Realdoll is building a fleet of AI-powered sex robots with customizable personalities

By Luke Dormehl Published February 3, 2017

Sex doll manufacturer Realdoll is dipping its toe (and we don’t want to know which other body parts) into the world of artificial intelligence and robotics with a forthcoming robot sex assistant that promises to form a “real bond” with its, erm, users.

The new system is made up of several components, which will roll out over the course of this year and next. It will begin with the Harmony AI app, scheduled for release on April 15, followed by the company’s “first robotic head systems,” set to launch by the end of the year. A virtual reality platform will ship sometime in 2018.

Recommended Videos

It’s not going to be cheap, mind you: the head alone will set you back $10,000. No pun intended.

“We are developing the Harmony AI system to add a new layer to the relationships people can have with a Realdoll,” Realdoll CEO Matt McMullen told Digital Trends. “Many of our clients rely on their imaginations to a great degree to impose imagined personalities on their dolls. With the Harmony AI, they will be able to actually create these personalities instead of having to imagine them. They will be able to talk to their dolls, and the AI will learn about them over time through these interactions, thus creating an alternative form of relationship. The scope of conversations possible with the AI is quite diverse, and not limited to sexual subject matter.”

From the sound of things, the Harmony system may be back-compatible with some of the existing dolls the company offers, although we’re sure that more clarification will be made available at a later date. With the AI system, users can choose from a range of personality traits (kind, sexual, shy, naive, brainy, etc.) and then choose how strongly these characteristics are engrained in their new acquaintance.

“We feel that this system, and this technology, will appeal to a segment of the population that struggles with forming intimate connections with other people, whether by choice or circumstance,” McMullen continued. “Furthermore, it will likely attract those who seek to explore uncharted and new territory where relationships and sex are concerned.”

We do worry about the resale value, though.

Former Contributor

I'm a UK-based tech writer covering Cool Tech at Digital Trends. I've also written for Fast Company, Wired, the Guardian…

Topics

Emerging Tech

Your holiday photos could help scammers figure out exactly where you are

AI can extract surprisingly precise location clues from innocent-looking Instagram and Facebook posts, making phishing scams much more convincing.

Photo on a holiday during sunset

That innocent holiday snap might be doing more than showing everyone how good the weather is. A new scam highlighted by The Guardian shows how criminals can use AI to analyse photos posted on Instagram and Facebook, work out where they were taken, and then use that information to make phishing messages look frighteningly legitimate.

The trick is surprisingly simple. Imagine posting a few family photos from Porto without mentioning the city anywhere. A few days later, a text arrives claiming that your bank detected unusual card activity while you were travelling in Porto and asking you to verify your details through a link. Because the message contains a detail that only someone who knows about the trip should know, it suddenly feels much more convincing. Except the scammer may never have known about the trip at all. The photo told them.

Emerging Tech

Meta and Google just got beaten by an AI giant you might not expect

Alibaba's Qwen AI model running on an Android phone.

The AI race usually revolves around familiar names like OpenAI, Google, Meta, and Anthropic. But when it comes to models developers can actually download and build upon, another company has raced to the front — and by quite some distance. Alibaba says its Qwen family of open-weight AI models has crossed 3 billion downloads globally over the past six months. That puts Qwen ahead of some of the biggest names in AI, including models from Meta and Google.

The difference is especially striking in Hugging Face data for 2026. Google’s open models have recorded around 418 million downloads, while Meta sits at roughly 227 million, according to Hugging Face’s latest State of Open Models report. Qwen, meanwhile, is becoming much more than one popular model. Alibaba has released more than 460 Qwen models, and developers have created over 300,000 derivative models from them.

Emerging Tech

Meta promised to stop creepy AI glasses videos, but they’re still all over Instagram

Instagram app

Smart glasses can be genuinely useful, whether you’re taking a hands-free photo, listening to music, or asking an AI assistant about whatever is in front of you. Unfortunately, putting a camera on someone’s face also opens the door to some much less pleasant uses. That’s becoming a problem for Meta.

Instagram head Adam Mosseri said in July that the platform was working to remove videos in which people use Meta glasses to film and harass others secretly. However, a new Business Insider investigation suggests the cleanup isn’t going quite as smoothly as Meta might have hoped. The publication found dozens of pickup and prank videos still available on Instagram weeks after Mosseri’s comments. Some came from popular or verified creators, and several showed women being approached and recorded in public through Meta glasses.