09/08/2021
Coding is more than transforming sentences into codes. My comment is Deep Learning can help to match sentences to code blocks but not sufficient to figure out control and data flow without other knowledge.
"...In June, a team of researchers. .. tested several types of models on APPS, including OpenAI’s GPT-2, GPT-3, and an open source version of GPT-3 called GPT-Neo. In experiments, they discovered that the models could learn to generate code that solves easier problems — but not without syntax errors. Approximately 59% of GPT-3’s solutions for introductory problems had errors, while the best-performing model — GPT-Neo — attained only 10.15% accuracy...."
Salesforce's CodeT5 model can understand and generate code with leading performance, researchers at the company claim in a paper.