Racing to claim a larger share of the global AI market, Google has released a version of its Bard AI that can chat with you in 43 languages as well as English.
The new version will speak out loud with users and can make comments about images, the company says.
The tricky part: teaching Bard not to blurt out words or comments any one of those 43 cultures find offensive.
To do that, Google hired native speakers of each language and told them to try to goad Bard into saying something taboo. When it did, programmers tweaked its algorithms through a process called “reinforcement learning” to not do it again.
However, Bard—like other chatbots—is less fluent in obscure languages, such as Icelandic, and more likely to make mistakes.
ChatGPT-4, the top-line version of OpenAI’s pioneering chatbot, is conversant in any language found on the Internet, a company spokesperson said.