펴보고 필요한 경우 진료를 통해 원인을 확인하는 편이 좋습니다. 참조 자료 :  미국피부과학회 탈모 관리 안내 1. 영양제가 도움이 될 수 있는 사람은 누구일까?  모발 영양제의 필요성을 생각할 때는 먼저 영양소 결핍이 확인되었는지 살펴봐야 합니다.  철분이나 아연 등 일부 영양소가 부족하면 머리카락이 눈에 띄게 빠질 수 있고, 의료진은 진찰과 필요한 검사를 바탕으로 보충 여부를 판단합니다.  식사를 지나치게 제한하거나 충분한 단백질을 섭취하지 못하는 경우에도 영양 상태를 돌아볼 필요가 있습니다.  그렇다고 머리카락이 빠지는 사람 모두에게 같은 검사가 필요하거나, 같은 영양제를 먹어야 하는 것은 아닙니다.  예를 들어 머리카락이 특정 부위에서 점차 가늘어지는 경우와 큰 수술이나 출산 뒤 일시적으로 많이 빠지는 경우는 살펴볼 원인이 다를...
Read more