In today’s fast-paced technology landscape, frameworks, languages, APIs, tools, and practices are constantly evolving. No doubt, all the Generative AI assistants are going to help developers/engineers/managers to keep up and be more productive in the entire Software Development Lifecycle (SDLC).
Ever since I tried ChatGPT, I have that pinned that on my bookmark bar and it is the first go-to resource for any questions, suggestions and solutions. The response is pretty fast when compared to other GPTs.
The other Copilot versions from Github, Amazon and Azure are either paid or in preview or available on trial basis. Recently, I had the opportunity to explore Amazon Q at work, which offers an interesting comparison. The first impression of Amazon Q is just OK. The responses took little time and in one case the source was from a personal blog link which was broken. Probably, the model was trained long back. But, I still wonder how could a personal blog be considered as a source to train a model? Broken links from official documentations are okay, but anyways.. the models are evolving, becoming more responsible and we have to accept that.
Unlike ChatGPT, which operates independently of cloud platforms, Amazon Q or Azure copilots are tightly integrated with the respective Cloud environment (i.e AWS Console or Azure Portal) or the IDE. These copilots have plugins available for few popular IDEs like VS Code / Intellij. The IDE integration can help streamlining workflows by keeping developers in the familiar environment. Thee portal integration would help devops and cloud engineers.
At a high level, these Gen AI assistants are built on a robust and fairly responsible foundation models. They are able to explain code, generate new code, optimize existing code and even fix bugs – making them valuable companions for developers.
At this point, ChatGPT stands out and still feels better for individuals. It supports uploading files and it is multimodal model with GPT-4. With it, one could do supervised, unsupervised or reinforced upskilling. ChatGPT could be your teacher for learning and understanding. Please do not use it for just copy/paste 🙂
Enterprises can go for Amazon Q, if majority of the development/infrastructure is with AWS. Same way, Azure Copilot would make more sense, if the organization is more into Azure services. This way, the service can reach broad set of people who use IDE or the portal. Amazon & Azure copilots do not support attachments, images or videos. These are mostly text based as of September 2024.
Amazon Q and Azure Copilot provide guard rails, governance and compliance options. We have to wait and see if these copilots can be trained with organization specific internal sources and how certain topics can be allowed only for selected roles or group of users. Example: The companies finance documents need not become available for anyone in the organization to queried, whereas Senior Management staff could get a summary or overview of the trend.
Understanding the basics of Prompt Engineering could help us get more creative and effective responses back from these GPTs.