Developers use AI to improve translation between programming or fix problems without any human assistance. However, some researchers envision a time where AI can write codes based on simple instructions given by non-developers.
Now, Microsoft and OpenAI have come together to announce their plans to develop CPT-3, one of the most advanced models for generating text to programming based on normal language instructions. This is the first commercial application of GPT-3 initiated since Microsoft shelled out $1 billion in OpenAI in 2020, securing exclusive rights to GPT-3.
“If you can describe what you want to do in natural language, GPT-3 will generate a list of the most relevant formulas for you to choose from,” said Microsoft CEO Satya Nadella at the company’s Build developer conference. “The code writes itself.”
Converting instructions into codes
Microsoft VP Charles Lamanna told WIRED the new tech can aid people to tackle complex challenges and empower them with little coding experience. GPT-3 will convert normal language instructions into PowerFx, a programming language that resembles Excel commands that were introduced in March by Microsoft.
The latest feature dwells on a neural network architecture called Transformer, which is also leveraged by other biggies such as Baidu, Google, and Nvidia to create large language models from text training data available on the web.
The future of coding is bright
Recently, a team of AI researchers developed a benchmark for automated coding with Python. This allowed them to measure how good large language models actually are. They found out that GPT-Neo outperformed GPT-3. The GPT-Neo is an open-source language model that comes with similar architecture as OpenAI’s flagship models.
The reason behind this, as explained by lead author Dan Hendrycks of the paper, is the fact that GPT-Neo is fine-tuned using data gathered from GitHub, a programming repository for coding. As researchers and developers continue to study how language models can help simplify coding, we can see big advancements in this field in the future. However, it remains to be seen how Microsoft, OpenAI, and GitHub will work together on this project.
