Have machine in the loop: It is time to flip the human-in-the-loop model

A few weeks back, you might have come across a viral story about Rosie, the dog that captivated the internet. Rosie was diagnosed with aggressive and terminal mast cell cancer. Her owner, Paul Conyngham, a software engineer, refused to accept the verdict and turned to AI. Using a cluster of modern tools like ChatGPT, Google DeepMind’s AlphaFold, genomic sequencing and mRNA vaccine science, he helped develop a personalised treatment.

With researchers at the Ramaciotti Centre for Genomics at the University of New South Wales, he identified mutations in Rosie’s tumour to create a bespoke mRNA cancer vaccine. Recent reports say Rosie’s tumours shrank significantly after the treatment, and she has years ahead of her.

It is true that Rosie was not saved by AI alone, it was a combination of cutting-edge technologies and dogged human persistence and optimism. Essentially, this was not just machine magic —it was human ingenuity, with technology in the loop.


STORY OF SABARI
That idea has stayed with me because, a little while ago, I encountered something similar closer home.

I teach a course on AI Literacy and Vibe Coding at Ashoka University. On the last day, one student left a deep impression on me. In the two hours of class when we were teaching students how to vibe code, he used Google AI Studio to build a quick tool to autocorrect WhatsApp messages.
His name is Sabari. He is 22 and fully visually impaired.

However, he has already lived a life richer than many people twice his age. He has been a motivational speaker since he was six, been appreciated by both APJ Abdul Kalam and Narendra Modi; he has a degree in international relations, and is now studying at Ashoka. But what stayed with me was not any of those achievements.

It was the way he spoke about AI — calmly, practically, without any of the theatre that surrounds the subject.

For many of us, AI is still largely a convenience. It helps draft an email, summarise a report, generate a presentation, clean up a note. Useful, yes. But for Sabari, AI is something else entirely. It is independence.

He described how he uses these tools in ordinary life. AI tools like ChatGPT help him “see” posters, “read” PDFs and identify currency notes. Multimodal AI systems can describe objects, recognise colours and identify people around him. To a sighted person, these may sound like small things. For a blind person, this is the difference between dependence and self-reliance—between having to ask for help and being able to proceed on one’s own.

But what impressed me even more was that Sabari does not see himself merely as a user of AI. He wants to build an app called RizzVision, which he describes as an “auditory mirror” for visually impaired users.

Its purpose is simple, practical and deeply human. How does a blind person know whether their clothes match, whether the shirt is inside out, if there is a stain and whether the shoes are from the same pair? For most of us, these are passing decisions, but for the sightless, these are daily obstacles to dignity and self-respect.

The app’s prototype, built without any funding, uses Claude Code for vibe-coding, Gemini’s multimodal capabilities, ChatGPT for guidance, and a training dataset of 400,000 images scraped from the web.

Strip away the jargon and what remains is something quite beautiful: a brilliant young man, using the most advanced technologies available to create capability where the world had left a gap. RizzVision, an app on the phone, grants the blind “sight” with sound.

Seen together, Rosie’s story and Sabari’s story suggest a better way to view our relationship with technolStrip away the ogy. For years, the gold standard for AI has been the ‘Human in the Loop’ model, the idea that humans are the final check on potentially dangerous or erratic machines.

It is a defensive posture, born of fear. I believe it is time to flip that script and talk about “AI in the Loop”

TOOL TO EMPOWER
In this paradigm, the human remains the driver, but the technology is woven into the loop of human intent to expand what is possible. When we put AI in the loop of a medical diagnosis, we save lives. When we put it in the loop of a student like Sabari, we create agency, where there was once a barrier.

The demonisation of AI often ignores the fact that technology, at its core, is a tool for empowerment. The “tech bros” may be reviled, but the technology they have unleashed is being used by 22-year-olds in India to build “auditory mirrors” for the blind.

In the future it will not be built by AI alone; it will be built by people like Sabari and Conyngham who know where technology belongs in the loop, and where it does not. This may become the defining test of our civilisation —not whether we can build more powerful machines, but whether we can place them in the service of life, dignity and humanity.

If we fail, AI will become just another amplifier of human violence and vanity. But if we succeed, it may do something rarer. It may help us become not less human, but more fully so. Which of the two we choose depends on us.
Meanwhile, Sabari gives me hope.

Views are personal.