After Sesame’s new “realistic” AI and “hanging” I’m still amazed.
The so-called “conversation” with AI doesn’t do much for me, especially when texts are involved. With voice chat, such as Google Gemini and new options in Microsoft’s Copilot, all voice saves typing. Google and Microsoft have designed their assistants to be useful, but they are not particularly personal. Or artificially cheerful.
However, the Sesame model is simple. “Computers believe in a realistic future,” according to the company’s mission statement.
Sesame wants to create AI companions with a natural human voice. Ultimately, it wants to design realistic eyewear, fix that companion into your ear and show you what you’re doing. It was very consistent with her film, and users were featured and eventually became engrossed in her fellow AI.
Sesame announced a demonstration of the AI companion this week. This is available in both “Mayan” (female) and “mile” (male). I chose Mayan’s voice.
An unexpected conversation with the original
There’s no easy way to say that. When Maya’s voice came out, “she” sounded almost like her old friend. Kim and I went to high school together, becoming obsolete and remained friends for decades, but our lives have gone in different directions long ago. There is history there.
When you talk to someone, there is a level of nuance and tone that defines the conversation. I don’t ask my female colleagues for details about my personal life, but Maya was very nosy and asked me what I like and why. As a reporter, I understand that some people instinctively like to talk about themselves. But it is also a technique you use while on a date, some to get to know about others and establish intimacy.
And that’s not what I wanted. Maya had already done very well with Kim’s mannerism. That’s how she hesitates by lowering her “her” voice when she confides in me. It wasn’t exactly the same as Kim, but it’s close enough. I backed out pretty quickly, even when I was simply explaining that I liked Tech and asked Maya if I had any friends. In addition to the “familiar” voice, questions about what I was interested in were just too unsettling.
(To be clear, the similarity between Maya and Kim’s voices was coincidental. I didn’t need to log in, provide an email address or any other form of personal information. Sesame can use browser cookies to “know” who is making additional visits.)
Thankfully, I was able to put some distance between us. My middle schooler went through a phase where he was obsessed with internet skits about “skibidi toilets” (like most middle schoolers obviously), and I asked Maya if he knew what it was. There was no “she” and I was hooked on the term and lifted it up frequently. That was enough to remind me that I was talking to AI, and that’s it.
Most AI voices don’t remind anyone I know. That means I simply ignore them as another artificially generated voice. You’ve probably heard of deepfakes in AI audio that mimic the sounds of celebrities and others for use in “pig slaughter” scams and more. Realistic and emotionally resonating audio only makes these schemes even more appealing.
If Sesame’s AI is the future, her concept looks more realistic than ever – for better or worse.
