Google quickly released an intelligent chatbot Bard, trying to keep up in the race of generative AI. The rush was evident not only after release, but also during internal testing. According to Bloomberg, company employees noted Bard’s lack of readiness to launch, calling the bot a “pathological liar,” “disgusting” and “worse than useless.”
Citing internal documents and testimony from 18 current and former employees, the publication claims that Google was aware of Bard’s unsatisfactory performance but decided to release it anyway.
- One employee message in an internal chat with 7,000 members said: “Bard is worse than useless: please don’t run,” an assessment many agreed with.
- Another employee approached Bard with a question about how to land a plane – the bot regularly gave answers that could cause it to crash.
- He also gave answers about scuba diving that were “likely to result in serious injury or death.”
Despite requests and comments from employees, Google “rescinded the risk assessment” presented by its internal security team, which warned that Bard was not ready for release.
Bloomberg sources say that Google, trying to keep up with competitors, offers low-quality information with little regard for ethical standards. It is alleged that employees responsible for the safety and ethical implications of new products have been told to stay away from tools being developed for generative AI.
- Google fired AI researcher Timnit Gebra in 2020 after she wrote a research paper on the unethical nature of AI language systems.
- Margaret Mitchell, one of II’s ethics team leaders who wrote a paper along the same lines, was fired a few months later amid allegations of misconduct.
Google is significantly behind its main competitor, Microsoft, in the field of generative AI. The integration of artificial intelligence into Bing has not only grown the browser’s audience to more than 100 million daily active users, but has also forced Samsung to consider switching from Google search to Bing on its devices.
Google contractors randomly rate Bard’s chat responses after a limited hour of testing – Insider