Founder member of The Horrors, synth polymath and passionate enthusiast for all things AI, Tom Furse spoke to us recently, and predicted how machine learning will re-define art in all its forms, he also admitted a growing hunger for innovation when it comes to music technology…

“Why go anywhere when you can go anywhere?” laughs Tom Furse, The Horrors’ inventive co-founder and creative journeyman tell us, when asked about his recent swing away from live performing. Aside from his continuing role as a The Horrors’ principal sonic architect, Tom’s individual exploits have brought two deep solo records, and a burgeoning passion for AI.

Using his own system to manufacture dense visuals, like those in his recent video for HAAi’s Baby We’re Ascending, as well as his ‘Relics’ series of generative art pieces. Furse explained to us how he foresees similar mind-blowing AI innovations eventually re-drawing the music technology landscape. But first, we asked Tom about his recent departure from performing live with The Horrors…

AMI: Last year you announced you weren’t going to be touring with The Horrors anymore, in your Instagram post you mentioned that you’re more of a ‘creator’ than a performer. Had that been an issue for a while for you, the need to write not marrying with being on the road?

Tom Furse: Yeah, it absolutely had been an issue for quite some time. Just because it’s such a different environment when you’re on the road. It’s great for all kinds of reasons, but if what you really want to do is make stuff, there’s pitfalls. It’s just really hard to find a quiet spot. I’d spend a lot of time with headphones on in noisy environments. I wasn’t always happy on tour. When Covid hit it was a bit of a lightbulb moment.

AMI: But of course, you are still a member of The Horrors, and you’re working on album six?

TF: Yeah, that’s slowly happening, we’re chipping away!

AMI: Your own career beyond The Horrors has been pretty varied, one of the things that has been interesting recently is your use of AI in your visual art and videos, not least the Baby We’re Ascending video you did with HAAi. AI seems like it’s quite a big area for you right now. How long have you been working with AI and what first attracted you to it?

TF: I’ve been using it for the last year. I just heard about it online, and I’d seen examples of [AI-generated visual art] and thought it was pretty cool. Then I heard an episode of the Interdependence podcast with Holly Herndon and Mat Dryhurst that really got me interested. I heard about an approach that married image synthesis with natural language input control. I really wanted to give it a go. So I did. It required the navigation of a virtual coding environment, it wasn’t like a nice easy user interface. It was a little bit tricky. But, I just started messing around with it and I haven’t stopped since.

After doing more of less the same thing with music for the past 15 years or so, this was like a much needed breath of fresh air – an entirely new medium that was much less explored than music. I think music is thirsty for new technology and sounds, and new places to go. I don’t think we’ve had that for quite a long time.

I’ve always loved exploring the element of surprise in music, with generative approaches. But, that’s more an illusion of surprise. This is much more strange and psychedelic. It’s quite something.

AMI: Have you used generative and AI-based approaches in music before?

Tom Furse: No, not really because they don’t really exist yet. I went looking for it, and everything I found was pretty wanting to be honest. The processes I’ve used before have been based on logic and maths systems. You can set quite complex patterns, but it’s in reality a very simple computer, ones-and-zeroes approach. It’s not really intelligent but it is fun. It’s what Brian Eno’s been doing for decades, he’s done some of the most interesting work in that area. But we need new stuff – we can’t be Brian Eno all the time!

AMI: Do you think that we’ve only seen the tip of the iceberg really with what AI can do, particularly in music production?

TF: Oh, God, yeah. It’s a hard thing to capture because I think historically what people have been trying to do is train AI on like MIDI files, but they don’t necessarily convey what is interesting or *good* about the songs. Or, the style of a song or a sound. It’s not terrible data but it’s quite raw data. It doesn’t really capture the essence of stuff, which is what the visual art-aimed AI approaches do.

With what Holly Herndon and Mat Dryhurst are up to, where they have captured the essence of Holly’s voice with Holly+ – there you can drag an audio file into it and a very good approximation of Holly’s voice will sing it. With the latest version you can’t tell the difference between her voice and the synthesised version. So, there we’re getting into some really interesting territory. The lines are going to get so blurred.

One of the best-selling vocal albums of 2026 or 2027 will probably be made not using a real vocalist. People will start making things like Chet Baker techno records. Whatever the maddest thing you can imagine will be possible. We’re seeing this already with image synthesis. People will be mashing up stuff. If you think about how postmodern culture is now, we’re all really primed for this.

I think that’s more how entertainment will go. Giving the type of AI technology that we’re currently seeing in visual art to musicians is going to be wild, beautiful, scary and psychedelic. I can’t wait. But, there’ll undoubtedly be a lot of pushback.

AMI: Then you’ve got the other side of the AI-paradigm; platforms like AIVA and Amper which can manufacture tailor-made soundtracks using AI, what are your thoughts on that side of things?

Tom Furse: I mean, I’ve heard them and I think it’s almost like the equivalent of Dall-E. When I first saw that I suspected that a certain kind of illustrator might not have as much work, and that might also be the case here. I feel that I have a slightly savage opinion on it. My heart says, maybe everyone needs to try a little harder – why do people settle for this kind of mediocrity? I don’t enjoy it, when I’m watching a film, if there’s just a really vanilla score there, I don’t enjoy making that kind of music when I’ve done library projects before, I just feel like we should be rewarding bolder experiments.

This is perhaps a wake-up call for everyone. If you’re worried then you’re really saying you’re as much of a skilled craftsman as the AI is, but then the art is somewhere lost in there.

Having done a few films, I do understand that it is a very cutthroat industry, and music is such a tiny consideration when people are budgeting. It’s really undervalued, so it doesn’t surprise me that people are seizing on the chance to undervalue it more. I think a better AI solution would be a system that enabled their samples to sound more realistically like a certain style…

AMI: Well, there are quite a few string sample libraries out there that are pretty indiscernible from the real thing…

TF: They’re great, and I do use string sample libraries but the lengths you have to go to to make them sound natural is often quite extreme. I’ve been recording some live strings recently and there really is such a huge difference. It sounds very convincing, but when you get the real articulation and expression from a real player, that hasn’t been beaten yet.

I think you have a lot more scope to do that if you’re synthesising out of nothing – rather than trying to construct recorded samples to play in this natural way. But that’s not there yet. I think that side of the industry will be able to realise a lot more, with a lot less. It’s going to be pretty wild.

AMI: Do you have a similar stance on synths, are you averse to soft synths and prefer the real deal?

Tom Furse: No, I’m synth-neutral. At first you could kind of tell there was a difference, but it’s harder to discern now. I use Arturia’s stuff quite a lot. There’s loads of great Max for Live developers doing interesting synths, there’s so much good stuff happening. I do think we’ve reached a little bit of a plateau in terms of synth methods and sounds.

When I first started buying synths around 2007/2008, there wasn’t really anyone making any good new analogue synths, but now you can get a Behringer TD-3 for under £100. The accessibility is there now, we’ve conquered that particular mountain but where can we go from here?

AMI: How many synths are in your studio right now, or is that a silly question?

TF: Like ten I think, I definitely used to have a lot more, but I’m slimming down a little bit. I’ve got an Arturia Polybrute, and that’s a really amazing workhorse. I realised there’s lots of things I didn’t need anymore. I’ve kind of gone for a more efficient set-up.

AMI: Your second solo album, Ecstatic Meditations was quite a blissful record. Do you intend to continue making music in that vein?

TF: At the moment, it’s more a question of time. I’m just quite busy with a lot of projects that are more visual based. I am doing a bit of music but it’s really nice to have a break for a little bit. I’ve got a piano downstairs and I’ll sit and play something, realise I’m on to something but then might not chase it – But I know that ideas are still there. I used to chase it relentlessly. When your identity is wrapped up in this niche of creativity, it’s good to have a rest and explore something else. When I do come back to make music, it feels more focused. Less like a job.

I also feel like I’m preparing myself for what will be an exciting new wave of music technology. I feel like this investment of time is a very positive thing right now.

Tom Furse: Well, I’m working on another video for HAAi right now, then I’m finishing up some artwork for Temples. I’ve got a record of mine I’ve mixed with Ghost Culture that’s sounding cool so I’m going to get to that when I have a bit more time. Yeah, my current situation is that I wake up in the morning and I think ‘Right, what am I going to make today?’ Things are pretty open.

Follow Tom’s artistic and musical adventures over at tomfurse.com