While many of us associate artificial intelligence with Chat GPT, AI technologies are already embedded in many everyday tools. From checking your commute time on a web mapping platform like Google Maps to perfecting your grammar with text editing software, chances are you already use AI in your day-to-day life.
Human Intelligence vs Artificial Intelligence
The human intellect is the product of billions of years of evolution.
It can only improve its substrate capacity - the three pounds of meat between your ears - by that same, slow, plodding method of trial and error based strictly on how any random mutation alters its function to improve your ability to avoid premature death before you can get laid.
The AI Buzzword Trap
Actually speaking we all got trapped in buzzword AI without knowing what it would serve us. We all fall into three categories
- Consumer of AI products and Applications
- Applicants or Redefining users of AI
- Builders or Makers of AI Platforms
Each of them involves different levels of cost, competition and control.
Let us see some examples of existing use of AI and its applications.
Longer wait time to get an X-ray or Scan Report has drastically come down from 4 hours to just 20 minutes with the use of AI-assisted X-Ray Reports says a study from Apollo Hospital, Chennai.
Healthcare managers should be exposed to case-based learning using Indian and global examples of AI in Radiology, Pathology, ICU, Population risk prediction and Operations for a better health transformation.
You might have heard about the quicker processing time in banks advertisement in recent times. That is realized with the help of AI. With the help of AI, the banks have cut the processing time from 28 hours to 15 minutes says IBM. This is extraordinary isn't it.
AI in Popular Culture: Lessons from I, Robot
How many of you have noticed these conversations in the movie I, Robot(2004)
NS4 Robots: [from flashback] You are in danger!
Detective Del Spooner: [from flashback] Save her!
NS4 Robots: [from flashback] You are in danger!
Detective Del Spooner: [from flashback] Save her! Save the girl!
Detective Del Spooner: But it didn't. Saved me.
Susan Calvin: The robot's brain is a difference engine. It's reading vital signs. It must have done...
Detective Del Spooner: It did. I was the logical choice. It calculated that I had a 45% chance of survival. Sarah only had an 11% chance. That was somebody's baby. 11% is more than enough. A human being would've known that. Robots,
[indicating his heart]
Detective Del Spooner: nothing here, just lights and clockwork. Go ahead, you trust 'em if you want to.
Don't ever jump into a conclusion that I am against AI.
There are also risk in over automation with AI without safeguards . Accountability requires new forms and norms. If you bring Agentic AI to decision making without human in the loop as a responsibility then the organization which employs it loses its hold and control naturally to the machine that finally leads to in-accountability and loss of trust.
Many a times we try to do something with AI but not actually using to solve the problem.
Systems designed with such AI i.e., without defining the problem then it is difficult to change later or upgrade.
I would like to give you a thought about the building blocks of AI. I strongly believe that only a Lion walks alone not the sheep that flocks together and fall together without thought.
In general the basic building blocks of AI are
- Data
- Computing Power
- Algorithms
- Model Training
- Inference(AI in action)
Say ChatGPT is only AI agent that utilizes Large Language Models. Same as Google's Gemini Microsoft's Co-pilot. These AI applications are nothing but prompt based LLMs where you prompt with a query, then it fetches from the already fed data with already trained models inside it.
AI at Scale: The Massive Power Demand
The hidden things are not taught to everyone and spoken. As an embedded system organization we feel it is our duty to educate. The infrastructure behind AI is of two different categories.
One is physical which are large data centers, GPUs, High performance Computing clusters and energy systems.
Second is digital which are datasets, model repositories, governance frameworks, and access protocols.
A Google report on August 2025 claims that a single text AI prompt consumes electricity of about 0.24 watt-hours.
Imagine how many AI powered devices in a country and how many prompts are given every time. It will be huge power thirst when a considerable portion of our population uses say 1% of 147 crore will be 1.5 crore approximately and say 100 prompts which will be 1500MW-hrs. Unimaginable right.
The Environmental Impact of AI Infrastructure
Global ICT is responsible for up to 3.9% of global greenhouse gas emissions. In India, data center capacity is projected to reach 2,073 MW by 2027, an 85% increase from 2025 levels. India has a 59% AI adoption rate, yet 50% of its data centers are located in extremely water-stressed regions like Bengaluru and Mumbai.
India hosts nearly 20% of the world’s data, but only 3% of global data centre capacity. In India, demand for data infrastructure is rapidly rising with the growth of AI workload and
the current installed capacity of nearly 960 MW is expected to reach 9.2 GW by 2030
Massive Electricity Consumption: AI models require continuous high-density power for training and its inference.
E.g. In Mumbai, the surge in AI-driven data centers has led to concerns over the city’s reliance on coal-based power to meet the 1,100+ MW load.
Cooling systems in data centers drink billions of liters of water to prevent hardware from melting.
E.g. In Bengaluru, data centers consume over 26 million liters of water annually, even as the city faced its worst water crisis in April 2024.
Training a single LLM can emit ~3,00,000 kg of carbon dioxide.
Emissions from one deep learning model equal emissions from five cars over their lifetime.
How will we cater to such huge demand and make ourself sustainable. Do we have the infrastructure and engineers to meet this.
Google's Project Suncatcher which is currently experimenting by putting data centers servicing AI workloads in low -earth orbit (AI in orbit) and feed them on Solar energy where cooling the systems remains a major hurdle till today.
ISRO is also reportedly studying space-based data center technology.
Microsoft worked on project Natick which tried underwater data centers to make cooling easier but abandoned the project despite it is promise.
The Real Bottleneck: Shortage of Skilled Engineer
The answer will be currently in progress with respect to infrastructure and No w.r.t human resource as per my understanding.
Looking at the positive side of this, we not only need AI engineers but also the engineers behind it who serve as the backbone of AI.
As mentioned earlier AI solutions includes data centers (for the storage and management of data), specialized processing units which is the core computational engine that powers AI workloads, ranging from general-purpose Central Processing Units (CPUs) to specialised accelerators such as Graphics Processing Units (GPUs), Tensor Processing Units (TPUs), and increasingly Neural Processing Units (NPUs) and computational capacity, including high-performance super computing clusters. These support the computational resources required to train large-scale AI models.
We equally need Electrical engineers who thinks for the future energy demand. VLSI and embedded system engineers who work on low cost/low power High Performance Computes(HPC) , NPUs, TPUs and GPUs.
Some spark would have tingled your mind by now and a visual might have definitely splashing before your eyes to the road of opportunities in embedded systems.
If you nod that embedded engineer is equally the need of the hour then you are one among us. But the catch here is the guidance and direction towards your journey.
Lets Learn together as more such articles to continue.
