DeepMind, one of Google's AI components, is allegedly developing 21 various applications for providing advice, planning, and tutoring, as indicated by The New York Times on Wednesday. Google is reputedly employing Scale AI, a $7.3 billion startup concentrating in training and asserting AI programming, to assess the tools. The testing includes evaluating if the tools are capable of providing relationship advice or aiding users with personal questions.
DeepMind, the "nimble, fast-paced" unit of Google's AI efforts, is working on developing at least 21 tools for providing life advice, planning and tutoring, The New York Times reported Wednesday. This comes after Google's AI safety experts had reportedly presented a slide deck in December warning of the potential for users to experience "diminished health and well-being" and a "loss of agency" when taking life advice from AI tools.To test the tools, Google has hired Scale AI, a $7.3 billion startup which specializes in training and validating AI software, and more than 100 people with Ph.D.s have been assigned to the project. Examples of potential prompts include a situation of having to decline a close friend's wedding invitation due to financial constraints.DeepMind has noted, however, that these tools are not meant to be used therapeutically. Furthermore, Google's publicly available Bard chatbot is programmed to only provide resources for mental health support when asked for therapeutic advice. This comes in light of controversy surrounding AI used in a medical or therapeutic context, such as when the National Eating Disorder Association was forced to suspend its Tessa chatbot for giving harmful advice in June.Although opinions are divided on the potential benefits of AI used in a short-term context, there is an agreement that the introduction of such tools requires careful consideration. A Google DeepMind spokesperson provided CNBC with a statement, saying: "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. Isolated samples of evaluation data are not representative of our product road map."For more details, see The New York Times.
Comments