"MIT researchers evaluated 41 different AI models on over 11,000 text-based tasks, finding that AI's performance often resembles that of a disenchanted intern, hitting minimum benchmarks but struggling with quality."
"The study revealed that AI models scored a 7, defined as minimally sufficient, in roughly 65% of tasks, indicating some reliability but significant room for improvement."
"Despite improvements, AI still falls short when the stakes or standards are raised, emphasizing the need for human intervention to refine outputs and ensure quality."
Recent research from MIT indicates that while AI is gradually improving in performing various workplace tasks, it often resembles a disenchanted intern, meeting only minimum benchmarks. The study analyzed 41 different AI models on over 11,000 text-based tasks, revealing that AI can achieve a score of 7, defined as minimally sufficient, in about 65% of tasks. However, AI still falls short in quality when higher stakes or standards are involved, suggesting that human oversight remains crucial for quality output.
Read at Fortune
Unable to calculate read time
Collection
[
|
...
]