
I hope you showed up hungry, cause we’re about to dig into a word salad dressed in hubris and seasoned with delusion. Dutch-born, UK-based Eline Van der Velden dropped an existential bomb two months ago when she announced her AI company Particle6 had created an AI “actress” who was about to be signed by an agency. The creation’s name was Tilly Norwood, and, predictably, she looked like a teenager. Van der Velden has largely been mum in the intervening months, and it turns out that was intentional; backlash to Tilly was so strong she had to call police over death threats. While I think Tilly is an abomination, death threats are never the solution. But now Van der Velden is talking — at length — with a verbose interview in The Hollywood Reporter. I’ll be kind: the article is a fascinating expression of the human mind. I appreciate that the THR reporter asked point blank why Tilly was a teenager, to which Van der Velden responded “She isn’t a teenager. She’s 24!” Sure, Van der Jan. Some highlights:
Creating Frankentilly: I played around for ages with text prompts… We would up res, we changed the skin, we changed the freckles, we changed her hair. I think we did about 2000 iterations until we finally had one image that we were like, “That’s it. That’s her. That’s who we want.” We wanted her to resonate with a specific drama audience. We wanted her to resonate around the world and not look too specific. [We wanted her to] represent Britain and the mix that it is. … It’s a long journey and a lot of human input and a lot of human work.
Tilly isn’t made from stolen material! Just to explain about the training and copyright thing, because people think [Tilly’s] based on people’s likeness and it’s trained on people. But then they should be compensated, right? So I think those third party tools need to pay for the data that they’re training on. However, what worries me is that the actual creators and the people, because it’s trained on the whole of humanity, billions [of people] — you, me, everyone’s been put into that training data — we won’t see any of that money, right? … What happens after that training is you are working with text in order to position pixels in a probabilistic position on a blank canvas. Unless your prompt says, “Make her like this person,” it’s not pulling that data from the data set.
People aren’t losing jobs because of AI… I think in our industry, people are losing a lot of jobs because of the state the industry is in. I don’t think AI is necessarily replacing those jobs. … It’s not got to the point yet where they’re saying, “Oh, we’re going to fire all these people in TV because of AI.” So I think that’s not there yet but it will happen, which is why I’m speaking out. Because I’m here to help people through this transition, this tough period, because it is going to be a transition.
…but all the jobs will soon be AI: There are lots of jobs if you become good at AI. So we are struggling to find people with the right skill set. And I think it’s that … we need to all upskill, to go with the new wave of tech that we’re going in and not get left behind. And I’m here to help. We’re training up lots of people in that, but we are using humans all along the way. There’s lots of jobs, right? We’re talking AI directors, AI DPs that are deciding the shots, AI production coordinators, AI production managers. There’s so much humanity that’s needed, otherwise it’s crap.
Her response to SAG-AFTRA’s denouncement of Tilly: I agree with SAG-AFTRA that she’s a computer-generated character, whether she can emote or not like a human can. So I would also say, yeah, we’re not expecting her to right now. And I want to watch real actors emote — I enjoy that very much. I’m an actor myself. I love the art, but I do want to prepare people that I think the tech will get there with good direction. And we’ve done that.
The mental gymnastics at play here are, like I said earlier, fascinating. The central cognitive dissonance of Van der Velden being an actor herself — she studied alongside Lily James and Daisy Ridley! — yet not seeing the problem with Tilly’s “existence,” is quite something. But that’s her personal dilemma. What I find more concerning is the crazy argument she gives for why Tilly isn’t built on stolen material. She’s basically pawning off the responsibility to those “third party” programs, the ones that technically do the stealing training. Then turns around and says, “But we didn’t do the training! And we didn’t literally say ‘copy so-and-so actress,’ therefore it’s not theft!” And the biggest fail, logic-wise, were her comments on AI jobs not taking human jobs but humans have to “upskill” to AI jobs if they want to have jobs in the future and we need humans to build the AI directors and cinematographers of tomorrow but AI is not taking human jobs! Then topping it all off with “I’m here to help!” Do us a favor and maybe help less?










She’s a failed actor living vicariously through her creepy Ai creation. It’s so pathetic and her acting like it’s not a serious thing is just as pathetic. Not to mention creating an Ai character who looks like the this one, she intentionally made her with the male gaze in mind. Truly gross all around.
I’m gobsmacked. What is real anymore when you can’t tell a photographic image from a genuine one? This trend is dangerous, in my opinion.
The destruction of our environment – especially in the US as half of all data centers built and planned worldwide are in the US – because tech bros have pushed the notion that we need to use AI for utter tripe like this is so very demoralizing. [full disclosure: I’m fighting development of a huge data center < .5 mil from my house and just found out there are two more planned in the next 2 townships over so there will be multiple large data centers using the same resources, polluting the air and harming local wildlife, beneficial insects, farm animals and domesticated pets in an overlapping 20-30 mile radius so this completely unnecessary use of this tech is infuriating0
Nothing to add, I agree it’s an abomination.