Google’s AI recommended promoting an employee that Brin didn’t notice

Imagine this. Artificial intelligence analyzes a work chat and recommends promoting an employee you didn’t even notice. This is exactly the case described by Sergey Brin, Google’s founder, in an interview. He told that inside Google there is a system resembling Slack, but with built-in artificial intelligence. This system is capable of processing all chat content and answering complex questions.

Brin experimented with this technology, giving it tasks like “Summarize the discussion” or “Assign tasks to employees.” He copied the artificial intelligence responses back into the chat, and colleagues didn’t even understand they were communicating with a machine. When Brin asked the system “Who in this chat should be promoted?”, the artificial intelligence chose a woman whom Brin himself practically didn’t notice and who wasn’t particularly active in discussions.

Interested in this choice, Brin talked to the employee’s manager. The manager confirmed and said “I think you’re right, she really works hard and has done a lot.” As a result, the employee was promoted. But for such deep analytical decisions, the system requires a large context window. And according to rumors, a version of Gemini with practically infinite context already exists inside Google.

Such technology opens tempting prospects – for example, loading an entire project’s codebase into one window and asking artificial intelligence to constantly improve it. Although Brin noted that at Google there are about 5 internal projects for any unusual idea, the main question remains the same “How well do they work?”