Cambridge startup Tenyks combats AI tech ‘terminator’
Celebrated Cambridge academic Stephen Hawking warned shortly before his death that if it wasn’t managed carefully Artificial Intelligence could spell the end of the human race.
While even hardened critics of DeepTech would not be so apocalyptic in their sentiments, Professor Hawking’s warning hit home and helped to temper gung-ho approaches to developing AI propositions.
Take a bow Cambridge startup Tenyks, founded by Cambridge University PhDs Botty Dimanov and Dmitry Kazhdan.
The company has just emerged from stealth with technology designed to help developers become more measured in creating AI-inspired software and ‘protect the world from the misuse of AI.’
Protecting the AI world from the tech ‘terminator’ is no bad selling point. The startup raised a modest $125k pre-seed funding from Y Combinator in August and previously collected non-equity assistance from Creative Destruction Lab. It is now set to fast-track growth from its home at Eagle Labs in Chesterton Road.
Tenyks is building an MLOps monitoring and validation platform that helps AI developers working with computer vision data to build more reliable software faster.Specifically, the platform helps Machine Learning developers understand soonest what’s wrong with any software and fix it. Its website says: “We started Tenyks to invent the way humanity interacts with AI to protect and delight – to protect the world from the misuse of AI but also to ensure that it is developed with passion, excitement, and joy.
“We set our goals ridiculously high and stick together to go further than anything previously imaginable. We start small, work hard and deliver fast, embracing the inevitable obstacles with open hearts because challenges fuel our burning desire for learning.
“Tenyksians make no distinctions between work and play. We simply pursue our vision of excellence, leaving others to decide whether we are working or playing.”
The net gain, according to the founders, is much faster prototyping, reducing the time it takes engineers to understand why a model is behaving strangely.
The technology is set to improve dataset quality and help engineers understand, visualise and curate dataset to uncover corrupted points, identify and fix incorrect labels and identify difficult edge cases.
The founders say their solution pinpoints edge cases and common failure patterns, removes spurious correlations, reduces unexpected failures by 80 per cent and boosts performance by five per cent.
For example, a medical imaging ML developer identified that 95 per cent of all predictions were invalid because the output depended on the wrong input signal.
Tenyks is supported by the University of Cambridge, the Impulse initiative at the Maxwell Centre and the Creative Destruction Lab.