These are the first “start-ups” in the AI ​​era: “A few years ago, there were only two of us who couldn’t set this up” | Technology

When Sam Altman, co-founder of OpenAI, was in Madrid in May last year, he uttered a phrase that seemed exaggerated: “Thanks to artificial intelligence, there will be billion-dollar companies managed by a single person. » His prophecy has not yet come true, but there are already mini start-up who see this future. For example, in Murcia.

Javi López and Emilio Nicolás launched on November 28 MagnificAI, an artificial intelligence (AI) application which makes it possible to improve resolution and retouch images in a way and at a speed difficult to imagine until now. In just a month and a half, they have 400,000 registered users, a small percentage of whom pay for three subscription options ranging from 37 to 275 euros. They got comments about X from Elon Musk and praise from digital artist Beeple, author of one of the most expensive digital works in history.

This story could be someone else’s to start up, except that’s not the case. “A few years ago, just two of us wouldn’t have been able to put something like this together,” says López, 42, in a video call from Murcia, where they are based in a coworking local. “Ahora hay a series of winds that you have of artificial intelligence, that mark an abismal difference and that cause you to go to the speed of the light, there is the world that is interconnected, that anyone can go viral from night to night Morning”.

López and Nicolás came up with three founding mantras that were not easy to maintain a few years ago: “Zero equipment, zero investment and don’t give away the product,” says López. “How long are we going to want to extend this over time? We will extend the investment for as long as we can, that is priming (business term for starting with few resources) mortality level. Team? Sooner or later this will be necessary. But great teams burn out a lot. There are many people facing many different types of problems and you take them home with you. However, being in front of a computer with GPT-4 is more bearable,” he explains.

This first virality is also linked to the interest generated by generative AI. Its current success lies not only in using tools that did not exist recently, but also in offering new services to many people who want to explore this technology. López explains: “Customers come from all over the world, only 5% are Spanish, and this shows very well in the revenues. » The businessman adds that in addition, it helped them “to have all the tailwind that AI represents and, within AI, generative AI, which not only has a tailwind, is a hurricane that animates you. You unfurl the sails and fly away.

López explains how he felt when Elon Musk responded to him about what appeared in his chronology and he made this meme so famous on the internet, which I think comes from the days of blade runner. It also pushed us a little more. Beeple also wrote to him privately that he used the tool daily: “I think he uses it in all his work now, because I have already trained my eye to identify if something was done with Magnific, I’m already seeing some patterns.” He can’t give many details about the relationship with Hollywood, but they showed him the usefulness of it in a film made by “a super famous actor and director.” hypermega famous,” he said.

López already founded Erasmusu in 2008, a community online of international students, which he sold a few years later and which he ended up leaving in 2021. He then began a sabbatical season, with his travels and other entertainment. Until Dall-E 2, the second version of the OpenAI tool that creates images from text, was released in 2022. He saw that it was promising and he devoted the free hours he had to it. He posted his experiments and games with AI on his Twitter accounts in Spanish and English, where he developed a community related to generative AI. “I was having a great time, I would have even paid to do what I was doing, which was enjoying, experiencing and playing,” he says. It was through this kind of motivated relaxation that the idea for MagnificAI was born: “If I hadn’t had time to play, I wouldn’t have found the key. »

In the summer of 2023, he talked about his idea with Nicolas. But they didn’t start working in earnest until October 2023. In just over a month, they had a product. They didn’t go in with a lot of pretensions: “In our heads, it was going to be a temporary freelance job, a toy, a personal project, but it blew up, we didn’t expect it. I think it’s really easy for many, with today’s technologies, to put them together. They’re like Lego pieces and all of a sudden you’re like, ‘what a cool castle I made,'” he says.

The feeling of being in a new era may be hasty, but López has few doubts: “The impression of recent years is that it feels like the year 2000 again, but instead of being 18 , in my case, and not having participated in it, I am now 42 years old and I had to live it, fortunately, with a lot of free time,” he explains.

MagnificAI uses the open stable diffusion model. Through testing and tweaking, López found that if he paired things together, changed settings, or tweaked the source code, something came out that could become the seed of a new product. “We can’t comment much on these details, it’s a secret,” he adds.

Its main users are people who already work with images, from film production to illustrations or design, and who need it to improve the resolution of their final projects. López explains: “The majority use it to last mile. Imagine someone who comes from interior design and has a give back and you don’t want to spend hours with lighting or textures. Then MagnificAI overtakes him and leaves it to him nickel plated“. The idea is sufficiently innovative to require a new name: “It is a new sector of activity which is the rise in power (resolution improvement) re-imagining,” he says. AI helps “reimagine” the original image and enhance it with specific text queries.

Many of those who contact them have ideas on how they could continue to grow. But it’s not easy. For the moment, they still have work to be able to apply MagnificAI to video or video games: “Some people tell us that if we had this for video, we would be worth a billion. But video is much more complex. Or that if we had it for real-time video games as the last layer, we’d be worth 2 billion. I tell them that’s all, and that if I had a flying car, it would have antigravity,” he jokes.

Just as AI has been a big driver for López and Nicolás’ product, it is also a big driver for their hypothetical competitors: “We have competition, although I don’t think they do yet to our level. On the other hand, we know that our defensive moats are not very wide, we are not panicking. We are neither scientists nor mathematicians. We understand that, sooner or later, only image enlargement has its days numbered,” adds López.

They do not exclude the idea of ​​a sale and continue to work internally, as they did with Erasmusu: “If it is purchased on economically acceptable terms Half good and it’s with a company that values ​​us and wants us to continue working on what we’re doing, even in MagnificAI itself, of course. We don’t have the American mentality of expecting it to be worth $3 billion in review cycles and a team of 400 people,” he says.

López deep down believes that they will not be the only ones to have found this new path in the AI ​​era: “I think that in the next 10 years we will see start-up individual entrepreneurs or less than five people, which is almost the same, charging 1 billion; Not that these are unicorns valued at a billion, but they will actually manage to reach this amount.

You can follow EL PAÍS Technology In Facebook And X or sign up here to receive our weekly newsletter.

_