When ChatGPT and the new Bing took the world by storm, Google was scrambling to offer an alternative to what appears to be the hottest new trend in tech. That’s where Google Bard comes in, a ChatGPT-like generative language AI model that you can converse with in a chat-like interface. Compared to ChatGPT, Bard still feels like it’s a lot less polished and like it will need a lot more updates until it’s at the same level. To make clear what progress Bard is making, Google has added a new "experiment updates" page to the chatbot, with the first changelog already listed.

The updates page is available on https://bard.google.com/updates. Unlike Bard itself, it doesn’t appear to be region-locked to the US and UK, and you can access it from anywhere in the world. The page lists the first update, with a changelog that notes the following updates, with the first one being the updates page itself:

2023.04.10

Welcome to Bard’s inaugural experiment update!

Experiment updates page

What: We’ve launched an Experiment updates page to post the latest features, improvements, and bug fixes for the Bard experiment.

Why: So that people will have an easy place to see the latest Bard updates for them to test and provide feedback.

There’s more when you click “Google it”

What: We’ve added additional suggested Search topics when people click “Google it.”

Why: People will be able to explore a broader range of interests with more related topics.

Updates to Bard’s capabilities

What: We've updated Bard with better capabilities for math and logic.

Why: Bard doesn’t always get it right on math and logic prompts and we are working toward higher-quality responses in these areas.

The biggest feature addition here is Bard’s improved math and logic capabilities. These are both areas that generative language models usually struggle with, and Bard initially outright refused to work with programming and coding prompts, with it even avoiding Excel formulas.

This new release is likely just the first one in a long row of direly needed upgrades, as the AI is generally in a much rougher state than its competitors — and let's be real, all currently available models need big improvements when it comes to fact checking and reliability.

It remains to be seen whether Google will continue to share each and every update or if the company will take an approach more in line with its Android apps in the long term. Here, the company usually tests multiple versions with different user groups simultaneously, and it only officially communicates updates once they roll out widely.

Thanks: Moshe E