A little-known AI company is making one of the boldest claims in the field, and the wider industry isn't sure what to make of it.
According to MIT Technology Review, a Miami-based startup called Subquadratic emerged from stealth mode last month with a striking announcement: it says it has solved a mathematical bottleneck that has been holding back large language models for almost a decade. In other words, the company claims to have addressed a long-standing limitation in how today's AI systems are built.
The reception, so far, has been skeptical. MIT Technology Review reports that the details Subquadratic offered were thin, and that many people were unconvinced by the announcement. The publication notes, however, that the company has stood by its claim.
Because the source material is limited, much about Subquadratic's approach remains unclear, including exactly what the bottleneck is, how the company says it overcame it, and whether independent researchers can verify the result. Extraordinary claims in AI tend to draw scrutiny precisely because they are difficult to confirm without published methods or reproducible results.
Why it matters: if a genuine breakthrough in a core limitation of large language models holds up, it could influence how the next generation of AI is built — but as MIT Technology Review's coverage makes clear, that's a big 'if' that the broader community has yet to accept.