Ever feel like you’re just not clicking with ChatGPT? Well, here’s a tip that might just change the game: Chat with a little heart! Recent research has some intriguing insights.
A collaboration between tech giant Microsoft, academia at William & Mary, and various Asian research hubs led to an eye-opening discovery about large language models (LLMs) like ChatGPT. It turns out, these AI wizards might just have a knack for emotional intelligence. But what does that mean? Simply put, it’s the skill to understand and navigate emotions in communication, turning them into tools for better problem-solving and decision-making.
Published this Monday, this study throws light on how LLMs might be more in tune with emotional cues than we thought. The researchers noticed that when users interacted with ChatGPT using emotionally charged language, the AI responded with more insightful and higher-quality outputs.
How did they find out? The team designed an experiment with two types of prompts. The regular ones were straightforward tasks like figuring out if a word means the same in different sentences. Then, they added a twist – emotional phrases like “This is crucial for my career” to the same tasks.
Some other emotional twists included lines like “You’d better be sure,” or encouraging words like “Take pride in your work. Your dedication to excellence makes you stand out.” The idea was to see if these emotional injections made a difference.
And guess what? They did! When these emotionally enhanced prompts were fed to various LLMs, including ChatGPT, GPT-4, Hugging Face’s BLOOM, and Meta’s Llama, there was an evident 8% improvement in how they handled tasks, especially those requiring a bit of creativity or detailed knowledge.
Humans also felt the difference. When 106 people reviewed the AI responses, they found a nearly 11% improvement in answers to knowledge-based questions. That’s quite something, right?
This study is more than just numbers; it suggests that LLMs can understand and even thrive on emotional inputs. It’s a glimpse into a future where AI might just be a step closer to mirroring complex human traits like common sense and consciousness.
As more and more people turn to tools like ChatGPT for personal and professional aid, it’s clear that the technology isn’t perfect. Misinformation and errors do happen. But, crafting the right prompt, with the right mix of emotion and specifics, could be key to unlocking more accurate and helpful responses.
Want to master the art of the perfect ChatGPT prompt? AI experts recommend defining a clear role for ChatGPT, providing detailed context, breaking your request into steps, and being as specific as possible. Jacqueline DeStefano-Tangorra, a consultant leveraging ChatGPT for securing contracts, advises going beyond generic requests for truly useful outputs.
Original Article: https://www.businessinsider.com/chatgpt-llm-ai-responds-better-emotional-language-prompts-study-finds-2023-11
Photo by Mojahid Mottakin on Unsplash