Quest for Day One Productivity
People are adapting to ChatGPT fast. I've heard many amazing examples, from automating mundane tasks to blowing up whole industries. Our friend is teaching French at a high school. She's using ChatGPT to generate learning tasks and assignments for students, which saves tons of time and effort. She calls it "Evil ChatGPT," and students do not like it at all. On the other side of the spectrum, I saw somebody developing a model to scan US patents and help generate new ones.
Despite the plethora of excellent examples, I see two extremes. I talked to a librarian about paying a fee for a lost book and asked if they're starting to use ChatGPT at the library. The replay was no. When I asked why, she said, "It's for the bosses to decide (if we should use or not ChatGPT)." I hope they find a way to learn and use it while bosses are thinking about what to do.
Speaking about bosses, ChatGPT, and learning. Some believe it is a game changer for productivity. I saw this post recently on Linked-in; a CTO is sharing the vision of making junior developers productive from the moment they step foot in the office (or login for remote workers).
Here's a problem. If you're a CTO and have yet to figure out how to use ChatGPT to change the tone of your message so it is less NET NEGATIVE - then what's the chance the AI technology will help make your developers more productive?
Ah, this never-ending quest for productivity in the world of software development. LinkedIn is particularly fixated on leaving no stone unturned in enhancing the developer's output. It used to be Agile and Scrum; you know how it usually goes: "This amazing ticketing system will assist you to storypoint user stories...". Now, they've stumbled upon a shiny new plaything.
It was also impressive how many people didn't want to consider the consequences after reading that post. There were dozens of comments in the "this is amazing" style, and just one person said it was one of the saddest posts. He then had to explain why. I decided to check what ChatGPT thinks about it, so this is what I asked after a few iterations (it was getting stuck on technological advancements - just like the origin post commentators)
Q: Act as developer considering new opportunities. Assess and evaluate culture first rather than technology.
You saw a linked post by CTO. Assess interactions, communication and culture inside company between engineers based on opinions and hints provided in post. Skip technological advancements, try to understand interactions of people. (link or text)
领英推荐
ChatGPT: The CEO seems to be focused on productivity and efficiency, and values the potential of AI to assist entry-level developers. This might suggest a company culture that emphasizes learning and growth, with a strong emphasis on leveraging technology to support employees. However, the statement about entry-level developers being a net negative impact might also hint at a potentially demanding or high-pressure environment
Q: Could it be a problem? Where else you see potential issues?
ChatGPT: Yes, it could be a problem. Labeling entry-level developers as a "net negative impact" might create a negative perception of junior employees, leading to a less supportive and collaborative environment. This could result in a culture where junior employees feel undervalued or discouraged from asking for help. Additionally, the emphasis on AI assistance may inadvertently reduce the importance of human mentorship and interpersonal relationships, which are crucial for personal and professional growth.
The CTO may have overlooked a few essential details. ChatGPT isn't a one-size-fits-all replacement for human-led learning and mentorship. And contrary to popular belief, senior developers might need ChatGPT's support and guidance more than juniors. Good luck convincing them to adopt AI assistants into their highly efficient workflow zones.
So, what's the latest news in the ChatGPT universe? I had a productive conversation with some developer advocates from OpenAI, and I highly recommend connecting with them. You can find links to their calendars for meetings on this page. I'm interested in two essential questions:
I learned there's an API of course, a few official SDKs, but beyond that, it's a relatively uncharted territory. Most UI projects are based on NextJS. I'm fond of chatbot-ui; deploying it on Vercel takes only a few minutes and provides instant benefits for the team with a better interface.
The prompt engineering discussion was rather amusing. I showed Victor the Value Validator prompt and attempted to explain its purpose. This was challenging, describing Victor as a person you'd talk to as a DevOps, pre-sales, or cloud architect, and you'd be trying to find the ideal solution for Victor's problems.
The dev advocate examined the prompt and immediately said, "Oh, there's no need for such extended definitions!" He attempted to write a shorter version using ChatGPT version 3.5, which, as expected, came far from the rich conversation I'd managed to create. Prompt engineering, folks – it's a thing that takes quite a few iterations to make the model behave.
Lastly, for any developers aiming to boost productivity or just starting to learn AWS CDK, check out this slimmed-down, barebones version of the CDK Typescript application in the Learning AWS CDK group. With just four files in the Typescript project, deploy a lovely VPC using three simple commands. It will save you the trouble of consulting ChatGPT about generating ~500 lines long Cloudformation.
And with that, I'm off to enjoy the Mexican Riviera with my family – ChatGPT did help to handle vacation planning but someone has to actually enjoy it.