It started, as these things sometimes do, not in a lab or a data center, but on a bike ride through Fort Lauderdale. Somewhere along the way, Rolv Heggenhougen had a simple thought: what if computers are doing a lot of work they don’t actually need to do?
That question turned into ROLV LLC, a 10-month-old startup with a bold claim. Heggenhougen says he’s found a way to make AI systems run much faster without changing the hardware they run on. In some cases, he says it can be over 100 times faster. Now, researchers at the University of Miami are working to independently test those results.
Heggenhougen didn’t come from the typical AI background. He studied math and physics in Norway, then business at the University of Miami, and later law. He started out buying and selling computer parts across different countries, then built a software company that turned everyday emails into ads. That company went public.
Years later, with no deep experience in AI, he decided to take on a problem most people assumed was already solved.
“I thought there’s gotta be a way that you can speed up matrix math,” Heggenhougen told Refresh Miami.
At first, he didn’t even fully understand the space he was entering. “I made a lot of stupid mistakes because I didn’t know the marketplace,” he said. But he kept going, testing ideas over and over again.
Eventually, something clicked.
To understand what he built, just imagine you’re doing homework, and half the questions on the page are blank. Instead of skipping them, you’re forced to go through each one anyway, wasting time doing nothing.
That’s basically what many AI systems do today.
“They scan also the blank pages,” Heggenhougen said. “My solution skips scanning those pages.” By doing less unnecessary work, the system can move faster and use less energy.
That could have a big impact because running AI systems is expensive. Companies spend huge amounts of money on powerful computers and electricity. If you can make those systems even a little more efficient, the savings add up quickly.
“There’s a huge business case here,” he asserted.
Still, going up against major tech companies is no small task. When he first shared his idea with someone from NVIDIA, the reaction was skeptical.
“One guy sitting somewhere in the world doesn’t come up with this,” he remembered being told.
That didn’t stop Heggenhougen.
Today, the University of Miami’s Frost Institute is testing his technology across different types of computers to see if the results hold up. For Heggenhougen, that outside validation is key. “Otherwise, I’m the only one who knows that I’m not cheating.”
Even if the tech works, there’s another David versus Goliath-sized challenge: getting big companies to actually use it. Many prefer to stick with what they already know, even if a new approach could save money.
For now, Heggenhougen is building ROLV on his own, funding it himself and trying to get in front of the right people. His plan is either to sell the technology to a large company or license it out so others can use it.
It’s a big bet, built on a simple idea: That maybe the future of AI isn’t just about building bigger, more powerful machines.
Maybe it’s about teaching them to stop wasting time.
READ MORE IN REFRESH MIAMI:
I am a Miami-based technology researcher and writer with a passion for sharing stories about the South Florida tech ecosystem. I particularly enjoy learning about GovTech startups, cutting-edge applications of artificial intelligence, and innovators that leverage technology to transform society for the better. Always open for pitches via Twitter @rileywk or www.RileyKaminer.com.
Source link