The development of the artificial intelligence has a challenge: artificial general intelligence (AGI), and it seems that OpenAI It already achieved this with its o3 system, after a test carried out on December 20.
Is this a model of intelligence comparable to the human level?
It was known that the AI model achieved a score of 85% in the ARC-AGI test, well above the other systems. In addition, he excelled in mathematics assessments.
This has led to the belief that OpenAI has already achieved Artificial General Intelligence (AGI), which has become a challenge for AI developers, notes a Gizmodo report, although there are still skeptical voices at this level of AI.
The test involves subjecting the AI to the “sample efficiency” of an AI system to adapt to something new.
For now, it is known that systems like ChatGPT (GPT-4) are not very efficient with samples. This was trained with millions of human texts, under certain rules, with the combination of words, but which lead it to perform repetitive tasks.
ARC-AGI testing makes the AI model able to cope with changes.