AI matches human skill in detecting online sentiments
A new study finds that big AI models like GPT-4 are now comparable to people at picking up on things like mood, political leanings, emotional intensity, and even sarcasm in text.
Researchers tested seven top AIs—including Gemini and Llama—on 100 text samples with help from 33 human participants.
Both humans and AIs still struggle a bit with sarcasm, but the speed and consistency of these models could be a game-changer for fields like journalism and public health.
AI could trend faster in social sciences
AI's knack for quickly analyzing huge amounts of text could help spot trends faster in social sciences or provide real-time insights during elections or crises.
The study did flag some concerns about bias in AI results, but overall it shows tech is catching up fast when it comes to understanding what people really mean online.