Financial Times Interview - Meta's chatbot personas

The Financial Times interviewed me about Meta's newly announced feature - chatbots with "personas". Some risks and potential mitigations.

➤ Background

👉Meta is expected to launch a new feature - users will be able to communicate with another using chatbots with a range of personas, such as Abraham Lincoln and surfer dude.

👉Zuckerberg said he envisaged AI “agents that act as assistants, coaches or that can help you interact with businesses and creators”

👉Link:

https://www.ft.com/content/fa76c8ce-cdfd-458c-baec-73dceb2d2ad5

➤Here are some risks to keep in mind, expanding on what I said in the article

☢️Nudging - Meta can use the chatbots to take nudging to the next level. Think of a chatbot that has been customized to convince a person like you in particular.

☢️Data collection - anything a person says in a conversation with a chatbot can be used for any purpose, including training to improve nudging.

☢️Predators and fraud - We know that predators and fraudsters already use chats to deceive people. Imagine a chatbot persona that is very effective at seeming like a 13-year-old girl.

☢️Isolation - We know that social media can increase isolation in teenagers and children. Imagine that instead of chatting with one another, they are only chatting with bots.

➤But there are guardrails that Meta could put in place to prevent that. For example:

💉Create a button to permit/forbid chatbots from contacting you.

💉For children, make the default setting "off" for interacting with chatbots.

💉Limit the permitted use cases for the chatbots. Forbid extreme versions of nudging, at least for certain products.

➤ Thank you to Hannah Murphy, the reporter, for a great article and for interviewing me.