Google Tests AI Assistant Offering Life Advice

Google Tests AI Assistant Offering Life Advice
Last updated Aug 17, 2023
Image credit: Wikimedia Commons


  • According to a New York Times report published on Wednesday, Google's DeepMind division is developing tools that could turn generative artificial intelligence (AI) into a life coach.
  • Google has reportedly signed a contract with Scale AI — a $7.3B AI software startup — to test at least 21 personal and professional tasks, including life advice or answering intimate questions.
  • The new AI chatbot will reportedly process complex inputs and respond with situation-specific recommendations. Whether it will be a standalone product or integrated into Bard is not yet clear.
  • The news comes despite Google's AI safety experts having previously said that users taking life advice from AI could experience "diminished health and well-being" and a "loss of agency."
  • When Google launched Bard in March, it restricted the AI chatbot from giving medical, financial, or legal advice. It does, however, reportedly share mental health resources with users who experience emotional distress.
  • In separate news, Google has also been testing Genesis — an AI software for journalists that can generate news articles. Last year, its AI safety experts warned that generative AI could lead to the "deskilling of creative writers."


Narrative A

Creating an AI life coach might be Google's worst idea yet. Previous attempts at making advice AI chatbots have shown that the technology is dangerously poor at giving health advice and even worse at advising about morality, with some bots being extremely easy to manipulate into accepting racist and homophobic advice and even promoting genocidal suggestions. Google's life coach is likely to backfire.

Narrative B

AI is revolutionizing how we live, work, and communicate, so it isn't unreasonable to believe that AI life coaching could soon be a reality for many people. The technology can provide individuals with personalized guidance and assistance they might not have had access to. It could also give vulnerable users customizable, up-to-date, and more affordable 24/7 support.

Nerd narrative

There's a 50% chance that an AI will first pass a long, informed, adversarial Turing test by 2030, according to the Metaculus prediction community.

Articles on this story

Sign up to our newsletter!