The short answer is scary. Our developers share insights from using Gen AI.
Artificial Intelligence (AI) and its applications have become a part of our lives—from facial recognition machines to financial robo advisers and even recommendations on YouTube.
But nothing has captured our imagination more than apps such as ChatGPT or Dall-E. Known as “Generative AI” or Gen AI apps, they create new content (e.g., images, text, code etc) in response to natural language prompts. Their answers are generated from the massive data that their large language models (LLMs) have been trained on.
Couple of months ago, our content strategists shared how they’ve been using Gen AI in Content Strategy work. Our developers are also using Gen AI apps in their work now. Here’s their experience so far.
We’ve found that Gen AI does well in these three roles:
Gen AI can be used to generate entire code snippets, data models or scripts. During our tests it seems we can’t generate a fully working app from a single prompt yet, but some LLMs are focusing on this area and things are improving quickly. For now developers’ skills are still needed in order to ask the correct question at the correct time, understand the output and to be able to use or debug the generated results.
Most code editing softwares are also starting to add features based on Gen AI for code correction, completion, or detection of security vulnerabilities.
Gen AI can be used to quickly generate wireframes or other designs. Figma, Adobe or Midjourney are expanding tools available to the designer, but conversion-related tools like Fronty or Uizard are also very valuable since they allow to speed up tasks that used to be done by humans only.
Just as it is important to craft the right prompt to get the best results out of ChatGPT, the right inputs for Gen AI design tools will also affect the relevance of the generated outputs. Trial and error is key here, but it opens new possibilities for designers.
Machine learning is not new, but it is starting to be used in combination with LLMs and other Gen AI tools to provide precise guidance and better learning experience. Even low code platforms like OutSystems are adding AI-powered tools to give contextual assistance and relevant information to developers.
You can also simply ask questions to ChatGPT and generate document templates with it, but never take the output as-is without verifying its accuracy.
Because nothing beats experience, we built internally an LLM-based app using traditional coding as part of our #PRexperiments on AI and other emerging tech. After the success of this first version using Pinecone, LangChain and OpenAI API, we decided to reimplement the same features using Flowise AI. This tool allows to manipulate complex AI related concepts in a low-code drag-and-drop web UI.
As you can guess, it was much faster to create this new version this time, and we could understand more deeply how these mechanisms work. Since all these tools are still being developed, the technological landscape is evolving rapidly. It’s important to stay curious and look for AI news or articles online to keep learning.
While AI apps have been helpful, we find the results can feel somewhat “generic” or templated. The solutions provided by Gen AI certainly wouldn’t have come from a more skilled or senior developer. Rather, they read more like “model answers” from a textbook that may be right but do not address a complex problem in a specific context.
So for now, don’t expect to be able to use AI’s work as-is for all use cases. Developers still need to have solid technical know-how to get the best out of AI—be it in coming with great prompts or modifying the code it produces. More importantly, we still need to use our own creativity and experience to create value. Another thing that is still missing too for Gen AI tools is understanding of the context. We need humans to analyse complex situations and inject that unique and magical sparkle into our work.
"In no more than three years, anything that is not connected to AI will be considered broken or invisible."
Clearly, Gen AI is still in its early stages and has its flaws. We generally expect LLMs to give correct answers but they’ve been designed to generate plausible content, not absolute truth. LLMs can make factual wrong answers appear very convincing, but this problem is improving by leaps and bounds everyday.
It’s also important to note that Gen AI technology seems to move towards diversification and specialisation. The rich variety of libraries and tools compatible with Hugging Face is a good example of this, some being made for very specific tasks.
So, will developers eventually lose their jobs when Gen AI becomes more advanced?
The short answer is “Yes, some developers will lose their jobs.”
The longer answer is “It depends on the type of developer you are and the type of projects you work on”.
These are the types of developers that will disappear:
But take heart! As some jobs become obsolete, new ones will be created as AI-powered tools free up developers from grunt work, enabling them to focus on more creative and strategic work.
Sam Schillace, deputy CTO of Microsoft has predicted that “in no more than three years, anything that is not connected to AI will be considered broken or invisible.” Here’s how developers can stay relevant.
The first step is to learn the technologies and tools used in AI development. This includes programming languages like Python or R, full ecosystems like Hugging Face or ChatGPT plugins, frameworks like TensorFlow or PyTorch, and follow news related to AI.
You should also apply your new skills as you learn them, because nothing beats practice. A good start could be to create an AI chatbot for specific requests. It may sound simple, but by doing this you will get first-hand experience with vector databases, summarisation frameworks, NLP and various kinds of API or LLM models.
While AI can automate many routine tasks, it is not yet capable of solving complex problems on its own. Developers who can combine AI technology with creative problem-solving skills will be in high demand.
The field of AI is constantly evolving; it's important to stay up-to-date with these trends to remain competitive in the job market. You can refer to this aggregator site to explore AI tools.
Still clueless where to start? Use this list:
Step 1: Learn the fundamentals of Natural Language Processing (NLP)
Step 2: Understand the Attention Mechanism in deep learning
Step 3: Deep dive into Transformers (Deep Learning Models)
Step 4: Fine-tune Pre-trained Models
Step 5: Explore NLP Tasks with Transformers
Overall, the impact of AI on developers will be mixed: as some jobs become obsolete due to automation, new doors will open up. Your job as a developer will be safe if you stay curious and hungry to learn new things: today, it’s Gen AI nipping at your heels; tomorrow, it’ll be another new technology.
Apart from your tech know-how, your greater value lies in your human ability and intuition to see the bigger picture, connect the dots, and solve problems creatively—tasks that AI can’t replicate yet.
Note: We'll continue to run more #PRexperiments on AI and other emerging tech to see how they can help us deliver our best work for clients. If you would like to explore using AI in your organisation, connect with us by emailing email@example.com.
A quick experiment to explore how AI can be co-pilots at work.