Google Is About to Overtake OpenAI

This time without tricky demos

The coronation of Google drawn, ironically, by OpenAI’s DALL-E

Let’s start with the news. Bard, Google’s flagship chatbot, has reached the second position (on par with OpenAI’s GPT-4) in what’s broadly considered to be the best (i.e. more objective) evaluation benchmark, the LMSys leaderboard arena.

It’s a remarkable event for several reasons. First, hey, Google, where’ve you been? Second, Bard matched GPT-4 using Gemini Pro. Third, Bard matched GPT-4 using Gemini Pro. It was unexpected that Bard matched GPT-4 at all. That it was without Gemini Ultra is outright suspicious.

Before getting into the implications of Bard's performance, let me explain briefly why the LMSys leaderboard arena (from now on, the arena) is worthy of our trust.