‘World Models,’ an Old Idea in AI, Mount a Comeback
1 min read
Summary
The next focus for AI research is the development of ‘world models’ which simulate different scenarios that an AI can use to make predictions and decisions.
Various luminaries in the field of AI are proponents of world models as they seek to create AI that is smarter and safer.
The concept of world models is not new and various fields such as psychology and robotics have used similar ideas previously.
The advantages of world models are that they give AI the ability to understand and interact with the real world more effectively, and give humans more transparency and explainability over how AI reaches its conclusions.
However, the issue with world models is that they are difficult to agree on what should be included and to what level of detail They are also hard to test for and create, with current AI systems using sometimes incompatible ‘heuristics’ to mimic the world.
Despite these concerns, the potential benefits of world models mean that all the major AI labs are currently focusing research in this area.
