Tech News
← Back to articles

How Can AI Companions Be Helpful, not Harmful?

read original related products more articles

For a different perspective on AI companions, see ourQ&A with Jaime Banks: How Do You Define an AI Companion?

Novel technology is often a double-edged sword. New capabilities come with new risks, and artificial intelligence is certainly no exception.

AI used for human companionship, for instance, promises an ever-present digital friend in an increasingly lonely world. Chatbots dedicated to providing social support have grown to host millions of users, and they’re now being embodied in physical companions. Researchers are just beginning to understand the nature of these interactions, but one essential question has already emerged: Do AI companions ease our woes or contribute to them?

RELATED: How Do You Define an AI Companion?

Brad Knox is a research associate professor of computer science at the University of Texas at Austin who researches human-computer interaction and reinforcement learning. He previously started a company making simple robotic pets with lifelike personalities, and in December, Knox and his colleagues at UT Austin published a pre-print paper on the potential harms of AI companions—AI systems that provide companionship, whether designed to do so or not.

Knox spoke with IEEE Spectrum about the rise of AI companions, their risks, and where they diverge from human relationships.

Why AI Companions are Popular

Why are AI companions becoming more popular?

Knox: My sense is that the main thing motivating it is that large language models are not that difficult to adapt into effective chatbot companions. The characteristics that are needed for companionship, a lot of those boxes are checked by large language models, so fine-tuning them to adopt a persona or be a character is not that difficult.

There was a long period where chatbots and other social robots were not that compelling. I was a postdoc at the MIT Media Lab in Cynthia Breazeal’s group from 2012 to 2014, and I remember our group members didn’t want to interact for long with the robots that we built. The technology just wasn’t there yet. LLMs have made it so that you can have conversations that can feel quite authentic.

... continue reading