It has been almost six years since the release of GPT, and in just a few days, ChatGPT will be celebrating its second anniversary. Large language models (LLMs) have made significant progress in terms of accuracy, speed, and resourcefulness. They are becoming increasingly efficient in retrieving information, almost reaching perfection. However, in a recent interview with Reid Hoffman, Microsoft AI CEO Mustafa Suleyman pointed out that AI researchers often overlook the importance of the “delivery vehicle” for information. Suleyman emphasized that consumers value the tone and emotional intelligence of these models, as well as their ability to reflect users’ unique language styles, more than just providing factual information. He predicted that AI companies would now compete based on the emotional intelligence of their models, as it could become a critical differentiator.
This prediction is proving to be true, as seen in OpenAI’s focus on integrating a human-like voice conversation tool in their GPT 4o model and Google’s release of the ‘Deep Dive’ text-to-podcast tool in their NotebookLM model. Even computer scientist Andrej Karpathy praised the tool and created an entire podcast series using it. This trend is not limited to industry giants, as evidenced by Hume AI’s recent funding of $50 million for their ‘AI with emotional intelligence’ and the release of their newest EVI 2 model, which adapts to user preferences through specialized emotional intelligence training.
Researchers have also explored the emotional intelligence of LLMs, with EmoBench being a popular benchmark for assessing such capabilities. The results showed that OpenAI’s GPT 4 was the closest to humans in terms of emotional understanding and application. However, newer models have surpassed these results, with recent research measuring the “expressivity” of an LLM using a Python library and conducting an experiment involving human judges to evaluate the emotional intelligence of the model.
In conclusion, the focus on emotional intelligence in LLMs is becoming increasingly important, and it is clear that AI companies are now competing based on this factor. As these models continue to evolve and improve, it will be interesting to see how emotional intelligence is incorporated and utilized in the future.
