Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
AI on the Lot, the tech conference focused entirely on AI and filmmakingtook place at Amazon MGM Studios in Culver City last week. I was there for part of the sold-out event among nearly 2,500 people, who were all milling about from soundstages to theaters to get educated on everything artificial intelligence can do for Hollywood.
Human creativity, intent, identity and impact were not front and center in the panel conversations I attended. I made it a point to make these topics a focal point of my sidebar chats with various AI representatives. The most eye-opening of these interviews was with Luke Arrigoni, CEO of Loti AI — a company that scours the internet for deepfakes of people (celebrities and otherwise) and takes them down.
Do you want to protect your likeness online? Luke is the guy to call.
A deepfake is basically a digital forgery of someone’s likeness, whether it’s the use of their face, voice or the whole package, to make content, like pictures, videos and audio recordings.
The Screen Actors Guild-American Federation of Television and Radio Artists, the performers’ union of which I’m a member, went on strike in 2024. The threat of generative AI using a performer’s likeness without consent or compensation was — and still is, if I’m being honest — an existential concern.
Deepfakes don’t just affect performers and celebrities, though. AI tools on social media are fertile ground for such, whether it’s for spreading misinformation on TikTok or for the nonconsensual nudifying of images of peopleas happened earlier this year with Elon Musk’s Grok AI tool.
Arrigoni (who was featured in Forbes’s Next 1000 list in 2021) met me outside the Culver Theater. Sporting a crisp blazer and casual button-down, he had a welcoming yet no-nonsense demeanor that made me hopeful our conversation would disrupt the monotony of the event’s AI worship. It did.
We ducked away from the conference hustle and bustle to talk about everything from the boilerplate details of his company to the fear surrounding AI and the guardrails needed to keep it from destroying humanity — his words, not mine. During a convention filled with techno-optimism about a technology many fear and refuse to understand, Arrigoni cut through the noise to get to what really matters.
The following interview was edited for length and clarity.
What does your company specifically do?
Arrigoni: We do two big things. The first thing is we do a lot of defense: We scrape the whole internet, bring it into our system, use face and voice recognition to find you out there where you don’t belong, and then we delete it. Think of it like search, find and destroy. It’s unauthorized deepfakes, even explicit material, scams, people photoshopping logos onto T-shirts — that kind of thing. We’re really good at the finding, and we’re even better at the taking down part.
The second big thing we do is offense. If any gen AI company wants to generate you, there are these rules and pricing that you get to control as a creative. If they’d like to use you or your work, they send it to us, and in real time, we’ll clear it, say, “Oh yeah, you can do that, you can generate that,” or we say, “Don’t waste your money on generation. If you generate that, I’m going to delete that.”
So then your product isn’t necessarily AI-specific, but it’s more so than that. It sounds more like one of the many guardrails that are necessary in this era to potentially protect people.
Arrigoni: It’s fighting fire with fire. We use face recognition and AI models to better understand the content and decide whether it should be removed or approved. Realistically, what we are is a human tool. You mentioned guardrails; we’re a rail system. We’re the fastest way for people to be generated in a safe way. We’re the easiest way for gen AI platforms to be both compliant and participate in a bigger market.
Hypothetical: I’m an actor and found a video online that claims to be me but isn’t. Do I need to go through law enforcement and get a subpoena to have your company handle the complaint?
Arrigoni: You don’t need any of that. You sign up, and within 10 minutes, we’ll be searching the internet for you.
What prompted you to start doing this?
Arrigoni: We actually started doing revenge porn takedowns first. That’s a horrible term for nonconsensual intimate images, but most people know it as its other name. And then we started doing that everywhere, and then SAG-AFTRA went on strike. I used to work at CAA [Creative Artists Agency]so I called them and William Morris Endeavor [WME] and other [talent] agencies and said, “Would your clients like a service where we can actually take care of this on a systemic level, where I can ingest large swaths of the internet, and then just take this down?” And everyone said, “That sounds like a no-brainer. Let’s go and do that.”
At the end of the day, do you feel hope or dread for what’s to come with AI?
Arrigoni: I think there’s going to be a ton of AI-generated everything; that’s already starting. People will lose their appetite for that, and then we’ll be left with real tastemakers and real people creating art, journalism or other things that AI cannot replicate. I have no faith that AI will suddenly be better than humans at generating top-tier content; there’s a reason for that. The reality is that AI is structured around the idea that if you give it a ton of data, it can try to predict the average of it, and all the data everyone’s looking at now is somewhere between the 50th and 75th percentile in terms of quality because the bottom quartile is illiterate and the top quartile really doesn’t publish much. So you have a system that wants to converge on just above average, but it has no mechanism to converge on the 99th percentile.
Speak to me like I’m 5, please.
Arrigoni: OK. Depending on what your appetite is, whatever kind of cool stuff you like, whether it’s a Christopher Nolan or a Jordan Peele type of movie — something that is weird and out there with its own cool aesthetic — that’s not going to be replaced by AI anytime soon. But if you’re making common content, there’s going to be real competition.
Do you see us getting to a point where deepfakes will be indistinguishable from the real thing?
Arrigoni: I think probably in the next year or two, it’ll be impossible to distinguish what is fake or not.
What does that say about how AI technology is affecting the general public regarding how we interact with media?
Arrigoni: This is an education problem, and it’s going to take at least a single generation to overcome it. But how do you get the cost of critical thinking to be lower so it doesn’t cost the average American so much mental power to think critically? I mean, you have to start in kindergarten, right?
This brings me back to that hope question because this all feels a bit bleak and dystopian.
Arrigoni: I’m very optimistic about what I think I can do. It doesn’t mean I’m optimistic about the current state of things, but I definitely feel like when there’s a larger problem in our ecosystem, I can figure out how to wield and bend the tools in my network to try and fix it. I feel the right amount of fear, but I’m also a little excited about what I think I could do to help.
Arrigoni: This is what tech likes to do: Tech likes to optimize over and over and over until it gets to the point where it removes either humanity, caution or guardrails. A lot of the recommendation algorithms are point-blank an exercise in the most extreme versions of technology. So what we’re doing at Loti is making sure we don’t actually take that last step, even though the market wants us to say, “What’s the next step?” I’m fighting it, and I actually have good investors that are aligned with that thinking, but I think more people need to start basically saying, “I know I can take the next step.” It will make things more profitable, but at the rate things are going, it’ll actually destroy humanity.
Wait, what did you just say?
Arrigoni: That sounds super extreme, right?
Yes, but people are also going to AI for all sorts of things they shouldn’t: ChatGPT is a doctor for people, helps students cheat on homework, etc. What do I know?
Arrigoni: I have a 15-year-old, and I made it really easy for him. I just told him, look, you can cheat, right? I know there’s no way I could stop you, or anyone could. If I were to tell you that your entire life would be easier if you put in the work when all of your peers will not, and suddenly everything — like being able to attract partners because you’re just more literate, being able to get jobs because you’re more knowledgeable, being able to have conversations with people and connect — all of that will be easier for you if you just do a hard thing right now. So, you can avoid it if you want. But I promise that putting in some work now will actually make your life easier. I don’t know how you teach that at a scale for everyone to understand it. But I think parents probably have to do with the OG guardrails of humanity — just like having a kid and making sure they don’t end up miserable, so yeah, I guess that’s probably one place you could start.
I’ll end this chat with the topic of the singularity — as one does. Do you think AI will be the catalyst that takes us to that tipping point where technology surpasses humanity?
Arrigoni: If it happens with the tech we have now, it’s going to suck. It would be like the most average person you’ve ever met suddenly is just everywhere. And that would suck.
