US companies like Open AI or Google They are in no way inferior in terms of technology. But the advantage they have is capital, says Just: “If we want to train similarly large models, we would need more than 500 million euros need. That’s unrealistic.”
“More efficient than competing models”
research results have shown that xLSTM is significantly more efficient in computer usage than competing models, says Just. This is due to its own memory, which helps the model to understand and classify data more quickly. It also has advantages when solving mathematical problems, as data is also recorded in its chronological sequence. The quality of the answers is at least as good, if not better.
First use cases
In a next step, the research results will be confirmed with much larger amounts of data. Just assumes that the advantages will increase. In parallel, the start-up is working on initial use cases. “It is important for us to find a place in the market that is defensible,” says Just. Initially, the company wants to focus on applications in the Industry focus.
As an example, he cites the management of Energy systemsWith numerous solar power producers, more and more electricity is being fed into the grid from more and more producers. The systems are not really prepared for this. Real-time applications based on artificial intelligence are able to calculate energy flows and adapt these calculations to different times of day, seasons and weather conditions.
“We see that we can produce results that no other model can at the moment,” says the NXAI CEO. In 3 to 6 months You want to be on the market with the first applications.
Prominent investor
NXAI is financed by the industrialist Stefan Pierer and the Linz-based digital company Netural X, in which the KTM boss is also involved.
Just does not want to reveal how much money has flowed into the start-up so far. However, thanks to the powerful investors, the company is fully financed until the end of next year. “The biggest cost we have is computing power,” says Just. The infrastructure alone costs at least tens of millions per year.
Regulation restricts
The EU regulation of artificial intelligence with the AI Actwhich came into force on Thursday, hit NXAI hard. “As a company that operates in the field of foundation models and basic research, we are among the most heavily regulated,” he says. It also prevents the company from thinking in the short term about end-customer applications. “To do that, we would have to validate every text that we use in training,” says Just. “That is also hardly feasible at the moment.”
