The AI scientist who says fundamental science already ended

A storage unit in Sydney and a one-way move to Houston

The episode opens on Muzamil’s birthday — 13 September — and on a guest he admits surprised him. Dr. Nasir Ahsan founded and runs Abyss Solutions, a robotics company that started in Australia and now operates out of Houston, with offices reaching from Rio de Janeiro to Aberdeen to the Gulf to Islamabad. He has been working on machine learning since 2004, long before most of the world had a name for it, and he has been to places in the ocean, he says, that no other human has reached.

The move from Sydney to Houston is where the real character of the company shows. The obvious assumption — Nasir notes Muzamil making it — is that a technology company belongs in San Francisco, and that any relocation is about optimising around California’s tax laws. That is not the reason. The pull came from the oil and gas sector. By 2018 Abyss had Chevron, Anadarko and BP among its clients, and the commercial gravity was American. “It’s a land of entrepreneurs,” he says. The appeal is not personal — he has no family there, no stake in the country — but the culture lets you “deploy ideas and transform ideas into products and commercialise them very quickly.” He packed in two weeks. Five years later his belongings are still in a Sydney storage unit he has paid for and never emptied.

A 7am introduction and a purchase order two days later

The first oil-and-gas contract is a small story Nasir tells with relish, because it captures how he operates. He was flown to Houston in 2018 to present at a routine conference. He had no entry into the sector yet. So at 7am, before the day got busy, he found the conference organiser and asked plainly for a chance to talk to people at Anadarko — he did not want to fly home having only given a talk. The organiser walked him to a corner and introduced him to Jim Ranney, the director of engineering and technology.

Two days later Ranney assembled his directors around a table and told Nasir to speak. Abyss did a couple of hours of demos and applications from other sectors — its standard move when entering a new market is to show capabilities and ask clients which of their problems it can solve. The room went quiet. Ranney leaned back and asked, “so when are we gonna pay them, boys?” A general manager scratched his head, knocked on the table, and said, “you know what, screw it, let’s do it.” That was the first purchase order. The lesson Nasir draws is about appetite for risk: in that culture, ideas can become products fast, and for an entrepreneur, that is the dream.

A rebel with no purpose, and the teachers who fixed that

The conversation winds back. Nasir was born in Karachi, his mother from the northern areas, the family settled in Quetta, his childhood split four years at a time between Islamabad and Saudi Arabia, where his father — a renowned Pakistani mathematician — taught at King Fahd University of Petroleum and Minerals. Nasir is the eldest, and, by his own account, nothing like his father: where his father is calm, methodical and timely, Nasir is loud, ambitious and drawn to risk. “I was a troublemaker, I was a rebel,” he says. His father gave him space, and used the hard method when needed.

He is unusually candid about how badly the rebellion ran. He left home at 16, did his A levels in Pakistan because his father wanted his children to remain Pakistani, and then, free of supervision at university, “went berserk.” He got into trouble, got locked up in Saudi Arabia, and at one point was beaten by what he describes as a hundred people. That beating was the moment he understood the trajectory was unsustainable. His diagnosis of it — and of a lot of Pakistani youth — is precise: it was not anger, it was a lack of purpose. Energy, he argues, is a blessing and a double-edged sword; the failure is in not recognising, managing and channelling it, and in cultures that try to suppress energy rather than direct it.

The turn came through teachers, whom he treats with near-religious devotion. His first mentor after his father was Dr. Krishna Rao, a travelling computer scientist who cycled around the Saudi heat while everyone else drove — different, and Nasir was always drawn to different. Rao introduced him to AI in its classical, logic-programming form, made him write proofs for hours, and trained him hard. When Nasir wanted to do a research master’s in Australia and needed a proposal, the method was almost comically practical: he typed “AI, machine learning” into Google, pulled a few research papers, lifted the “future work” section, converted it into a proposal, and sent it to Australian universities. That is how he ended up applying machine learning to molecular biology, trying to model the causes of cancer.

From tool-maker to the field — and a torpedo idea

Two years into molecular biology, he hit a wall that defined his career. He felt he was becoming “a tool-maker for biologists” — building models around data and assumptions he did not control, with uncontrolled variables he could not interrogate. He wanted control over his data, his desk and his research problem. He wanted to feel the science, not mine someone else’s spreadsheet.

The pivot to robotics had a second source, and it is the most charged passage of the episode. He met a Pakistan Navy commander, Junaid Hussain, on exchange at the University of New South Wales. The commander told him about the days after 9/11, when, under Musharraf, Pakistan’s navy stood off against the American carriers headed for the invasion of Afghanistan — an ant and an elephant, in Muzamil’s framing. The young, hot-blooded Nasir asked why they hadn’t simply fired. The commander explained the engineering of it: a missile would be intercepted, and even a hit on an aircraft carrier would be absorbed by its internal compartments. That answer planted an idea — that Pakistan needed to work on defensive technology, an autonomous torpedo that could dodge and reach — and it pulled him toward field robotics. He returned to Pakistan against the advice of every friend who had stayed abroad, took a position at NUST working on drones and underwater vehicles, and describes those frustrating, bureaucratic years as some of the most gratifying of his career, because he was building something for his country.

He went on to the Australian Centre for Field Robotics for a PhD, proposing an underwater absolute-positioning device — a kind of GPS for the ocean — built on geomagnetic parameters, the way turtles navigate thousands of kilometres by sensing magnetic field strength and inclination. That PhD, and his Pakistani background, also left him permanently flagged for security clearance, even now, travelling on an Australian passport to the United States.

Why he walked out of academia

After the PhD came a Rio Tinto post-doc on fully autonomous coal mining, and the result he is proudest of. Miners drilling to understand the geology of a seam were sinking 108 exploration holes before they could mine an area. In a single month, Nasir replaced that with an alternative technique that needed six holes — and showed it could go to four. A Rio Tinto manager told him he had just saved them five years of money. Then, when he returned from an ocean-exploration mission, he was handed a termination: the coal programme was being shut down. “Is that how expendable I am?” he asked. That was the moment he understood academia, however prestigious, was “a golden cage.”

His argument against the field is broader than his own firing. Fundamental research, he says, is finishing, killed by the American “publish or perish” model that drives people to easy problems. The old idea of a university professor free to work on whatever they want has, in his telling, disappeared.

‘It’s all just pattern recognition’

This is the spine of the episode and the source of its title. When Muzamil raises the singularity — the popular idea that AI will soon help crack quantum physics and revolutionise innovation — Nasir refuses it. What people see, he insists, is not science. “Science is discovery. Science is mathematics. Science is understanding the nature of things.” What is actually happening is the commercial materialisation of concepts invented decades ago, made viable now by miniaturisation of sensors and processing power and Moore’s law. Neural nets, he points out, were taught as a core concept in the 1950s. In his own field, the fundamental contributions came up “until the fifties,” and most of what followed is recycling — new tyres and a fresh coat of paint on the same car.

The AI of today, he argues, “is nothing more than pattern recognition.” He reduces it to a line: machine learning is figuring out the value of m and c in y = mx + c, then feeding in x to get y. Driving a car needs more, but it is still function-fitting. Human intelligence — creativity, intuition, emotion, ambition — is “way beyond basic logic,” and the reason machines cannot reach it is mathematical, not a matter of hardware or software. “Can you characterise infinity? Can you characterise love?” Mathematics, he says, is a limited language, and your ability to express a concept defines your capability to solve the problem. Asked about Neuralink, his answer is the same: a glorified pattern recogniser. When Muzamil floats the cyborg idea — a hybrid super-species rather than a purely robotic one — Nasir genuinely pauses, says it is a proposition he had not considered, and admits it is worth thinking about.

What Abyss actually builds

The company, founded in 2014, came out of the same lineage — marine robotics, mining, agricultural drones — applied to one thing: AI for robotic perception, planning and decision-making. The vision Nasir preaches is “Intel inside” for field robots. Abyss does not make the drones; it makes the intelligence that goes on them. He is dismissive of what passes for advanced robotics today — Boston Dynamics, in his view, “is just hardware,” machines following dumb waypoints and pre-programmed instructions. What does not yet exist, and what Abyss is built to deliver, is a robot in the field that can notice a damaged pipe, weigh whether to raise an alarm or spray a marker or apply a rust-stopper, and make that decision in real time. The purpose, as he frames it, is to protect and conserve the world’s most precious resources, and to send machines where it is too risky, too remote or too large-scale to send humans.

He chose his co-founders deliberately — his wife Mahina, who handled the early finances, grants and tax work; Masood from Afghanistan; Ibrahim on the technical side — all of Muslim origin, a choice he can only explain as chemistry and a sense of safety. He is frank about the near-death moments: standing in front of a cash-flow sheet with two weeks of runway and thirty or forty people on payroll, while the founders told their non-Muslim colleagues that God would take care of it. “When you’re cornered and you rely on your faith, that’s when you see your strength.”

Pakistan in 2050, with and without Abyss

Muzamil closes with his standard question: extrapolate the last thirty years and tell me where Pakistan is in 2050, no wishful thinking. Nasir answers it twice. Take Abyss out of the picture, and he is blunt: Pakistan is not a conducive environment for development, it does not entice investment or commercialisation, and on current trends it will be no better in 2050 — possibly worse. But the problem, he insists, is fixable fast. The single priority is revenue generation. The country’s most precious asset is 220 million people who can export services; the answer is to train them, build cost-effective outsourcing, stop taxing the only asset there is, and push the healthy consumer market toward a freer market so people have something to spend on. Without opportunity, the best leave, and the brain drain pulls the rest into a vicious cycle.

Put Abyss back in, and the picture changes. Abyss Pakistan is the only subsidiary with a separate mission, and Nasir states it without hedging: “to become so embedded in the future core economic infrastructure of Pakistan through robotics and AI that no decision can be made without our influence.” While India is still chasing IT offshoring, he wants to equip young Pakistanis to make the country an export centre for robotics and AI — and notes that all of Abyss’s global delivery already runs through a building in Islamabad. He is honest that when he first named Pakistan to his board, people laughed and asked why not India.

The funding mechanism is concrete. One percent of revenue — for a company he aims to grow to five billion dollars in a decade or so — routed into two things: governance and education. He wants to become a top investor in Pakistan not for returns but to improve governance through micro-economies and micro-finance, and to fund schools that teach what he thinks Pakistan actually lacks. Not maths and chemistry, which everyone can learn, but integrity, honesty, critical thinking, the value of quality work, and confidence. He does not want to own the outcome or run the schools; he wants to set a system in place that keeps running. What gives him hope, he says, is that the noise among ordinary people is rising — through globalisation, the old ways are changing, and the people who run them know it.