11 Mar 2026
Yesterday at the 60th Edition of Cobotalks organised by I-Hub Foundation for Cobotics Technology Innovation Hub of IIT Delhi, I had a chance to attend a talk on physical AI by Dr Santanu Chaudhary, Former Professor, Department of Electrical Engineering, IIT Delhi. This talk came just weeks after the India AI impact summit but I was excited nonetheless to learn more about some of the academic aspects of AI.
Dr. Santanu emphasized that while chatbots have taken over the mindspace there is more to AI than just interactions with a server on the cloud. The chatbots represent a more general-purpose intelligence, they are quietly disrupting several industries in the ‘knowledge-work’ space but physical AI is narrow task specific intelligence that has applications in industrial automation and handling tasks that are either too dangerous for humans or too much of a chore.
Autonomous vehicles are a great example of physical AI. They have demonstrated that it is possible to make AI do tasks that were traditionally performed by humans. An AV is essentially a robo car with a brain that is surprisingly efficient for the processing it provides. Only 100 watts of power consumed.
But self-driving cars are still task-specific. Can we do more with physical AI? When we try to make robots more human-like it does raise some important philosophical questions. Is cognition merely information processing? Is learning from observation the same as reasoning or understanding? Can robots become aware in a humanistic sense? Then there are questions about ethics that go even deeper.
While these questions peek into what future robots could look like at present, robots lack the intelligence that chatbots possess. So while you can have a very intelligent server that intelligence becomes severely constrained when it's embedded into a compact robot.
But do robots even need human level intelligence? That is definitely the goal of many researchers in the field but it seems that those in industry are quite satisfied with a robot that can perform specific tasks with ease and reliability. That is good enough intelligence for them.
Physically we've in the last few years seen robots evolve through leaps and bounds. Gripping ,movement ,and sensing have all become better. Some robots have even demonstrated athleticism and performed acrobatics. From a bio mechanical point of view these are remarkable achievements. Things are only going to get better from here.
Yet there are hurdles. 2d chips are reaching their computing limit. 3d stacks would continue to push Moore's law if engineering challenges could be overcome but as stack height increases losses would add up. With the current computing architecture a general purpose intelligence in a robot seems unlikely.
However, task specific intelligence is doable. If we look at nature, intelligent humans make a disproportionately small amount of total mass living beings. Nature does the heavy lifting through ‘non intelligent’ organisms. The microbes ,the soil detritivores -earthworms,cockroaches,ants and other insects that silently toil away day and night to make our world livable. As do the bees above ground. Without their work the world we have today would be inconceivable.
Why can't robots fill that niche in the human ecosystem? Some robots like Tesla's Optimus perform superhuman feats by lifting really heavy weights— that's already useful.
Intelligent Robots are not the only way to achieve automation. Intelligence can be encoded in multiple ways. Everytime a craftsman makes a die for forging he encodes intelligence into lifeless objects that can be used to create thousands of products. A pump drawing water from a stream to irrigate the lands is semi intelligent automation encoded electromechanically ,performing the task that would take hundreds of human hours. The pump may not be intelligent but it doesn't matter because those who made the pump encoded intelligence into its design to perform one specific task repeatedly.
Mechanical unintelligent automation might be more useful industrially and economically than intelligent Robots learning how to do common chores. It is important therefore to recognise and target high impact areas even as we pursue our vision of enabling truly intelligent machines capable of reasoning. That's what nature seems to suggest.
“The true ship is the shipbuilder”-Emerson
Do you care about international relations? Would you like to be a part of a non profit that seeks to foster international collaboration? Partner with us & use your skills in Science/Engineering/Research/Team Building/Consulting/Administration/Law/PR/Comms/Business to shape the future. Let's do this.
ABOUT bhū
bhū is a self funded non profit organisation dedicated to advancement of science and promotion of international relations.
We aim to promote international harmony through creation of specific councils and bodies for regulating and overseeing international issues and accelerate developments in nanotechnology, material science ,electrostatics, fluids, plasma science,thermodynamics and advanced manufacturing.
Let us work together
https://akshatjiwannotes.blogspot.com/p/bhu.html
Akshat Jiwan Sharma
Materials science/International relations/Partnerships
Mobile/whatsapp:+919654119771
email:getellobed@gmail.com
Comments
Post a Comment