Anshul Ramachandran at VOICE & AI 2023
Modev's VOICE & AI conference blurs the lines between human ingenuity and artificial intelligence to create something truly memorable and insightful. At the event's 2023 edition, we were fortunate to feature Anshul Ramachandran, Head of Enterprise and Partnerships at Codium, where he took us on a journey through the best practices for using AI tools as a software developer.
With his easy-going demeanor and knack for simplifying complex topics, Anshul kicked things off by asking the audience a simple question: "Who here codes?" The sea of raised hands confirmed that we were all in the right place, ready to explore the AI-powered toolkit revolutionizing how we develop software.
Let's look at what was discussed.
Simplifying Coding through AI
Imagine this scenario: we have a contact form that needs email validation. Typically, this would mean a time-consuming search for the right piece of code. But Anshul provided us with a different approach, one in which AI takes care of these tedious tasks. Using Codium's handy side panel within Visual Studio Code, he simply queried the AI, "Where is the contact form stored?" In an instant, the AI combed through the codebase with precision, quickly identifying the exact spot where the contact form resided.
AI can explain code snippets, sparing us the awkward dance of interrupting senior devs with our queries. It was like having a virtual mentor, always ready to decode the mysteries of the codebase with a simple tooltip explanation.
However, the real magic happens with the autocomplete feature. With a casual comment like "validate email address," the AI conjured up a regex pattern that would make even Stack Overflow proud. And just like that, with a few keystrokes, the contact form was equipped with a shiny new validation function - no manual coding required.
This is a testament to the power of generative AI in the software development lifecycle. From clueless to code-complete, the AI assistant guided him through the process, even churning out unit tests and documentation on the fly. It is a glimpse into a future where developers can focus on creativity and problem-solving, leaving the mundane to their AI counterparts.
The Current State of AI's Potential
When it comes to generative AI tools, the technology is still maturing. There are areas where it's robust enough to be trusted, such as autocomplete, chat, and search functionalities.
Anshul shared a Venn diagram that captured the crux of the situation—where the usefulness of AI tools intersects with their robustness. He cautioned against getting carried away by the hype of AI capabilities that aren't quite ready for prime time - and instead, guided us to focus on where AI can provide real, reliable value today.
Latency and quality in AI tools is extremely important. Even a few seconds of delay could render the suggestion useless for features like autocomplete. It is a reminder that in the world of coding, time is of the essence, and AI tools must keep up.
The Importance of Due Diligence
Diving deeper, let's explore what makes a high-quality AI product. It is important to understand the data used to train the model and its architecture. For example, if you're working with a cutting-edge front-end library and the AI model was trained on data from two years ago, you might not get the best suggestions.
Context awareness is also something you should focus on when it comes to AI tools. Code is a unique beast, with information scattered across files and directories. The best AI tools are those that can pull relevant context from your entire codebase to provide accurate suggestions rather than hallucinating about functions that don't even exist.
When it comes to best practices, It is best to use autocomplete like a passive AI that you shouldn't have to prompt. It's there to assist, not to be instructed. You can compare it to texting, where you either take the suggestion or keep typing without missing a beat.
For chat functionalities, instructive AI, where you actively seek out assistance, use chat as we would with ChatGPT, providing clear, descriptive prompts to get the most out of the tool.
Wrapping Things Up
Anshul wrapped up his session with a call to action. He urged us to embrace AI tools as assistants in our coding journey, not as replacements for good development practices. He also reminded us that AI is not infallible and that we should always review the code it generates.
Anshul's talk was a treasure trove of insights for any software developer looking to harness the power of AI tools. From the live demo to the deep dive into the mechanics of AI, he showed us that the future of coding is not just about writing lines of code; it's about collaborating with AI to elevate our craft to new heights.
Integrating AI into our development processes isn't just about enhanced efficiency; it's about redefining the nature of problem-solving within the industry. By offloading the more routine aspects of coding to AI, we free up our intellectual resources to tackle more complex, creative challenges. This symbiosis between developers and AI is the crucible for the next generation of software innovation. And Modev will be there, providing a platform for those building the tools to transform our collective future for the better.