Google Officially Releases Bard, Its ChatGPT Competitor
Google today announced that it’s officially releasing Bard, the new generative language model that it unveiled several weeks ago. While it’s still a research product, Google is eager to get Bard into the hands of testers.
Bard is a lightweight and optimized version of LaMDA, the large language model (LLM) based on the Transformer neural network architecture that Google first launched back in 2017. LaMDA itself was trained on a huge corpus of data from the public Internet encompassing trillions of words, and contains up to 137 billion parameters.
Google announced Bard in early February as a counter to ChatGPT, the incredibly popular generative language model launched by OpenAI in late November. While Google researchers are responsible for many of the breakthroughs in LLMs, the launch of ChatGPT gave OpenAI much of the momentum in this burgeoning new category of tech.
When news broke in January that Microsoft was investing $10 billion into OpenAI and collaborating with it to incorporate GPT-3.5 into its Bing search engine, Google sensed that its lead in search was slipping and it declared a “code red”.
With the official launch of Bard today, Google appears eager to make a more forceful case that its LLM deserves to be considered the leader in the segment. When prompted with a question, Bard generates a response by selecting which word is likely to come next.
“You can use Bard to boost your productivity, accelerate your ideas, and fuel your curiosity,” company officials write in a blog post today. “You might ask Bard to give you tips to reach your goal of reading more books this year, explain quantum physics in simple terms, or spark your creativity by outlining a blog post.”
Instead of generating one response, Bard will often give users a series of possible responses. The user is then asked to pick the response that’s the best starting point.
There are downsides to LLMs, however. Google warns the prospective Bard users that since it’s trained on a wide range of publicly available information, biases and stereotypes may show up in its output. And sometimes it will just be wrong.
“Bard is an experiment,” Google warns the user. “Bard will not always get it right…When it doubt, use the ‘Google it’ button to check Bard’s responses.”
Bard is currently restricted to users in the UK and US who speak English, like the original Bard, William Shakespeare, did so many years ago. You can join the waitlist for Google’s Bard at bard.google.com/.
This story first appeared on sister publication Datanami.