For someone who only became familiar with the concept of artifical intelligence in the early 2020s, it may appear as though AI is all about machine learning: neural networks with various architectures, "trained" using a relevant corpus of training material, offering stochastic responses that, if the model is trained well using a suitable corpus, will have a high probability of being right or, at least, near optimal.
However, conventionally the field of artificial intelligence is much broader, and until recently, neural networks occupied only a small, mostly experimental, niche.
In particular, there is the subfield called symbolic AI, sometimes called GOFAI (Good Old-Fashioned AI) in jest: and within the field of symbolic AI, there is one of the oldest families AI applications, computer algebra systems (CAS).
The MACSYMA system, as it was called back then ("Project MAC's Symbolic Manipulator"), was born in the late 1960s at the Massachusetts Institute of Technology (MIT).
In 1982, a version of MACSYMA was made available to United States government agencies, among them the Department of Energy (DOE). In 1999, this version, DOE Macsyma, was released as open-source software under the GNU Public License (GPL). Using the name Maxima, it has been actively maintained ever since by a group of volunteer developers.
In terms of its mathematical capabilities, Maxima is mostly on par with leading commercial computer algebra products, including Maple and Mathematica.
Maxima also has some unique strengths, among them in particular its powerful tensor algebra implementation. The itensor package of Maxima specifically implements abstract index notation, which is particularly useful in applications that range from general relativity to modern machine learning.
Maxima is also easily scriptable, making it useful as an embedded computer algebra system. Maxima is integrated within the WISPL chatbot, allowing language models to invoke Maxima on request for complex calculations.