Go language shines for AI-powered workloads, survey says

The Go programming language is viewed by developers who build AI-powered applications and services as a strong platform for running those workloads. Nevertheless, Python is viewed by Go developers as the language to use when starting AI-powered work, according to a recent survey of Go developers.

In the semi-annual Go Developer Survey for 2024, conducted in January and February  by the Go team at Google, survey respondents who build AI-powered applications and services shared a sense that Go was a strong platform for running these types of workloads in production. A majority of respondents working with AI-powered applications already use Go or would like to migrate to Go for AI workloads.

The most commonly documented paths for beginning with AI, however, were Python-centric, the survey indicated. This has resulted in many organizations starting AI-powered work in Python before moving to a more “production-ready” language. The most common kinds of AI-powered services respondents reported building were summarization tools, text generation tools, and chatbots.

The results of the Go Developer Survey 2024 H1 were published April 9. The survey received a total of 6,224 responses. In other findings:

• When respondents were asked which generative AI models their organization was using, 81% said OpenAI ChatGPT or DALL-E, followed by Meta Llama at 28%, Mistral AI / Mixtral at 18%, and Google Gemini, Imagen, or PaLM at 13%.
• When respondents were asked which libraries or services they used to integrate with generative AI models, 69% said OpenAI, followed by Hugging Face TGI or Candle at 22% and LangChain at 20%.
• Lack of time or opportunities were the most commonly cited challenges to reaching Go-related learning goals.
• Go developer sentiment remains high, with 93% of respondents satisfied with Go during the past year.
• 80% of respondents said they trust the Go team to “do what’s best” for developers while maintaining and evolving the language.
• Almost a third of respondents said they had participated in the Go developer community either online or in person during the last year.
• Most respondents write Go code on Linux, at 61%, followed by macOS at 58%, Windows at 23%, and Windows Subsystem for Linux at 17%.
• The preferred editor for writing Go code was Visual Studio Code, at 43%, followed by GoLand/IntelliJ at 33% and Vim / Neovim at 17%.
• The most relevant security or compliance concern cited when working on Go services was insecure coding practices, with 42% citing this issue.