Main benefits of our technology
The traditional approach to ML
Standard procedure is often a struggle when you need to collect a dataset in advance, hire and manage a team to look after your AI models, evaluate their performance, and re-train regularly to adapt to an ever-changing environment.
Collect the dataset
Train the model
Evaluate the results
Re-train to optimize
We offer an alternative!
“Deep learning is an amazing technology and hugely useful in itself, but in my opinion it’s definitely not enough to solve AI, [not] by a long shot”
“My view is throw it all away and start again, I suspect that means getting rid of back-propagation.(…) We clearly don’t need all the labeled data.”
“We need systems that can handle changes and do continual learning, lifelong learning and so on (…) This is a long-standing goal for ML”
In ML we distinguish
the following types of learning:
Our Focus
Continual Learning
Use a pre-trained model, then adapt it in real-time with data coming through available sensors
Combine self-supervised, unsupervised, and reinforcement learning together to evaluate what and how to learn and how to drive perception and actions.
Read more
Supervised learning
Collect the dataset, provide correct answers, and optimize the model. But what if you don’t have the necessary data? Don’t have resources to label all of it? And a team to maintain it all the time?
Read more
Reinforcement learning
Define a target and measure the success rate to guide action. A very good way to direct AI behavior in games and simulations where the objective is clear (e.g. win the game, get the most points possible, get to the destination).
Read more
Unsupervised learning
Analyze the internal relations and similarities between available data samples, and learn to reconstruct them. Useful to some applications, but without any base to rely on – it is hard to use it successfully on its own.
Read more
Case studies
View all
