This post was originally published on this site.
Alex Kantrowitz writes the Big Technology newsletter on Substack and hosts the Big Technology podcast.
I’ve had a gnawing feeling lately about AI. Although the prophecies that chatbots would take our jobs have so far been false, the underlying technology’s evolution is making me believe it may soon automate much of my work, and perhaps yours too.
More than two years after ChatGPT’s release, generative AI’s core use case is extending beyond chat. There are now early iterations of software that operates your computer on your behalf. There are AIs that will do serious, if imperfect, research for you. And perhaps most unnerving, there are AIs that simulate human voices and personalities, going out into the world to seek and share information, and even entertain. It may not be time to panic, but I’m starting to rethink some of my calmer assumptions about where this all leads.
A few weeks ago, I met with Evan Ratliff, a journalist who cloned his voice with AI, attached it to an AI model, and had it talk to friends, family, and even a therapist. Ratliff captured the experience in a podcast series called Shell Game. And as he relayed the finer details to me in person — I wasn’t risking meeting his AI bot on Zoom — he shocked me with one anecdote.
Ratliff says his voice bot conducted an interview with a tech CEO, and the bot was able to obtain better information than he, the human journalist, could. Before the call, Ratliff prompted his AI clone with questions and instructed it to ask anything else potentially relevant. The tech CEO played along with the interview — he works in AI voice tech, for what it’s worth — and gamely responded to the bot’s questions. When Ratliff listened to the recording, he was surprised to find the CEO really opened up.
“He was a little more forthcoming with the AI than he was with me,” Ratliff told me. “There’s a quality of, you don’t necessarily feel like there’s someone there and you might be a little more intimate than you would have otherwise. And that can be very valuable in an interview for a reporting project.”
As a fellow reporter, I’d long assumed that our work — obtaining new information and disseminating it — would insulate us from AI automation. It’s a job that requires building relationships, eliciting information from sources, and communicating it effectively. That was supposed to be as human as it gets. But Ratliff’s bot showed otherwise.
The typical rebuttal here is that AI still can’t fake personality, but I am sorry to report that it can. Last year, Google released NotebookLM, a product that generates lively podcasts hosted by AI voice bots from just a few links or documents. The podcasts aren’t perfect. But as AI practitioners say, they’re worse now than they’ll ever be. And they’re really not bad today. Already, multiple members of my podcast audience have asked me if I rented my voice to Google for the project. I did not. But the insinuation is unnerving.
I doubt this stops with our profession. Sales and customer service are directly within AI’s reach, with more fields to follow. Sometimes loudly and sometimes in a hush, tech executives tell me that AIs in these fields can be more patient, more knowledgeable, and better listeners than humans. And as the technology advances, it’s getting easier to build and integrate the applications.
Sebastian Siemiatkowski, the CEO of Klarna, a financial services provider, tells me that his company was able to reduce outsourced human customer service after deploying AI agents, without a noticeable quality dropoff. “When we started exposing this chat AI agent to customers and they had the opportunity to interact with it as an alternative to a human agent, the customer satisfaction on that was equal to a human agent in many cases,” Siemiatkowski says.
This might sound scary, but I still believe it’s unlikely that AI will fully take over your job, or mine, anytime soon. The technology is showing signs that it can take on more advanced work, but it still struggles to multitask and it often requires a human guiding it, as in the case of Ratliff.
Some say these shortcomings, too, will be overcome in a matter of time. AI optimists say human-level artificial intelligence, or AGI, is just two or three years away. But generative AI pioneers like ex-OpenAI chief scientist Ilya Sutskever have suggested that the proven methods for improving these models are about to be exhausted. So an imminent robot takeover is far from certain.
Still, as AI extends beyond the chatbot — and toward something that can research, take calls, and even pontificate — it’ll likely become a force multiplier, used to scale up individuals’ effort and help them cover more ground. That might lead to less hiring, smaller companies, or potentially fewer overall jobs.
And now I’m less confident in our broader ability to weather this change without pain.