Grammatical Framework for No-Code AI

No-Code AI refers to a code-free way of building AI solutions, usually through a graphical, drag-and-drop approach. This idea is related to Smart Undertake, where we were building smart contracts out of natural language descriptions.

Instead of using a graphical interface, the intention here is to use a natural language description of the AI problem to be solved and produce using the Haskell-based Grammatical Framework (GF) to translate to a solution in, for example, Scikit Learn.

These days, this type of transformations are all the rage doing them with soft methods such as using GPT-3 as code generator.

Using GF is much slower and laborious but it lends itself to a Bazaar-style open source development. This process has much better chances of maintaining the semantics of the generated code. Soft methods have been known to introduce many bugs, including security-related ones.