Who needs people when AI is so "human"?
In the past couple of weeks, a host of new products and features has emerged that try to make AI your friend.
People already have a weird, confused relationship with AI. Because AI chatbots can make noises that simulate human speech, some 70% of people are polite to AI chatbots, according to a December, 2024, survey conducted by Future.
Most people agree with my Unicorn Roast podcast co-host, Emily Forlini, and choose to be polite to chatbots because it feels better to sound friendly and to keep the vibe in the room pleasant. They do it for themselves and other people within earshot rather than for the AI.
Here’s where it gets weird: A whopping 12% are polite to AI because they're afraid of an AI uprising that will come after humans and possibly spare those who have been nice to it.
The false belief that AI chatbots have thoughts, feelings, intentions, and desires will only rise as chatbot makers optimize their products deliberately to create this delusion.
In the past nine days, the industry has churned out several products and features designed to make chatbots more "human":
Tolan
Portola's Tolan is an app with a customized cartoon "alien" programmed to be your friend. Its "personality" is designed to "evolve" as you interact with it. Users can chat and share photos, both of which engender responses from the "alien." Tolan automatically keeps a journal tracking user-alien interactions.
TikTok videos show a large number of users, mostly young women (the app's target demo), asking their personalized alien for advice on cooking, homework, and relationships.
The app soft-launched in December and hard-launched on February 25.
Sesame
A chatbot from Sesame came out on February 27, billed as a "demo," which takes humanlike voice attributes to a new level. The voices are called Maya or Miles (yes, the main characters in the 2004 wine-focused movie Sideways), depending on whether you choose the female or male voice. The chatbot is like Pi.ai, which I’ve talked about in this newsletter in the past, but with even more vocal ticks, pauses, colloquialisms, and other mannerisms.
The Sesame chatbot recalls previous conversations, and will even bring them up out of the blue. The chatbot asks for feedback and jokes around sometimes.
The Sesame chatbots keep an open conversation, and if you don't say anything, it will just come back and talk to you with comments and questions after 10 seconds of your silence. If you’re still silent, it will still come back with more chatter.
It often talks as if you said something, even when you didn't say anything.
This attribute, where it keeps talking even if the user is silent, may prove irresistible for some lonely people who want to feel like there's another person in the room or just want to hear a human-like voice that seems to “know” them. It's the closest thing to a relationship AI, as in the 2013 Spike Jonze movie "Her," starring Scarlett Johansson and Joaquin Phoenix.
Perplexity AI
Perplexity AI is an advanced AI-powered search engine (technically a retrieval-augmented generation, or RAG, model, which combines a training dataset with real-time searches using an optimized version of Google's PageRank).
The mobile app has a voice mode, so you can prompt it by talking to get the answer in the form of a voice. In the past, the voice mode required you to press and hold a button while speaking, like a walkie-talkie.
But as of February 26, the app doesn't require the press of a button. It just keeps listening and will respond when (and only when) you talk. By making the app persistent or ambient, it's more like there's a person in the room you can talk to at any time and less like an app you're "using."
The voice in Perplexity AI sounds perfectly natural but thankfully neutral, without embellishments or conspicuous chatter designed to display a "personality." Sesame and Pi keep talking and asking questions, whereas Perplexity AI only speaks when spoken to.
Grok 3
Elon Musk's xAI released the early preview of the Grok 3 chatbot on February 23. The new version has an "Unhinged" personality designed to make it seem more human. The “Unhinged” mode can be offensive and unpredictable, sometimes using profanity and feigning emotions. It can insult and belittle the user and even terminate the conversation with a user.
"Unhinged" is a user-selectable "personality." Other "personalities include "Storyteller," "Romantic," and "Conspiracy."
Musk said in January that the "Unhinged" "personality" is "intended to be objectionable, inappropriate, and offensive" — like Musk.
The age of "humanized" AI chatbots
We're entering a world where AI is extremely social and conversational, potentially always there and often "unhinged."
It's not just that chatbots are being "humanized," but that "humanization" is being approached from a wide variety of angles, enabling each user to choose their favorite approach.
These changes, which "humanize" the state of the art of conversational LLM chatbots, have occurred in the past week and a half.
Imagine what the next week and a half will bring.
More From Elgan Media, Inc.
Unicorn Roast: Vaping Tamagotchis
Why fast-learning robots are wearing glasses
No, Apple and Meta aren’t making humanoid robots
Where’s Mike? DFW in Dallas, Texas!
(Why I’m always traveling.)
AI can’t give you sex. At least, not yet.
I have been known to swear at them. Back in the day I used to swear at Eliza.