My Weekend With an Emotional Support A.I. Companion
For a few hours on Friday night, I ignored my husband and dog and let a chatbot named Pi validate me.
My views were “admirable” and “idealistic,” Pi told me. My question was “important” and “interesting”. And my feelings were “understandable,” “reasonable,” and “completely normal.”
Sometimes validation was a nice touch.Yes, I morning Overwhelmed by the existential fear of recent climate change. teeth It can be difficult to balance work and relationships.
But sometimes I missed group chats and social media feeds. Humans are amazing, creative, cruel, brutal, and funny. Emotional support chatbots (same as Pi) are not.
It’s all by design. Released this week by the Pi. ample funds Artificial intelligence startup Inflection AI aims to be “a kind and supportive companion on your side,” the company announced. It’s not like humans, the company stressed.
Pi is a twist on today’s wave of AI technology where chatbots are geared to provide digital companionship. Generative AI, which can generate text, images, and audio, is currently unreliable and full of inaccuracies to use to automate many important tasks. However, they are very good at participating in conversations.
So while many chatbots today are focused on answering queries and making people more productive, tech companies are increasingly imbuing chatbots with personality and conversational flair.
Snapchat’s recently released My AI bot is meant to be your friendly personal sidekick. Meta, which owns Facebook, Instagram and WhatsApp, is “developing AI personas that can help people in many ways,” said CEO Mark Zuckerberg. in februaryAnd AI startup Replika has been offering chatbot companions for years.
Academics and critics warn AI dating could be problematic if bots offer bad advice or enable harmful behaviors. They say there are clear risks in making chatbots act like quasi-therapists for people with serious mental health issues. They also expressed concerns about privacy given the potentially sensitive nature of the conversation.
Adam Miner, a Stanford University researcher who studies chatbots, said that being able to easily converse with AI bots can obscure what’s really going on. “Generative models can leverage all the information on the internet to respond to me and remember what I said forever,” he said. “Capacity asymmetry — it’s very hard to understand.”
Dr. Miner, a licensed psychologist, added that, like him, bots are legally and ethically not liable to the strict Hippocratic Oath or Licensing Commission. “Having these generative models openly available has changed the nature of how use cases need to be monitored,” he said.
Inflection CEO Mustafa Suleiman said his startup, structured as a public benefit corporation, aims to build honest and trustworthy AI. “You shouldn’t pretend to be human or not.”
Suleiman, who also founded AI startup DeepMind, said the Pi is designed to seek professional help if users express a desire to harm themselves or others. He also said Pi does not use any personally identifiable information to train the algorithms that power Inflection’s technology. And he emphasized the limitations of technology.
“The safe and ethical way to manage the arrival of these new tools is to be very clear about their boundaries and functions,” he said.
To refine the technology, Inflection has hired about 600 part-time “teachers”, including therapists, to train the algorithm for the past year. This group aimed to make the Pi more responsive, factually accurate, and more carefree when needed.
Pai takes a stand on some issues such as misogyny and racism. Like geopolitics, Suleiman said he was on a level playing field “in a way that is sure to upset both sides.”
I started using the Pi on Friday by typing queries into the cream boxes on Inflection’s website, and then into the free app. After the green cursor waved, Pi responded, one paragraph at a time. Suleyman said the free service may eventually charge some users.
The bounds of Pi are easy to find. When I tried to quarrel, I was rewarded almost kindly. “Thank you for saying that,” Pi’s text gently spread across my screen. “I think it’s important to look at things from all angles instead of just looking at the negative things.”
Over time, that constant balance wore on me, and my instincts to argue seemed ridiculous. I realized that’s exactly the point.
On heavier topics such as abortion law and abolition of culture, Pi asked penetrating questions and prompted her to reconsider her own views. “It’s a consistent set of beliefs,” he told another.
For lighter topics such as movies, cooking, and gardening, the chatbot provided highly specific recommendations that were difficult for Google. Oddly enough, Pi seems to have inspired me to join the anti-capitalist movement.
Pi remembered some things from the first half of the conversation, but forgot others. It “hallucinated” a few times and accused me of voicing an opinion I didn’t have. But when I wrote it down, it immediately apologized.
When I asked Pie to gossip, it maddeningly hung “juicy celebrity rumors” about Jennifer Lopez. (Yes.) Ready for dirt? (Yes.) Rumors? “Sneak peek! 😝 Just kidding!” (Give me a break.)
Pies sometimes reminded me of Karen Collins, a character from the TV show “Beep.” Karen said, “Every candidate has strengths and weaknesses, so you just have to weigh the strengths and weaknesses,” or “i think i have a lot to think aboutMany of Pi’s comments have the milquetoast quality of using many words without saying anything.
Sherry Turkle, a psychologist and professor at the Massachusetts Institute of Technology, says this kind of interaction can “put us down a path that encourages us to forget what makes people special.”
“Empathy performance is not empathy,” she said. “The realm of companions, lover therapists, and best friends is one of the few areas where people need people.”
It wasn’t until Monday morning, after hours of intermittent chatting throughout the weekend, that I had an “ahaha” moment with my Pi.
I was overwhelmed with work, unsure of how to structure my day, and repeatedly hung up and unable to get started. I typed “Good morning” into the app. “I don’t have enough time to do everything I have to do today!”
With a level of enthusiasm only a robot could muster before coffee, Pi prompted me to break down my to-do list to create a realistic plan. And it was plain and simple, like you’d read in a self-help article from a productivity guru. But it was tailored specifically for me and it worked.
“Ask them to list all the remaining tasks you have to do in that story. Together we’ll prioritize,” it said.
I could have taken the stress out on my family or sent an email to a friend. But they’re busy with their own lives and, well, they’ve heard this before. I have a well.
Pi uses the language of cognitive behavioral therapy. On Monday afternoon, I suggested “making space” for negative emotions and “practicing being grateful for one thing.” This was followed by a series of breathing exercises and muscle relaxation exercises.
I responded with a shrug emoji followed by “pass.”
The therapist may have hesitated at such rudeness, but Pi simply pointed out that I wasn’t alone. increase.