- Modular claims that its “Python++” language is 68,000 faster than Python.
- Clears the Path for Artificial General Intelligence (AGI) Advancements.
A new programming language called “Mojo” is making waves in artificial general intelligence (AGI) development. It combines the elegance of Python with the speed of C++ and Compute Unified Device Architecture (CUDA), offering developers a powerful tool to create AGI systems. After accumulating over 120,000 developer sign-ups for its playground and extensive discussions about the language’s capabilities, Modular, the company behind it, has officially released it for local download. Initially available for Linux users, support for Microsoft and Mac platforms is on the horizon.
One achievement in the early stages of its release is the successful implementation of “llama2.py” by Aydyn Tairov, known as “Mojician” in the Mojo community. Tairov’s implementation outperformed Andrej Karpathy’s “llama2.c” (Baby Llama), written in C, by 20%. This early success validates Modular’s claim that it is significantly faster than Python.
What sets Mojo apart is its speed, reportedly being 35,000 times faster than Python. The language’s developers recognised the limitations of existing programming systems that primarily rely on graphical processing unit (GPU) accelerators for operations, only falling back on CPUs for specific tasks. The language was designed to support the full spectrum of operations in a single language, offering seamless integration with Python code and a scalable programming model for targeting GPUs. Simple code adjustments to existing Python programs can boost computation speeds by over 68,000 times.
The language’s software development kit (SDK’s) initial release includes essential development tools such as the driver, Visual Studio Code Extension, Jupyter kernel, and upcoming debugging support. These tools facilitate efficient program development, enhancing productivity and flexibility. Despite its potential, the company has chosen not to open-source the language. Some liken the language to a closed-source Python, hindering its adoption. While there is speculation about the possibility of open-sourcing portions of language in the future, similar to the swift programming language’s incubation period, the delay in clarifying licensing issues has deterred potential users.
For the language to play a pivotal role in AGI development, the company may need to expedite the process of open-sourcing the language. Open source remains a vital direction for the future, and any delay in making the language more accessible could impede its widespread adoption within the AGI community.