After Google released the Gemini Ultra model for public use last night, comparisons to other models in the market have begun, especially with GPT-4, which is currently leading the market.
Ethan Mollick, a professor at the Wharton School of the University of Pennsylvania, had the opportunity to use Gemini Ultra a month before the general public. He posted his sincere opinion, stating that Gemini Ultra (also known as Gemini Advanced) is on par with GPT-4, making the AI industry now have two leading models instead of just GPT-4.
However, Mollick also mentioned that Gemini Ultra has not surpassed GPT-4 decisively. Both models have different strengths and weaknesses, but overall, they are on a similar level. GPT-4 excels in coding, while Gemini excels in explaining and communicating with visual and search results combined.
Mollick also points out that both models have similar limitations, stating that “GPT-4 is full of ghosts. Gemini is also full of ghosts.” This occurs after conversing with the models for some time, suggesting it could be a common phenomenon for models of this size.
In terms of chatbot personalities, Mollick found that Gemini is more friendly and helpful, while ChatGPT is relatively bland. As for security concerns, both models have guardrails, but they have different approaches. Gemini has a higher likelihood of responding negatively to questions or negative narratives.
TLDR: Google’s Gemini Ultra model has been compared to GPT-4, with both models having their own strengths and weaknesses. While Gemini Ultra is on par with GPT-4, it has not surpassed it significantly. Both models have limitations in terms of generating irrelevant responses. Gemini has a friendlier personality, while GPT-4 is better at coding. They also differ in their approaches to security.
Leave a Comment