All indicates that AI mathematics is here to stay. Aiming at maximizing “Human in the loop”, we developed an ExaktAI Workspace. Its beta testing is open to everybody at exaktai.ai/workspace.
The Workspace started in the context of the ExaktAI Validation project. We wanted to validate AI mathematics using the Python mathematics stack, SymPy, NumPy, SciPy and Matplotlib, on equal footing with Maple, Mathematica and MATLAB. The missing piece was an equivalent, full-featured interface for the Python stack, so we decided to develop one.
Making the Workspace multi-dialect then appeared natural. With AI mathematics advancing quickly, having different computational dialects tied to particular engines seemed increasingly like an obstacle to making full use of their combined computational power. Such an environment, we think, is appropriate for reproducing and extending AI mathematics, and for doing the mathematics ourselves.
Among other things:
* The project resulted in a full-featured CAS interface for the Python scientific stack: SymPy, NumPy, SciPy and Matplotlib, all of it included in the Workspace. It brings a genuine CAS workflow to the Python ecosystem. This may be an interesting move towards popularizing computational mathematics, especially relevant now, as AI mathematics advances.
* The Workspace reads Maple, Mathematica, MATLAB, Jupyter and Python files natively, without requiring Maple, Mathematica or MATLAB to be installed on the computer. This makes it possible to use lecture notes written in different computational dialects through a single interface, present them, or translate them from one dialect to another, and run them using the Scientific Python stack where translation is supported (a capability that is growing).
Feedback and suggestions are welcome.
Edgardo