The AI oneupmanship continued this week with simultaneous announcements of new features and improved AI models from Google and OpenAI.
At its Made By Google event on Tuesday, Google unveiled its new Pixel mobile devices with an AI assistant powered by its Gemini models. The Gemini Live interactive voice chat feature is a direct challenge to OpenAI’s yet-to-be-released voice assistant.
Gemini Live enables a user to interact with their Android device in a free-flowing conversational manner, much like the chat assistant OpenAI demonstrated a few months ago. While OpenAI has only made its chat-enabled GPT-4o available to a limited number of users in an alpha release, Google says it is rolling out Gemini Live to its users.
Gemini Live is available now to Gemini Advanced subscribers, but only in English initially. Google says the rollout of Gemini Live has started and will be available on Pixel, Samsung, and other Android phones.
The 10 voices users can choose from continue Gemini’s star constellation theme with names like Vega, Dipper, and Ursa.
Here’s the part of the keynote where you can watch the demo of Gemini Live. It seems Google learned a lesson from its staged Gemini demos and did all of these demos live this time around.
OpenAI’s cheeky reveal
OpenAI seems to time its announcements of new products and updates around Google, Apple, and Meta events. Announcing an updated ChatGPT-4o model yesterday was unlikely to have been a coincidence.
The AI rumor mill has been working overtime with speculation that OpenAI was about to launch a new foundational model. We don’t have GPT-5 yet, unfortunately, but it turns out that the anonymous model that appeared on the Chatbot Arena a few days ago is an updated OpenAI model called ChatGPT-4o-latest.
When Senior Vice President of Devices & Services at Google, Rick Osterloh, kicked off Google’s Keynote, he proudly announced that Google’s experimental Gemini 1.5 Pro model topped the LMSYS leaderboard.
That announcement didn’t age well, with LMSYS announcing today that ChatGPT-4o-latest has seen OpenAI regain the top spot.
Exciting Update from Chatbot Arena!
The latest OpenAI?ref_src=twsrc%5Etfw”>@OpenAI ChatGPT-4o (20240808) API has been tested under “anonymous-chatbot” for the past week with over 11,000 community votes.
OpenAI has now successfully re-claimed the #1 position, surpassing Google’s Gemini-1.5-Pro-Exp with an… https://t.co/9lJlASI9UW pic.twitter.com/gxCDuBOi9N
— lmsys.org (@lmsysorg) August 14, 2024
The new ChatGPT-4o rankings in LMSYS Chatbot Arena see it topping each category:
- Overall: #1
- Math: #1-2
- Coding: #1
- Hard Prompts: #1
- Instruction-Following: #1
- Longer Query: #1
- Multi-Turn: #1
The bump we saw in ChatGPT’s performance over the last few days is likely down to this new model being deployed.
OpenAI has announced its DevDay 2024 events will take place in multiple cities this year. They’ll start in San Francisco (1 October) and then head to London (30 October) and Singapore (21 November). OpenAI emphasized that it would not use these events to announce a new frontier model, so don’t expect a GPT-5 release anytime soon.
Will Google, Apple, and Meta add those dates to their calendars for product announcements to steal OpenAI’s thunder?
