Bard welcomes a big update: finally supports Chinese! How is the actual experience?

Text: VickyXiao Juny Editor: VickyXiao

Original title: “Bard welcomes a big update: finally supports Chinese! Go and “trick” it”

With both ChatGPT and Claude ushering in major updates, Bard under Google is finally not far behind and has released a new version, but compared to ChatGPT’s artifact Code interpreter, Bard’s update is more to further improve Bard’s user experience.

However, for Chinese users, this update is of great significance: because Bard has finally added support for Chinese this time.

Newly added 40 languages Chinese dialogue without pressure

Bard has added more than 40 languages this time, in addition to Chinese, it also includes Arabic, German, Hindi and Spanish.

The Silicon Stars immediately chatted with Bard. Bard certainly answered some general questions fluently. Take a “Chinese Level 10” topic to test to see if it doesn’t understand the profoundness of Chinese:

Not bad answer.

But when the Silicon Stars asked it to write a seven-character quatrain, Bard still overturned:

More New Experiences

Additionally, Bard has added voice support. The new version of Bard adds a small megaphone icon, click on it, and you can hear Bard read out the answer. This is especially useful for users who want to hear the correct pronunciation of words or listen to a poem or play. This function now supports more than 40 languages, and Chinese is also supported.

Plus, users can easily adjust Bard’s responses. Users can now change the tone and style of Bard’s responses to five different options: simple, long, short, professional or casual. For example, if you feel that Bard’s answer is too long, you can use the drop-down menu to shorten the answer. This feature is currently only supported in English.

At the I/O conference, Google announced that it would bring the functionality of Google Lens to Bard. And this update realizes the integration of Google Lens. Users can now upload images with images, and Bard will analyze the image content and information to help. This feature is also currently only available in English.

Source: Twitter

In addition, this Bard has also made some adjustments at the product level.

• Pin and Rename Conversations: Options to pin, rename and choose recent conversations are now visible in the sidebar so users can revisit these prompts later.

• Export code to more places: Allows users to export Python code to Replit in addition to Google Colab.

• Share responses with friends: Shareable links allow users to share ideas and creations with others.

**What is the strength of the Palm 2 behind the Bard? **

Google’s Bard is trained based on its own PaLM 2 model.

The original PaLM is a large language model announced by Google in April 2022, using 540 billion parameters for training, about three times that of GPT-3. The new version of PaLM 2 has been further enhanced and improved on PaLM, with multilingual, reasoning and coding functions.

PaLM 2 received additional training on multilingual texts, covering more than 100 languages, and was remarkable for its ability to understand, generate, and translate nuanced texts including idioms, poems, and riddles, and also passed the “mastery” level of advanced language proficiency exam.

In terms of reasoning, PaLM 2’s data sets include scientific papers and web pages with mathematical expressions, and have super logical, common sense reasoning and mathematical abilities.

At the same time, PaLM2 has been pre-trained on a large number of public source code datasets, and the ability to write code is stronger. In addition to Python and Java, special-purpose code generation in Prolog, Fortran, and Verilog is also included.

It is worth noting that PaLM 2 has been developed into different versions, which can be targeted at different customers and deployed in different enterprise environments.

At present, PaLM 2 has four specifications, from small to large, they are Gecko, Otter, Bison and Unicorn. Among them, the smallest Gecko can run on mobile phones and can process about 20 Tokens per second, which is about 16 or 17. word. In other words, instead of spending a lot of time and resources to create and tune PaLM 2, developers can use it directly for deployment.

However, as far as Bard’s current performance is concerned, it is always at least one step behind ChatGPT. ChatGPT’s Code Interpreter plug-in is finally fully launched. Many people have said that Code Interpreter is GPT 4.5 wearing a plug-in mask. This Bard update, in addition to Supporting Chinese makes Chinese users very excited, but there are not many other surprises.

When will Bard be able to amplify his moves?