AI: Is your code still private? How can you measure success?

Written by Philip Sadkov

‍

You might be inadvertently leaking your code outside the company. 

‍

Does this sentence scare you?

‍

Both ChatGPT and Copilot offer options to limit training on your data. Now, ask yourself: how often do you check these settings? Daily, weekly, hourly? Are you truly aware of how your inputs are used? Of course, it's better to be safe than sorry, so let's take a closer look.  

I lean towards ChatGPT for its out-of-the-box code privacy, allowing me full control over what I share. It's worth investing a bit more time to craft safe examples based on your issues – the answers you need will still be there. Copilot, with its built-in code suggestion feature, streamlines the process, allowing for quicker responses with less typing, but the price for this convenience is direct access to your code.

Our company, Cloudgeometry, endorses the use of both these AI tools, so we're all set to embrace them with open arms! To wrap up the security talk: I strongly advocate for ChatGPT, but you will need to manually monitor what data you feed into it.

Now that we've taken a closer look at ensuring AI does not inadvertently let others help themselves to your code,  let's dive into the real meat of the matter – how effective are AI tools helping you? 

Some recent studies show that the average developer's productivity jumps by 47% using AI.  Sounds impressive, right? However, the truth is, currently, there's no concrete method to quantify the effectiveness of AI tools. Speaking from personal experience, I use AI daily –how does that make me faster, better, and cooler than those who don't? 

Think back to when Google Translate first hit the scene. It revolutionized language translation, moving us away from traditional dictionaries. Over the past decade, it's become my go-to for crafting and fine-tuning messages. What are you using now? AI? :)

Focusing on Copilot, it's like autocomplete on steroids – a real game-changer. It doesn't just save your fingers from typing fatigue; it adapts to your specific project context, sometimes even predicting substantial code blocks. Official reports from Copilot suggest a trend towards assessing developer satisfaction, rather than productivity in terms of lines of code. From this perspective, and mine too, AI tools like Copilot are indispensable in a developer's toolkit, even if my experience is limited to pet projects.

Bottom line – my personal recommendations on where AI tools shine:

• Extended Autocompletion 👉🏼 Absolutely essential.
• Inline Documentation 👉🏼 Experiment with AI for docs, comments, and docstrings.
• Unit Tests 👉🏼 A great starting point to experience AI's capabilities, but don't rely solely on it.
• Exploring New Tech 👉🏼 AI's a fantastic ally here. 

Want to spin up an app in a new language in 40 minutes? With AI, it's a breeze! Remember, these tools are here to augment our skills, not replace the human touch. Embrace them, experiment, and enjoy the journey into the future of tech!

‍