Beyond the Buzz: AI Alone Won’t Transform Manufacturing

#AI #ArtificialIntelligence #AIinManufacturing #IndustrialAI #FactoryOfTheFuture #ManufacturingTech #Industry40 #AsimTewari #SmartManufacturing #AppliedAI #IITBombay

“In manufacturing, AI alone isn’t enough. But AI plus engineering—that’s where the real transformation begins,”

“If AI can read your retina, your lips, and even your thoughts—why can’t it run your factory? The answer, as Prof. Asim Tewari reveals, lies in the constraints of real-world engineering.” - Prof. Asim Tewari, Department of Mechanical Engineering, IIT Bombay.

March 2025 : the second episode of his insightful series, Prof. Asim Tewari, Chair Professor at IIT Bombay aIn nd one of India’s foremost experts in AI for manufacturing, takes us deeper into what AI can—and more importantly—cannot do in industrial environments.

While recent advances in artificial intelligence have stunned the world—think ChatGPT, AI-generated art, and self-driving cars—Tewari brings a sobering and essential perspective for manufacturing leaders.

AI's Superpowers: Eye, Lip… Even Mind Reading

Prof. Tewari begins with jaw-dropping examples of AI's capabilities. From retina-based health diagnostics to lip-reading from news broadcasts, and even MRI-based image generation of human thoughts, deep learning systems powered by massive data and GPU computing are achieving what once seemed like science fiction.

Generative AI, he explains, can now create realistic human faces, compose original music, and even run virtual influencers with social media accounts. “There’s an AI CEO running a beverage company in Europe,” he says. “She—or rather, it—never sleeps, never takes a vacation, and never makes emotional decisions.”

It’s a surreal shift. But amidst this technological explosion, Prof. Tewari brings the focus back to where it matters for Pro MFG readers—the factory floor.

So Why Hasn’t AI Taken Over Manufacturing?

Here’s the paradox: AI can drive autonomous cars in chaotic city environments. Yet, it struggles with controlled shop floor operations.

Why?

Because data quality and variety are everything in deep learning. In manufacturing, most process data is narrow-band and highly repetitive. "Even if you have 10 years of production data," Prof. Tewari says, "if all your process parameters remain within the same controlled range, AI won’t learn anything new. There’s no variety."

Moreover, factories often contain hidden variables—complex dependencies and phenomena that aren’t measured or even measurable. These variables are critical to quality outcomes but are invisible to current AI systems.

And then there's the issue of trust and risk. When an AI model suggests a radical change to a blast furnace, no plant head is going to risk millions of dollars without extensive validation. AI hallucinations—where the system makes confident but false predictions—can be disastrous on the shop floor.

Engineering Constraints Still Matter

Manufacturing is governed by principles—mass balance, energy conservation, thermodynamics. “AI cannot afford to hallucinate when one kilogram of material must yield exactly one kilogram of product,” says Tewari. “In the real world, physics trumps fiction.”

This is where many AI initiatives in industry hit a wall. The tools are powerful, but they lack the hard-coded knowledge of engineering science that has evolved over two centuries.

The Path Forward: XGen AI—Merging Intelligence with Engineering

Prof. Tewari doesn’t leave us with a problem—he offers a roadmap.

At IIT Bombay, his team is developing XGen AI—Expert Generative AI models that blend traditional engineering wisdom with cutting-edge machine learning. These models respect physics-based constraints while unlocking AI’s pattern recognition and prediction power.

“In manufacturing, AI alone isn’t enough. But AI plus engineering—that’s where the real transformation begins,” he asserts.

For manufacturing leaders exploring AI adoption, Prof. Tewari’s insights are a vital reality check. While AI’s potential is enormous, blind faith is not the way forward. Engineering judgment, data quality, and contextual understanding remain irreplaceable.

And yet, the future is coming fast. Whether you're a skeptic or a believer, one thing is clear—those who combine AI with domain expertise will define the next era of industrial leadership.

Pro MFG Media Newsletter

Get regular manufacturing insights that matter and enable business.