‘Far from human level’: AI models score below 25% on job tasks, UC Berkeley study finds
SMRTR summary
A UC Berkeley study tested popular AI models on over 1,500 real-world professional tasks across 55 industries and found that none scored above 25%. The best performer, OpenAI's ChatGPT-5.5, only passed 24% of tasks, while every model scored 0% on the hardest challenges. Researchers say repetitive jobs face the greatest automation risk, while decision-heavy roles will likely remain secure longer.
SMRTR provides this summary for quick context. The original article belongs to Reddit.
Read the original article