In the race towards AGI, alignment is one of the most important problems to solve for true acceptance in the enterprise. At Sparkolaxta Inc., We’re solving that problem.
Without alignment, who does your AI work for?
With a unique combination of academic experts and industry veterans (ex-Anthropic), we’re consolidating a lean team around this critical issue. With offices in Palo Alto and Brno, CZ — we’re staying small but thinking big.
We are not currently accepting investment at this time. Having gone through many rounds of VC and mergers & acquisitions, we understand how those events — while necessary — can slow the pace of scientific advancement.
Reinforcement learning (RHLF) is an expensive, over-wrought system that does not scale and ultimately does not represent human-like alignment. The RM rϕ is not the “true” human preference function! Sparkolaxta’s research revolves around so-called constitutional AI, specifically RLAIF. We believe our framework of SFT models in conjunction with per-organizational RLAIFs provide a unique and enterprise-ready way to tame this long-standing problem.
Stay tuned!