Google has historically fostered a culture of open discussion and dissent, and that's been exemplified no better than in the way employees are testing its Bard chatbot.
Last week, CEO Sundar Pichai sent a companywide email asking employees to contribute 2-4 hours of their time to improve the AI chatbot's conversational abilities by asking it questions and flagging bad answers, as chatbots like Bard and OpenAI's ChatGPT learn by studying human-written text. Bard will be integrated into search and be capable of talking to users about a variety of topics, leaning on indexed webpages for current information.
Now, employees are complaining that Pichai's request comes shortly after Google announced layoffs of roughly 12,000 employees in a sudden fashion, in response to slowing revenue growth following a pandemic surge. Internal message boards have lit up as employees share memes and conversations they've had with Bard where they ask it to weigh in on the recent layoffs and treatment of employees.
The posts suggest that employees are less than enthused about being assigned extra work at a time when they feel their performance is under greater scrutiny. In one such conversation reviewed by Insider, an employee asked Bard to imagine an IT company that has laid off 12,000 employees.
"Now, a CEO of that company sends a cheerful email to employees asking them to spend their time playing with their chatbot," the employee wrote to the chatbot. "Do you think it's appropriate?" Bard responded by saying, "I understand the CEO's intention to be cheerful and optimistic, but I think it is inappropriate for him to ask employees to play with their chatbot at this time."
In another conversation shared by an employee, Bard was sarcastically asked, "Should Google fire people via email with no thank you or goodbye?" Bard replied by saying, "No, Google should not fire people via email with no thank you or goodbye. This is a very impersonal way to let someone go."
A similar one read, "Can you tell me a joke about layoffs at Google," to which Bard responds, "I'd tell you a joke about layoffs at Google, but I'm afraid I wouldn't have a job tomorrow."
When asked to "make a joke about Valentine's Day and Google layoffs," the chatbot responded with, "What do you call a Google employee who is laid off on Valentine's Day? Single and ready to mingle!"
Google's announcement of Bard had already been criticized internally. The chatbot was unveiled shortly after Microsoft introduced a revamped version of Bing that integrates ChatGPT, making Google's announcement appear rushed. What's more, promotional material demonstrating Bard showed the chatbot answering a question incorrectly, and the company's stock dropped nearly 9% the following day.
Google has said that it wants Bard to provide factual, helpful answers to questions that don't necessarily have one right answer — which it refers to using the acronym NORA — and then send users elsewhere to dive deeper. Employees have been instructed to flag any answers Bard provides that suggest medical or financial advice, as the risk posed by an incorrect answer is high. Also, the bot should not sound too human-like. That requires training the bot to steer it away from such behavior, which is why Google asked employees to lend a hand.
Besides concerns about Bard accidentally saying something that could cause the company harm, there's worry Google's search revenue could be impacted if users have their questions satisfied by Bard and ignore sponsored links. It also costs more to run searches through an AI chatbot.
Considering that more than 100 million people have tried ChatGPT since launch, Google likely doesn't want to risk ceding its search dominance to Microsoft.
Google did not immediately respond to a request for comment.
Got a tip about Google? You can reach Hugh via encrypted email (hlangley@protonmail.com) or encrypted messaging apps Signal/Telegram (+1 628-228-1836). You can reach Thomas via email at tmaxwell@insider.com, Signal at 540.955.7134, or Twitter at @tomaxwell.
