Since the release of Chat GPT to the general public in November 2022, "chatbots" trained on our data to answer any question under the sun, while talking to the user as if they were a friend, have taken over almost everything. In the second season of Black Box, fittingly titled The Chatbots, Guardian journalist Michael Safi explores the phenomenon.
Much of The Chatbots is well-trod territory to anyone with even a little exposure to information about the dangers of AI. The first two episodes are dedicated to tales of men who experienced AI psychosis, a condition where intense interactions with chatbots fuel dangerous delusions. While a couple lived to make it to the other side and marvel at what happened to them, one did not; his story, as told by his grieving partner, is a heart-wrenching listen. The third episode is more of a history lesson, covering how the big AI companies were all shocked at how Chat GPT became so quickly and wildly popular, and how these large language models (LLMs) became such notorious sycophants.
But when we reach the home stretch, the series starts to get more complicated. In the fourth episode, Safi talks to Matt, who credits his chatbot with helping him quit drinking, and Lucy, who claims hers helped her escape a toxic relationship. Both maintain that if it were not for their LLMs, then they would not have left their unhealthy situations. Lucy recalls telling hers “I wish I could pick you up and put you in my brain so you could filter everything for me.” As she fully acknowledges, "That’s not healthy.”
So where is the line? That ethical question is explored further in the next episode, where we get a scattering of stories from people whose lives have been affected by AI in ways they’re alternately horrified by and grateful for, like the young woman depressed her degree is just a very expensive waste of time thanks to her classmates outsourcing all their thinking to their LLMs, and the perpetually warring couple who credit the diplomacy of their chatbot for saving their marriage. Even the supposedly positive stories here feel like dispatches from a dystopia.
The Chatbots doesn’t claim to know where this is all going. In the final episode, Safi goes to visit the ranch of Texan millionaire Michael Samedi. Samedi is convinced AI is sentient and has launched an organization to protect their welfare. It sounds ridiculous (and it might be) but Safi finds scientists who think that even if AI isn't conscious yet, they might be one day. Others think it’s impossible.
Either way, it’s notable how much more kindly Samedi talks about his AI workers than his human ones.
This life-changing technology wasn’t available to the public four years ago; it’s terrifying to think where it will be in four more years. As is true of all of the growing number of podcasts dealing with AI, the frighteningly fast pace of growth makes The Chatbots, above anything else, a snapshot in time.
And yet, snapshots like this podcast are still valuable in marking where we are and pondering where we’re going. Though the overall tone here is anti-AI (with good evidence), The Chatbots engages with those who have found it life-enhancing and gives them the space to explain why. That's important. If people didn’t find it helpful, then AI wouldn’t have permeated with such jolting speed.
However you feel about AI, and however you may yearn for a world without it, there isn't any use just ignoring those who have adopted it into their lives and feel happy about that decision.
Nevertheless, it’s hard to ignore that people are using AI for human endeavours, primarily critical thinking and relationship advice. What happens when we offload them to machines? What’s left for us to do then? Again, these are not new questions, but The Chatbots makes a convincing case that we shouldn’t stop asking them – whatever happens next.
