December 20, the new Openai AI system called O3 overcame an important line. When passing the Arc-gi test of thinking, the model gained 85% – as much as people gain on average. In addition to this, she brilliantly coped with complex mathematical tasks.
Until now, the best result among such developments has not exceeded 55%. Many researchers believe that O3 has made a serious breakthrough on the path to creating artificial general intelligence (Agi), although skeptics still doubt the significance of achievement.
It is important to understand the features of the ARC-Agi test. The main indicator in it is “sample efficiency”, that is, the ability to adapt to new conditions, having a minimum of examples. Simply put, the test determines how many samples of a new situation are needed to catch the principle of its work.
Modern language models like Chatgpt based on
GPT models are actively used for automatically writing texts, dialogue systems, solving various problems that require processing a natural language. They open huge prospects in many areas, including marketing, education, medicine, etc.