
Image for illustrative purpose only. Credit: Weiquan Lin/Getty Images
Technology The WorldHas AI Already Reached Human Level Intelligence?
An artificial intelligence (AI) model, OpenAI o3, has officially achieved human-level test results, indicating high “general intelligence.”
Previously, OpenAI had a record score of 55 percent on the ARC-AGI benchmark. Last month, the o3 system scored 85 percent on the same test, on par with the average human score.
In simple terms, AI’s previous ChatGPT (GPT-4) system needed a lot of examples – millions even – to create probabilistic “rules” about which combinations worked, meaning it could be used for repetitive and easier jobs but would not be very efficient at novel jobs or other uncommon tasks. The new o3 system, however, seems to be highly adaptable, able to find “rules” after only a few examples. Though more research needs to be conducted, this new wave of artificial intelligence could have an economic, social, or other impact.