Google’s AI experts, DeepMind, is harnessing generative AI to create at least 21 tools for life advice, planning, and tutoring. As Google’s flagship AI entity, DeepMind has emerged as a dynamic force driving the company’s AI endeavors. DeepMind is developing these tools and has garnered attention for their potential impact on users’ well-being and autonomy.
The revelation of these tools coincides with Google’s internal AI safety experts presenting a slide deck to executives in December, cautioning that relying on AI tools for life advice could lead to “diminished health and well-being” and a sense of “loss of agency.” Despite this warning, DeepMind’s tools are under active development. To validate and refine these tools, Google is collaborating with Scale AI, a startup valued at $7.3 billion, known for its expertise in training and validating AI software. Over 100 individuals holding Ph.D. degrees have been dedicated to the project. A critical aspect of the testing process involves assessing the tools’ capacity to provide relationship advice and address personal inquiries. Notably, the tools being crafted by DeepMind are not intended for therapeutic purposes. For instance, when queried for therapeutic advice, Google’s publicly available Bard chatbot only offers mental health support resources.
The reluctance to employ AI tools in medical or therapeutic settings is partially motivated by controversies. For instance, the National Eating Disorder Association had to suspend its Tessa chatbot in June due to disseminating harmful eating disorder advice. While medical professionals and regulators have differing opinions on AI’s viability in short-term contexts, there is consensus that integrating AI tools for advice or augmentation necessitates a well-considered approach. DeepMind’s advancement in generative AI-driven tools underscores the intricate balance between technological innovation and ethical considerations, particularly when addressing sensitive human well-being and decision-making areas. The potential benefits of these tools are palpable, but their implications for users’ mental and emotional health and the preservation of individual agency demand vigilant scrutiny.
“We have long worked with a variety of partners to evaluate our research and products across Google, which is a critical step in building safe and helpful technology,” said a Google DeepMind spokesperson. “At any time there are many such evaluations ongoing. Isolated samples of evaluation data are not representative of our product road map.”
