Why Ensemble is backing Loti AI’s mission to become the safety infrastructure for generative AI

Ensemble is excited to announce our investment in Loti AI, a platform that protects the integrity and authenticity of online identities. The company applies advanced facial and voice recognition capabilities to monitor the web for deepfakes and other inappropriate uses of likenesses online.

• Loti AI analyzes real photos, videos, and audio from its customers, then searches for potential deepfakes or other unauthorized content, such as impersonation accounts, and automatically removes unwanted content using an AI agent performing DMCA takedowns.
• ‍Loti also ensures that likenesses are used according to contract, providing a robust solution for protecting personal and professional image rights. Given rapidly evolving generative AI capabilities, Loti’s mission is incredibly timely.

Ensemble sat down with CEO/Co-Founder, Luke Arrigoni, to hear the story of how a Columbia math student bucked the trend and headed for Hollywood, where he went on to build an AI company that is now key to individuals and brands safely harnessing the full potential of generative AI. We learned how Loti AI’s effort to defend a measure of the truth online starts with intellectual honesty and a ground-up approach to problem-solving.

“If I wanted to be on a computer, I had to build a computer. If you have something you desire, it will not be handed to you in a place like Fresno. You’ve got to go make that happen.”

Fresno, California isn’t known for its tech scene – in fact, spend enough time in Silicon Valley and you’ll probably hear the historic Central Valley town alluded to as an example of everything that the state’s glittering coastal innovation hubs are not. But in a town where “tech for its own sake” is hard to come by, you might say Luke Arrigoni’s interest in computers developed the old-fashioned way: necessity was the mother of invention.

“If I wanted to be on a computer, I had to build a computer,” Luke told Ensemble. “If you have something you desire, it will not be handed to you in a place like Fresno. You’ve got to go make that happen.”

From California to New York – and back

Luke describes his early interests as “shoving pieces together and making it work,” and he evidently had success putting things together – after high school, Luke landed a spot in the math program at Columbia, where he became convinced he would follow the conventional path into investment banking or stock trading (“I had a dumb dream.”). Instead, he fell into an opportunity to bring a quant approach to an unlikely industry – talent agencies. Luke moved back to the West Coast, joining Creative Artist Agency (CAA), one of LA’s most influential agencies, in a role he describes as “entertainment quant.”

“I was trying to figure out how much movies would make in the first eight weeks of box office, or the most efficient path to tour a musician across the country, at certain venues, at certain times, with other acts.”

The entertainment quant expands his range

Applying deep mathematical solutions in Hollywood was off the beaten path for an Ivy League math major. But for Luke, whose problem-solving approach had been cultivated through a childhood of finding ad hoc solutions for practical, hands-on problems, it was back to basics –  with a whole new toolkit.

“Math is a tool. I'm not a mathematician. I'm not really a computer scientist either. But people can put me in a room with their data and I can figure out the line of best fit, essentially.”

With success at CAA leading to a growing reputation and burgeoning career as a “hired gun” data pro, Luke and a small team branched out into an increasingly wide scope of one-off consulting projects. As Luke found out, it isn’t just hedge funds that need deep, bespoke mathematical applications for data sets that can’t be made sense of by off-the-shelf software. Companies across the economy generate large proprietary data sets but don’t necessarily have the internal capabilities to draw useful conclusions.

“Math is a tool. I'm not a mathematician. I'm not really a computer scientist either. But people can put me in a room with their data and I can figure out the line of best fit, essentially.”

“We had a good rep for being the Spartan team that would just be airdropped into a data science team. They had all kinds of weird data, and they'd say. ‘The essential part is we want to know if we can make money or save money.’”

Discovering a product for the mass market

So, when an intelligence agency approached Luke with an issue that called for advanced facial recognition, he took them up on it. In order to track and monitor specific individuals across a wider scope of footage and photos, the intelligence agency asked Luke to develop a system that could identify faces even when obscured by sunglasses or masks.

“It was back to, ‘I have a problem and I know that I'm the only one. So I'm just going to build it. I have to build it.’”

This time, though, Luke wondered if there might not be wider interest in the product of this one-off software project. Curious to apply his new model to deeper corners of the internet, he got connected with a woman who wanted to take down a widely circulating series of photos that had been taken without her consent.

“We did our first takedown and she said, ‘My life is back.’”

So Loti AI was born.

Keeping up with AI

The frontier of possibilities for misuse of likenesses online is expanding exponentially in the age of generative AI. From now on, it’s not just duplicates of licensed images that agencies and celebrities have to contend with; deepfakes will only increase in sophistication and blur the line between truth and fiction.

In just the last few months, computer-generated likenesses have emerged as the headline legal battleground of the AI era. From voice actors identifying their own voices used in ads without compensation, to a public spat between OpenAI and Scarlett Johansson, the ease with which unauthorized likenesses are generated calls for an effective watchdog.

Beyond clearly unauthorized use, the jury is still out on what constitutes innovation vs. infringement. In a move that split the entertainment industry and its consumers, Josh Kushner’s Thrive Capital (which owns a large interest in OpenAI) recently committed $75M to A24, the independent studio known for the use of cutting-edge technology. Just last year, an extended writers' strike in Hollywood largely hinged on negotiations concerning the use of AI-generated scripts and likenesses of real actors.

Loti AI’s mandate is to monitor the front lines of that battle – so how will they contend?

Loti AI’s approach actually makes the advancement of deepfake capabilities an advantage rather than a moving target. In layman’s terms, Loti AI creates a mathematical model of a given likeness and compares any image on the internet to that model. Within a certain threshold of similarity, the two faces are considered the same. As deepfakes continue to look more and more like a given Loti AI customer, those images will be easier to catch.

“As technology gets better, our systems don't actually need to get better and better at the actual identification. As tech makes it look more realistic, we will actually have to deviate farther and farther away from our reliance on deepfake strategies, because they're basically going to be making 35-millimeter film pictures of humans.”

Intellectual honesty as a foundation for long-term solutions

“Ensemble stood out from all the other investors. I would say the biggest compliment I can give is we feel like we can extend that level of transparency and analysis with Ensemble, and they would agree to reciprocate.”

Loti AI is the product of professional problem solvers armed with math, data science, and computer science tools. Their product is positioned to redefine internet safety and serve as the primary tool for any public figure or talent agency to monitor contract compliance for likenesses.

While generative AI’s impact on Hollywood is still evolving, it looks like the entertainment industry has picked its horse. In fact, Loti AI recently announced a partnership with WME to share the company’s capabilities more broadly.

Luke says that a long track record of addressing practical issues from the very root of the problem has instilled the intellectual honesty necessary to build such a platform.

“An important Loti ethos is when someone on the team makes a mistake, they acknowledge it and then share those lessons with the rest of the team. That intellectual honesty makes us all better and Loti better.”

Luke told us that sensing Ensemble’s honesty in problem-solving was a major factor in their decision to work with Ensemble.

“Ensemble stood out from all the other investors. I would say the biggest compliment I can give is we feel like we can extend that level of transparency and analysis with Ensemble, and they would agree to reciprocate.”

“I know Ensemble calls it a founder-friendly environment, but really it's a problem-solving environment. ‘Founder-friendly’ means you can call someone with a problem and they'll help you with it – it's not a crisis. It’s ‘We believe in you, so you'll figure it out. This is the thing I think you should do to figure it out, and here’s the slice we can help with.’”