Why AI second selves will be your greatest mentor — and opponent

If you don’t want AI to replace you, you need to use it to make you better at what you do, argues artist Harry Yeff in conversation with Accenture Song’s Mette Stannow.

In the fourth episode of the sixth season of the NEXT–Show, Harry Yeff (Creative Technologist who performs as Reeps100) and Mette Stannow (Managing Director, Accenture Song) discuss AI as an accelerant on human skill and creativity, through the creation of digital second selves.

Watch the complete episode

If you live in a world of creativity, be it a commercial one, like Stannow, or an artistic one, like Yeff, now is a potentially worrying time to be alive. The rapid rise of generative AI in the last couple of years has transformed the creative landscape. How are we to react?

Yeff, at least, is clear about how he sees AI fitting into both his future and his work. “Augmentation is quite a tired term, but at the very heart of it is a really important idea. Now that the masses have more and more access to powerful AI, we are starting to think about how we can use it to have these second selves.”

He sees his core job as collecting as many human vocal techniques as possible — and pushing them to the very edge. And he had experienced the power of AI to help him with that in his teens.

And so began a multi-year experimentation with a range of systems, from bespoke models to public APIs, exploring how the tools could augment his ability with his voice. His contemporaries absolutely hated the early version of his experiments. “There was a lot of discomfort, which I never felt, I felt excitement.”

A 20-year journey with AI

That excitement had roots in his childhood. Yeff grew up in a “rough situation” and in what he describes as a neurodivergent home. While his parents did their best, it was a very unstructured environment. In his early teens, his first obsession was chess, and he was good enough to become a tournament player. He was surrounded by middle-class children, with coaches and easy transport to matches, which his family couldn’t match. His, Dad, hoping to support him, bought him a chess engine, a very early, very primitive chess AI.

“It was unrelenting, it never got bored of playing me, it never questioned me, so my skill grew exponentially,” says Yeff. “20 years ago, I had a very early human/AI partnership, that I embraced. 15 years later, in the world of voice and beatbox culture, when I had the opportunity to do research, I remember that sentiment.”

If AI helped him do that with his chess skills, could the same be done with his vocal skills? To find out, he’s been creating a “second self” — AIs trained on his own work, that he can collaborate with. He’s devoted 20,000 hours of his life to his skills, developing a lexicon of phrases and sounds, yet he was seeing new patterns and phraseologies that were emerging from the AI — not from him, but somehow of him, without being him.

“There was something in that latent space that was very, very fascinating to me, as a world-class hyper-specialist expert on something.”

The second self as creative catalyst

Yeff’s experience was that, as he reached the pinnacle of his skill set, there was less challenge, less to learn from and draw from. He loved hearing and being challenged with these new ideas coming from his second self. “That was the ‘ah-ha’ moment, when I realised I was being challenged in new ways,” he says.

“A lot of creatives are afraid,” says Stannow, “because we’ve seen AI going from helping us analyse to truly helping us create. What would you advise designers and photographers to do?”

Yeff responds to that by dividing human/AI partnerships into three categories:

  1. Mentorship
  2. Collaboration
  3. Opposition

Most people are familiar with the first two emerging from technology. “Even a pocket calculator can do things humans can’t,” he says. “So, we accept that. The difference is now that these AI systems can respond directly from your contribution. The scary part is the opposition, when the system feels like it will take away opportunities from you. You need to seek out that opposition.“

Why? Because it will force you to grow as a creative. Again, he reiterates that hearing AIs trained on his own vocal skills exposed him to new ideas, causing discomfort and growth. “Creatively, are you taking full advantage of how these systems can push your limits?”

AI as a mutative force

“There’s been examples where systems have generated ideas that change how humans approach expertise. It’s happening in game theory all the time. It’s really important because it allows a framework of where these systems might go. These systems in Chess or Go are better than the world’s best players. But all humans have become augmented — they’re able to play in a different way.”

There’s even a name for these AI-augmented human players: centaurs.

Yeff cites an example from Go, when an AI made a move that the human world would consider bad, but which its opponent described as “beautiful” because it changed his perspective on the game. This, Yeff suggests, is when AI moves from generative to mutative. It changes how we play.

“There are occasionally these moments when something truly new happens, and that’s going to happen more and more. There are so many opportunities — if we embrace it.”

Second selves beyond the creative realm

Will this also apply outside the arts, to the creative industries or even the personal realm, asks Stannow? Outside the realm of art, these second selves might manifest more as “agents”, that will generate insight through mentorship, collaboration, and opposition.

“What gets in the way is the narrative and stories we have around AI, mainly from science fiction,” Yeff says. “But AIs allow us to pull information out of the latent space to improve your craft, whatever it is.”

Yeff is excited by the potential of AI to bring mentorship and opposition to parts of the country, or the world, that, through poverty or insecurity, had less access to technology or opportunity. “It’s really, really important to me that emerging tech isn’t just for institutions, but for individuals to get new opportunities.”

Authenticity in an AI world

So, it could increase social mobility and access to education, Stannow agrees. But what happens to authenticity? If the technology can replicate what exists, what happens next?

“Well, the word here is ‘regulation’ and that brings with it ‘consent’,” says Yeff.

Most people don’t have access to models trained on just their own data. It’s just these big systems trained on massive data sets—often unethically. He’s pleased to see work going into ethical data sets, but he thinks people should collect their own data so that it can be used. And that applies to brands and studios as much as to individuals.

He sees a future in which the systems that generate new work are traceable, much as food sourcing and origin are increasingly monitored. Will it get to the point when “human-made” is a marketing label?

Countering AI angst

And yet, for all Yeff’s excitement, fear persists. But how do we deal with and overcome it? Stannow asks if we need to work with that inherent angst.

“I think it’s going to change jobs, how people work, more than take away jobs, across the board,” says Yeff. “When it comes to creation, discipline or expertise, growth comes from discomfort. Life can bring you discomfort, or you can walk in and own that discomfort.”

What do we need to get right to make sure that happens? We need more AI literacy, and the path to that is more narratives and stories around the use of the technology, Yeff suggests.

“Education is the most important thing, and then rewriting the narratives around the technology,” he says. “If you’re worried or insecure, you can’t see things clearly.”

This post is based on the conversation between Harry Yeff, artist, musician and technologist, and Mette Stannow, Managing Director, Accenture Song, on the NEXT–Show in November 2023.