fact-ua.com

AI has reached human-level performance in the "general intelligence" test.

Искусственный интеллект достиг человеческого уровня в тестах на общий интеллект.

On December 20, the o3 system from OpenAI achieved 85% on the ARC-AGI test, which is significantly higher than the previous best score for AI — 55% — and aligns with the average human performance. It also demonstrated strong results in a very challenging math test.

The creation of artificial general intelligence (AGI) is a goal pursued by all leading research laboratories. At first glance, OpenAI has at least made a significant leap towards this objective.

While skepticism remains, many AI researchers and developers feel that something has shifted. For many, the prospect of AI now seems more tangible, urgent, and imminent than previously anticipated.

To understand the o3 results, it's essential to know about the ARC-AGI test. This assessment measures the AI's ability to adapt to new situations by determining how many examples are needed for it to comprehend their functioning.