Google just unveiled their latest Large Language Model (LLM) in Dec, named Gemini. It is multi-modal and reportedly performs better than GPT4.
Does it really? Let's find out.
In my limited experiments for text-based tasks, GPT4 is still better than Gemini Pro in Constrained Generation, Fact Checking, Advanced Coding and Tool Use.
It's performance is still superior to ChatGPT and Llama 2, so that's a good first step.
Probably with more data collected from users on Gemini via Bard, it will likely get even better.
Also, Gemini Ultra (the best version) may be better, and I look forward to evaluating it when it is out.
0:00 Introduction
0:40 Classification and JSON Prompting
1:11 Creative Generation
3:09 Fact Retrieval and Checking
5:06 Math
5:50 Coding
7:46 Tool Use
9:57 Verdict
~~~~~~~~~~~~~~
AI and ML enthusiast. Likes to think about the essences behind breakthroughs of AI and explain it in a simple and relatable way. Also, I am an avid game creator.
14