We aim to build a system capable of understanding knowledge, to answer questions and get things done.

PROBLEMS WE WORK ON

Making use of humanity’s vast collective knowledge is
hard and the tools we have are not sufficient

Information is spread out and heterogeneous.
This makes it hard to access and make use of it.

Existing ways of managing information and software
require significant manual effort.

Many machine learning methods can not leverage our
understanding of the world which limits their
applications and robustness.

Our system strives to solve these problems. It aims to provide
services across different domains, acting as analysts,
personal assistants, data scientists, or scientific advisors.

METHODS WE USE

The goal we pursue is not easy, in order to tackle it we need to be on the lookout for the most useful methods. In particular our work heavily relies on these new and exciting scientific fields:

PROBABILISTIC PROGRAMMING

TO REPRESENT
UNCERTAIN KNOWLEDGE

CATEGORY THEORY

TO CAPTURE
STRUCTURE AND COMPOSITION

COGNITIVE SCIENCE

TO UNDERSTAND
REASONING AND LEARNING

GOING BEYOND TODAY’S APPROACHES

Today there are many approaches to AI
each with their own strengths and weaknesses.

Rather than using them in isolation, we aim to create a system where they can be composed to get the best out of all, while avoiding their pitfalls.

Approach
Strengths
Weaknesses
Fast for larger data
Hard to combine models Opaque reasoning
Easy to combine models Transparent reasoning
Slow for larger data
Fast to query
Limited to static facts
Reliable facts and algorithms
Requires manual maintenance

News feed

PlantingSpace
PlantingSpace@mas.to