Society
AI goes nuclear, but what are the risks?
As technology companies invest in small modular reactors (SMRs) to meet energy demands for AI data centers in the future, how safe are they?

As AI fever runs high, BBC reported a US-based start-up, Digital Realty that plans to use small nuclear reactors to power their AI data center in Portland, Oregon. But why?
In an interview with BBC, Stephanie Hare, an AI commentator and technology researcher, said that powering data centers in general are very energy-intensive, leaving behind a massive carbon footprint in addition to the usage of water.
Gallons of water, for instance, functions as a coolant to counteract overheating in machines when it busy processes user requests.
Hare noted that computers there can use up to ‘half a liter’ of water to process requests from a single user at a time.
However, operating an AI data center is going to consume even more power.
“A normal data center needs 32 megawatts of power flowing into the building. For an AI data center it’s 80 megawatts,” said Chris Sharp, Chief Technology Officer (CTO) of Digital Realty, to the BBC. But it’s not just Digital Realty though participating in this enterprise.
Small Modular Reactors can generate one-third the energy of a conventional nuclear power plant and are said to be cheap based on design.
In 2023, The Verge reported Microsoft potentially showing interests in using ‘small modular reactors’ (SMRs) to fuel their AI data centers. These reactors split uranium nuclei with slow-moving neutrons, very much like conventional nuclear power plants.
However, lending a nuclear reactor to commercial establishments comes with its challenges. For one, only skilled workers can be relied upon to operate properly and manage the nuclear reactor.
Whereas for another, is for the nuclear reactor with its safety mechanisms to manage waste. However, scientists at Stanford University and University of British Columbia had worked out some technical flaws in SMRs. They reached the opposite conclusion to what SMR advocates had to say. They said there’s going to be more radioactive leakage owing to the small design that can’t absorb and take away byproduct neutrons from the chain reaction.
However, these generate one-third the energy of a conventional nuclear power plant and are said to be cheaper to design and manufacture. But how soon can they be deployed?
In the US, their Nuclear Regulatory Commission has authorized one such SMR design, by NuScale although it will be demonstrated only in 2029.
Spencer Lamb, Chief Commercial Officer at British data center developed Kao Data, said in the same BBC report, “I’ve heard about SMRs, but it will take a long time to deploy a nuclear-configured data center in the UK, and AI is happening now.”
BBC interviewed Dr Doug Parr, who’s chief scientist of the non-profit environmental activist group, Greenpeace UK, who labeled the unfolding story about SMRs powering AI data centers as mere ‘hype’. He said tech companies will develop cold feet when they realize that SMRs would prove to be much costly when they’re finally demonstrated. “Unrealistic hype lies behind the cost estimates for SMRs,” said Dr Parr. “This hype will fall away as delays and difficulties emerge.”
Paradoxically, we’ll never know how safe a technology is, unless we’ve already tested them.
However, Dr Michael Bluck a nuclear engineer at Imperial College London, UK was more optimistic – at least in a technical standpoint. He said, “There’s no reason why a small fast reactor can’t power a data center, except that you have to get it past the regulator.”
What about public trust though? The BBC doesn’t cover that. Won’t they have the final say in this case, since it involves nuclear energy? At least in history, nuclear energy has been a point of contention in the West, with public suspecting whether authorities were truly capable of ensuring safeguards against radioactive leakages and waste management. In Germany, policy failure to reassure the public actually led to the wide-spread phase out of nuclear reactors. Public trust is hard to achieve, but it takes the government and scientists to trust them back.
In the UK back in 1957, local farmers in Cumbria, England had suspected radioactive leakages from the Sellafield nuclear plant. However, authorities and scientists didn’t pay attention to the farmer’s concerns of a leak, until farmers strenuously lobbied to get the site checked for by scientists – later positively verifying the claims, leading to the shutdown of the plant.
The point isn’t that nuclear energy is somehow more unsafe compared to other forms of energy, say renewable energy. The numbers of countries operating nuclear reactors have actually expanded to 32 countries, including developing countries, with some 436 reactors operational of today.
Yes, catastrophe has occurred in the past – there’s the infamous Chernobyl and Fukushima events. The US alone had witnessed the Three Mile Island nuclear disaster in 1979. But we don’t want that to happen again.
The point is – paradoxically – we’ll never know how safe a technology is, unless we’ve already tested them.
But before that we need to keep the dialogue on as we discuss and discover hidden risks.
Earth
In ancient India, mushy earth made for perfume scent
Kannauj, a city in the Indian state of Uttar Pradesh, offers a sustainable alternative in producing perfumes using traditional modes of distillation.

A sweet scent typically lingers around in the air at Kannauj, an ancient city in India’s most populous state of Uttar Pradesh. It’s an imprint of the countless occasions when it had rained, of roses that bloomed at dawn, and of sandalwood trees that once breathed centuries of calm.. Though mushy smells are not unique to Kannauj, the city utilized traditional distillation methods to make perfume out of these earthly scents.
Kannauj has had a longstanding tradition in perfume-making since four centuries ago. The city, colloquially known as the country’s ancient perfume capital, still uses rustic copper stills, wood-fired ovens, and bamboo pipes leading to sandalwood oil-filled vessels, or attar as it is colloquially known, to make their perfume. Though it gives a pre-industrial look, a closer peek would reveal an ecosystem of complex thermal regulation, plant chemistry, sustainability science, and hydro-distillation chemistry at work.
When synthetically-made but sustainable perfumes, and AI-generated ones share the spotlight today, Kannauj’s tryst with perfumes offer an alternative, sustainable model in traditional distillation, which is inherently low-carbon, zero-waste, and follow principles of a circular economy; all in alignment with sustainable development goals.
Traditional perfume-making is naturally sustainable
In industrial processing, hydro-distillation is a commonly done to separate substances with different boiling points. Heating the liquids produce vapors, which can later be liquefied in a separate chamber. Perfumers in Kannauj follow the same practice, except it promises to be more sustainable with the copper stills, a process colloquially known as dheg-bhakpa hydro-distillation.
There’s no alcohol or synthetic agents in use. Instead, they heat up raw botanicals – such as roses, vetiver roots, jasmine, or even sunbaked clay – to precise temperatures well short of burning, thereby producing fragrant vapor. The vapors are then guided into cooling chambers, where they condense and bond with a natural fixative, often sandalwood oil. Plant residue is the only byproduct, which finds use as organic compost to cultivate another generation of crops.

Trapping earthly scent to make perfume
In the past five years, Kannauj’s veteran perfumers noticed a quiet, but steady shift in their timely harvest and produce. Rose harvests have moved earlier by weeks. Vetiver roots grow shallower due to erratic rainfall. Jasmine yields are fluctuating wildly. The local Ganges river, which influences humidity levels essential for distillation timing, is no longer as predictable. For an entire natural aromatic economy built on seasonal synchrony, this uncertainty has rung alarm bells.
“The scent of a flower depends not just on the flower itself,” Vipin Dixit, a third-generation attar-maker whose family has distilled fragrance for decades, said to EdPublica.
“It depends on the weather the night before, on the heat at sunrise, on the moisture in the air. Even the soil has a scent-memory.”

As a result, perfumers in Kannauj have begun to adapt, applying traditional wisdom through a modern scientific lens. Local distillers are now working with botanists and environmental scientists to study soil microbiomes, measure scent compounds using chromatography, and develop community-based rainwater harvesting to ensure sustainable crop health.
One of the most surprising innovations is trapping petrichor — the scent of first rain — through earth attars. Clay is baked during extreme heat waves, mimicking summer conditions, then distilled to trap the scent of rain hitting dry soil. This aroma, called mitti attar, is one of the few scents in the world created from an environmental phenomenon; and not a flower.
At a time when the world is scrambling to save biodiversity, the humble attar may become a template for green chemistry — one that doesn’t just preserve scent, but also restores the relationship between science, nature, and soul.
Society
How Scientists and Investigators Decode Air Crashes — The Black Box and Beyond
The final report may take months, but it will be critical in issuing safety directives or revising standard procedures.

As rescue and recovery operations continue following the June 12, 2025, plane crash in Ahmedabad, aviation safety experts are now focusing on the technical investigation phase. With 241 lives lost, the search for the cause isn’t just about accountability—it’s about prevention.
The Black Box: Aviation’s Memory Keeper
1. What Is the Black Box?
Despite the name, the black box is actually orange — for visibility. It consists of two components:
- Cockpit Voice Recorder (CVR): Captures conversations and audio from the flight deck.
- Flight Data Recorder (FDR): Logs dozens to hundreds of parameters — speed, altitude, engine status, control inputs.
These devices are housed in titanium or steel and can withstand:
- Temperatures above 1,000°C
- Underwater pressures up to 20,000 feet
- Crashes with up to 3,600 G-force
They also emit underwater locator beacons for up to 30 days.
2. Forensic Engineering & Flight Reconstruction
Beyond black boxes, investigators use:
- Radar data and air traffic control logs
- Wreckage analysis for structural failure clues
- Satellite-based tracking systems like ADS-B
- Weather data for turbulence or wind shear insights
Forensic teams often reconstruct the flight path virtually or even physically using recovered debris to determine failure points.
3. Human Factors & AI in Modern Investigation
New tools like machine learning and human factors analysis are used to identify procedural errors or lapses in judgement.
In many modern investigations, AI helps:
- Filter large datasets (e.g., over 1,000 flight parameters per second)
- Detect patterns missed by the human eye
- Predict similar risk scenarios in future flights
What Happens Next in the Ahmedabad Crash?
Authorities, in coordination with the Directorate General of Civil Aviation (DGCA), are likely:
- Retrieving and analyzing the black box
- Interviewing air traffic controllers
- Reconstructing the aircraft’s final seconds using both data and simulation
The final report may take months, but it will be critical in issuing safety directives or revising standard procedures.
Society
Researchers Unveil Light-Speed AI Chip to Power Next-Gen Wireless and Edge Devices
This could transform the future of wireless communication and edge computing

In a breakthrough that could transform the future of wireless communication and edge computing, engineers at MIT have developed a novel AI hardware accelerator capable of processing wireless signals at the speed of light. The new optical chip, built for signal classification, achieves nanosecond-level performance—up to 100 times faster than conventional digital processors—while consuming dramatically less energy.
With wireless spectrum under growing strain from billions of connected devices, from teleworking laptops to smart sensors, managing bandwidth has become a critical challenge. Artificial intelligence offers a path forward, but most existing AI models are too slow and power-hungry to operate in real time on wireless devices.
The MIT solution, known as MAFT-ONN (Multiplicative Analog Frequency Transform Optical Neural Network), could be a game-changer.
“There are many applications that would be enabled by edge devices that are capable of analyzing wireless signals,” said Prof. Dirk Englund, senior author of the study, in a media statement. “What we’ve presented in our paper could open up many possibilities for real-time and reliable AI inference. This work is the beginning of something that could be quite impactful.”
Published in Science Advances, the research describes how MAFT-ONN classifies signals in just 120 nanoseconds, using a compact optical chip that performs deep-learning tasks using light rather than electricity. Unlike traditional systems that convert signals to images before processing, the MIT design processes raw wireless data directly in the frequency domain—eliminating delays and reducing energy usage.
“We can fit 10,000 neurons onto a single device and compute the necessary multiplications in a single shot,” said Ronald Davis III, lead author and recent MIT PhD graduate.
The device achieved over 85% accuracy in a single shot, and with multiple measurements, it converges to above 99% accuracy, making it both fast and reliable.
Beyond wireless communications, the technology holds promise for edge AI in autonomous vehicles, smart medical devices, and future 6G networks, where real-time response is critical. By embedding ultra-fast AI directly into devices, this innovation could help cars react to hazards instantly or allow pacemakers to adapt to a patient’s heart rhythm in real-time.
Future work will focus on scaling the chip with multiplexing schemes and expanding its ability to handle more complex AI tasks, including transformer models and large language models (LLMs).
-
Earth4 months ago
How IIT Kanpur is Paving the Way for a Solar-Powered Future in India’s Energy Transition
-
Space & Physics3 months ago
Could dark energy be a trick played by time?
-
Society4 months ago
Starliner crew challenge rhetoric, says they were never “stranded”
-
Earth2 months ago
122 Forests, 3.2 Million Trees: How One Man Built the World’s Largest Miyawaki Forest
-
Space & Physics3 months ago
Sunita Williams aged less in space due to time dilation
-
Society4 months ago
Sustainable Farming: The Microgreens Model from Kerala, South India
-
Space & Physics6 months ago
Obituary: R. Chidambaram, Eminent Physicist and Architect of India’s Nuclear Program
-
Learning & Teaching4 months ago
Canine Cognitive Abilities: Memory, Intelligence, and Human Interaction