Is generative AI the catalyst for redefining modern engineering? Are we prepared for a world where intelligent tools shape how we create and code? This new wave of AI is transforming embedded systems, IoT, and the future of digital innovation.
I am a long-term resident of the United States, having worked for an average of one of the large software service companies. However, I do not represent a company. Today, I am a representative of a boy who began reading a particular magazine in 1998 or 1999, and that habit led me into the fantastic world of technology.
The question to ask is why. Why am I here? I understood that I would have a chance to pique your curiosity. It is that inquisitiveness which has been the basis of all my learning.
Why generative AI matters
Everyone is now discussing ChatGPT, OpenAI, Google Gemini, Microsoft Copilot, and Amazon Q as tools for writing embedded software and source code. All large corporations are lagging behind generative artificial intelligence (AI). It is my purpose to make you see why.
I began my career by reading electronics magazines as a teenager. In 2005, I was standing at my university gate as early as the morning, awaiting news about embedded systems. I had the chance to design embedded systems and build a career around it. 27 years later, I still enjoy the ride and look forward to new technologies.
I have always thought that we need to continue learning. Unless we continue to learn, we will fall behind. When we read news of people losing jobs due to AI, the reality is that they are not losing jobs to AI, but to a lack of learning. This is what I learned seven or eight years ago, and I have been making learning a priority ever since. I experimented with cloud and digital development, and even attended MIT, which was a fantastic experience.
Learning the basics of AI
AI did not start yesterday. It started during the 1950s. The development of GPT-4 and subsequent models spanned more than 70 years. The same ecosystem includes machine learning (ML), deep learning, and generative AI. ML is an AI subdivision. One subcategory of ML is deep learning. Deep learning is a subdivision of generative AI.
Nowadays, generative models can generate language, images, and even videos. These models have been trained on large datasets, some with billions of parameters, and their training enables them to generate new content that is similar to human creativity.
The emergence of large language models (LLMs)
Large language models (LLMs) are robust AI systems that can comprehend and generate text similar to human language. The rate at which ChatGPT reached one million users indicates the scale of the technology. The Netflix business reached that point in three and a half years. ChatGPT took five days.
The list of LLMs is vast and is unlikely to end in the near future, as companies and countries continue to develop their own models. India has also announced plans to establish its own LLM.
Embedded systems and the problems that we encounter
As embedded developers, we face challenges daily. In addition, there is increasing complexity, shorter deadlines, price pressure, reduced reusability of applications across platforms, tedious debugging, absent documentation, and complex integration. They are the usual areas of concern to every embedded engineer.
On one side of the technological environment, LLMs dominate and on the other, our hidden problems. These two worlds are brought together, and this is the real value.
Fig. 3: All about embedded systems and IoT devices
The generative AI benefits embedded in development
Generative AI may be helpful in embedded development in four key ways:
- Automated code generation
- Driver code generation
- Auto documentation
- Test automation
Fig. 4: How Gen AI is helpful
At least seven powerful methods of generative AI can transform workflows, such as debugging and refactoring, understanding codebases through natural language, and creating personalised learning environments.
Cursor, Copilot, Gemini Code Assist, Amazon Q, and ChatGPT are tools that can significantly increase the productivity of embedded and IoT developers. This is why some of these tools are my personal favourites: they facilitate code generation, documentation automation, and a better understanding of complex systems in less time.
Strategy, risks and the road ahead
The implementation of AI must be planned appropriately. Developers do not have to experiment. Organisational executives should specify their approach to AI adoption, which tasks are essential, and how control will be exercised. The importance of AI governance has been highlighted, particularly given that AI can hallucinate or generate misinformation. AI poses a severe threat without validation and guardrails.
Gartner provides a roadmap comprising strategy, use-case identification, organisational preparedness, culture, governance, and engineering. All of these are crucial to long-term success.
What comes next
In the coming years:
- Embedded development will be forced to use copilots.
- Edge devices will be used to run real-time LLMs.
- Optimised AI-ready hardware and software environments will be built by hardware vendors such as Intel, Nvidia, and STMicroelectronics.
We are at the dawn of a gigantic change. The trick lies in doing things step by step. Learn continuously. Stay curious. Take advantage of the available tools. Explore. Experiment. Grow.
The article is based on the talk at EFY Expo, Pune, titled ‘Adopt GenAI Technology for IoT-enabled Embedded Systems Development’ featuring the opening speech by Abhijeet Desai, Group Project Manager, HCLTech America. It has been transcribed and curated by Vidushi Saxena, Journalist at EFY.
