We are building AI 2.0
Q2X Labs develops independent research in the the United Kingdom. Early results from our R&D are strong, and we're now raising to take the work from lab to scale.
92% less compute.
Infinite context.
Transparency.
Progress in AI has outpaced the systems built to make it efficient, trustworthy, and accessible.
Most gains in model capability today come from throwing more compute at the problem. That approach requires huge investment in data centres and energy costs.
Q2X Labs is developing a new class of AI model built around a different set of trade-offs. Our early research results suggest this approach is not just theoretically sound, but practically achievable and we're now moving from lab validation to a funded programme of work.
A different approach to building AI models.
Using our new core approach which we have termed SNE 'Spherical Naigation Engine' we are able to train a model using 92% less compute than traditional transformer approaches. The SNE approach also translates the saving to inference compute costs.
Spherical Navigation Engine
The core of this technology.
Concepts not language
The technology uses 'concepts' not 'human language' as its base.
Knowledge vs Prediction
We believe it is better to 'know' than 'predict'.
The initial numbers are strong.
A summary of what we can share publicly today.
The concept works, we are now looking to prove this at scale.
If you are interested in getting involved in the next stage of this project please get in contact.