This week on the EU-Startups Podcast, we sit down with Zuzanna Stamirowska, CEO and Co-founder of Pathway, the corporate behind the world’s first post-transformer AI mannequin.
Pathway is creating an structure designed to adapt and suppose extra like people, shifting past the restrictions of conventional transformers. To this point, they’ve raised over €15 million, providing native reminiscence, temporal consciousness, and full visibility into how the mannequin works, combining these capabilities with one of many quickest data-processing engines in the marketplace. Trusted by organisations equivalent to NATO, La Poste, and Formulation 1 racing groups, Pathway is aiming to convey enterprises a brand new degree of contextualised, experience-driven intelligence.
On this episode, we discuss why AI wants new foundations, how Pathway updates reasoning as new proof arrives, and what “glass-box AI” appears to be like like in observe. We additionally contact on long-range context, knowledge effectivity, the patterns they see throughout NATO, La Poste, and F1, what their high rating on Hugging Face unlocked, the potential Zuzanna is most enthusiastic about for the subsequent 12 months, and way more!
Video model of episode 146:
Audio model of episode 146:
Key Takeaways:
- AI’s subsequent leap is architectural, not incremental.
- Reminiscence and time unlock adaptation transformers can’t contact.
- Context beats parameter rely.
- Glass-box visibility will turn out to be an enterprise requirement.
- Much less knowledge, extra reasoning: the long run is effectivity over scale.
