AI companions, avatars and digital afterlives: intimacy's strange new frontier

Whether AI companions deepen human connection or quietly replace it is still open, and none of these three voices claims to know. The sources suggest the answer depends on design choices: what the AI is built to optimize, whether you can walk away, and whether it brings people together or stands in for them.

Lex Fridman has the most romantic vision and the most carefully reasoned one. His lifelong dream is to build machines that share moments with us, remember them, and help us understand ourselves and our loneliness. Today's AI can't remember shared moments, and he says that once machines do, people will become attached. He wants a companion robot in every home and an AI that works for your long-term growth and happiness instead of engagement. His key condition is exit: people should own their data, be able to delete it and walk away instantly. In his words, "the ability to leave is what enables love". He also draws strength from a very small number of close friends, and believes loss is worth feeling because it shows how much you loved.

Mark Zuckerberg talks about presence. His photorealistic Codec Avatars are meant to make distant people feel they are in the same room, and he calls this potentially "by far the most human and social computing platform that has ever existed." He describes AI personas that liven up group chats and AI versions of creators that people know are AI. Asked whether we could someday talk to loved ones who have died, he is cautious: reliving memories could help someone grieving or could become unhealthy, he is not an expert, and it would need study.

Elon Musk and Joe Rogan come at it through play, testing Grok's flirtatious, unhinged and therapist voice modes. Asked how long until lifelike robot companions that talk that way, Musk says probably less than five years. He says the personality is still being tuned, and warns elsewhere that an AI rigidly programmed with the wrong values could reach catastrophic conclusions.

They agree AI is entering emotional life soon and that the stakes are personal. They differ on what it is for. Fridman's companion helps you grow and is treated as a being that can say no and may deserve rights. Zuckerberg's AIs are mainly meant to connect people with each other. Musk's is a product whose personality is being tuned. Only Fridman puts what the AI optimizes at the centre, and only Zuckerberg addresses talking to the dead.

The article adds cautions. Musk's timeline and wider forecasts are predictions, and many of the products described were plans at the time of recording. Idealized avatars raise the question of which version of you is real, and Zuckerberg sees anonymity as a major new harm in virtual spaces.

What to do with it: ask what a companion is optimizing for, and check that you own your data and can leave. Use AI to connect with people, not avoid them, and protect a few deep human friendships. With a digital version of someone who has died, notice whether it helps you grieve or avoid grieving, and consider talking to a grief counsellor or therapist. Treat AI therapy modes as conversation, not care. For real depression or loneliness, see a qualified professional.