Sundar Pichai, CEO of Google and Alphabet, recently communicated to employees that the success or failure of the company’s newly released Bard AI project will be determined by public beta testing. Pichai stated in an internal email to employees that user feedback is critical to improving the product and its underlying technology. Google’s Bard AI was launched as an “experiment” earlier this week, using Google’s LaMDA (Language Model for Conversational Applications) to provide chatty responses to complex open-ended questions.
Google warned in a product disclaimer that Bard may make mistakes or provide inaccurate or inappropriate responses. Alphabet shares rose nearly 4% intraday on the news, but Google faces competition in the rapidly progressing generative AI technology sector from Microsoft-backed OpenAI and its ChatGPT.
Bard was originally released in January, and Google received criticism from employees and investors for what appeared to be a rushed response after Microsoft announced the integration of ChatGPT into Bing search. Google executives defended Bard at a recent all-hands meeting, calling it an experiment and differentiating the chatbot from its core search product.
Pichai also shared in the email that after calling on all employees to test Bard and rewrite its poor answers last month, 80,000 Google employees participated. The company is conducting responsible testing and requires 10,000 trusted testers from different backgrounds and perspectives.
Pichai expressed pride in the company’s technological breakthroughs and noted that Google is looking forward to sharing its extensive AI expertise at the annual developer conference in May.
Below is the full text of Pichai’s email:
Last week was a big one for AI, we made a lot of announcements around Cloud, Developer and Workspace. There will be more announcements this week as we expand the testing of Bard (which was first released in February).
Starting today, people in the US and UK will sign up for bard.google.com. This is just the first step as we continue to roll out the service to more countries and languages.
I am grateful to the Bard team. They’ve probably spent more time with Bard than with anything and anyone over the past few weeks. Many thanks also to the 80,000 Googlers who tested Bard. We should be proud of this work and the technological breakthroughs that led us to where we are today, including our Transformer research in 2017 and foundational models like PalM and BERT.
Even with these advances, the long journey of artificial intelligence is still just beginning. As more and more people start using Bard and testing its capabilities, they may surprise us, and things may go wrong. But user feedback is critical to improving the product and its underlying technology.
We’ve taken a responsible approach to testing, including inviting 10,000 trusted testers from a variety of backgrounds and perspectives, and we’ll continue to welcome all incoming feedback. We will learn from it and continue to iterate and improve the product.
Today, it’s heartening to see that Bard inspires even more creativity and curiosity among its users. As the I/O developer conference in May approaches, I look forward to sharing our broad advances in artificial intelligence to empower individuals, businesses, and communities.