Modern “sustainable”‘ innovations in architecture are failing to slow climate change, but revisiting ancient knowledge and techniques found in traditional architecture could offer better solutions.

This is the argument of architectural historians Professor Florian Urban and Barnabas Calder in their new book “Form Follows Fuel: 14 Buildings from Antiquity to the Oil Age”. The authors argue that energy availability has been the biggest influence on architecture throughout human history.

Their extensive study is the first to calculate energy inputs for a range of historical buildings, demonstrating how different types of fuel, from human labor to fossil fuels, have fundamentally determined building designs across civilizations and eras.

“The history of architecture can be told as a history of energy,” the authors explain. “Today’s architecture is accordingly the outcome of four centuries of effort, innovation and ingenuity directed at maximizing the proportion of architectural production and operation that could be powered by fossil fuel heat.”

This argument comes at a critical moment in architectural history, as the building sector currently accounts for 37% of all human climate-changing emissions. Despite decades of research and discussion, the environmental impact of buildings continues to rise.

Urban and Calder document how the shift to fossil fuels begins in the 17th century and transforms architecture more profoundly than any other development in human history. This transition reversed the previous dynamic, where labor was cheap and heat expensive, creating an architectural model which depended on energy-intensive materials and processes that reduced human input.

“If form follows fuel, ours is fundamentally an architecture of intense fossil fuel consumption,” the authors explain.

Even as society becomes more aware of emissions and carbon footprint, and more efforts are made to build sustainably, the authors prove that today’s architecture comes at a catastrophically high energy cost.

They explain how globally influential minimalist designs often depend on massive energy consumption, for example, the Seagram Building in New York, widely praised for its simplicity, received an energy efficiency rating of just 3 out of 100 from the US Environmental Protection Agency, and cost more energy to build than the entire labor cost of quarrying, transporting and placing 5.5m tons of stone for the largest of the Egyptian pyramids.

“Mies’s famous dictum that ‘less is more,” turns out to be missing a word: ‘less is more carbon,'” the authors explain. “Per square meter of floor space, it used four times as much energy as the average American office building in 2012.”

By contrast, pre-modern buildings like the Scottish blackhouse achieved remarkable thermal efficiency using only local materials and passive design strategies. Examples of buildings like these show how humans have always before been able to provide the interior space and thermal comfort needed for survival in a harsh climate, while being fully sustainable and recyclable.

The authors’ studies span 4,500 years of architectural history, from the Great Pyramid of Giza to Kuala Lumpur International Airport.

The authors offer practical solutions for contemporary architects by unpicking the specific energy costs of different building elements and materials. For instance, their research demonstrates how structural stone tenements used significantly less energy throughout their life cycle than similar brick buildings, providing quantifiable metrics to inform modern sustainable design decisions.

Professor Urban says, “With regard to energy consumption, the world has never had so many pharaohs. Not only special buildings like the Seagram, but even our most mundane buildings use more energy than the most extraordinary structures of the ancient world.”

As architects and policymakers search for solutions to the climate emergency, “Form Follows Fuel” challenges assumptions about sustainability always meaning technological advancement, and provides an alternative approach to low-carbon architecture.

“The historical conditions of life without fossil fuels often look like poverty to those living in today’s energy-rich societies,” the authors explain, “but while luxuries were sparse and ill-distributed, materials local, and technologies comparatively simple for most non-fossil-fuel buildings, they had one immense advantage at a global scale: they collectively used resources at a rate within the bounds of what the planetary ecosystem could sustain.”

In an increasingly divided world, one thing that everyone seems to agree on is that artificial intelligence is a hugely disruptive—and sometimes downright destructive—phenomenon.

At WIRED’s AI Power Summit in New York on Monday, leaders from the worlds of tech, politics, and the media came together to discuss how AI is transforming their intertwined worlds. The Summit included voices from the AI industry, a current US senator and a former Trump administration official, and publishers including WIRED’s parent company, Condé Nast. You can view a livestream of the event in full below.

“In journalism, many of us have been excited and worried about AI in equal measure,” said Anna Wintour, Condé Nast’s chief content officer and the global editorial director of Vogue, in her opening remarks. “We worry about it replacing our work, and the work of those we write about.”

Leaders from the world of politics offered contrasting visions for ensuring AI has a positive impact overall. Richard Blumenthal, the Democratic senator from Connecticut, said policymakers should learn from social media and figure out suitable guardrails around copyright infringement and other key issues before AI causes too much damage. “We want to deal with the perfect storm that is engulfing journalism,” he said in conversation with WIRED global editorial director Katie Drummond.

In a separate conversation, Dean Ball, a senior fellow at the Foundation for American Innovation and one of the authors of the Trump Administration’s AI Action Plan, defended that policy blueprint’s vision for AI regulation. He claimed that it introduced more rules around AI risks than any other government has produced.

Figures from within the AI industry painted a rosy picture of AI’s impact, too, arguing that it will be a boon for economic growth and would not be deployed unchecked.

To train artificial intelligence (AI) models, researchers need good data and lots of it. However, most real-world data has already been used, leading scientists to generate synthetic data. While the generated data helps solve the issue of quantity, it may not always have good quality, and assessing its quality has been overlooked.

Wei Gao, associate professor of electrical and computer engineering at the University of Pittsburgh Swanson School of Engineering, has collaborated with researchers from Peking University to develop analytical metrics to qualitatively evaluate the quality of synthetic wireless data. The researchers have created a novel framework that significantly improves the task-driven training of AI models using synthetic wireless data.

Their work is detailed on the arXiv preprint server in a study titled “Data Can Speak for Itself: Quality-Guided Utilization of Wireless Synthetic Data,” which received the Best Paper Award in June at the MobiSys 2025 International Conference on Mobile Systems, Applications, and Services.

Assessing affinity and diversity

“Synthetic data is vital for training AI models, but for modalities such as images, video, or sound, and especially wireless signals, generating good data can be difficult,” said Gao, who also directs the Pitt Intelligent Systems Laboratory.

Gao has developed metrics to quantify affinity and diversity, essential qualities for synthetic data to be used for effectively training AI models.

“Generated data shouldn’t be random,” said Gao. “Take human faces. If you’re training an AI model to identify human faces, you need to ensure that the images of faces represent actual faces. They can’t have three eyes or two noses. They must have affinity.”

The images also need diversity. Training an AI model on a million images of an identical face won’t achieve much. While the faces must have affinity, they must also be different, as human faces are. As Gao noted, “AI models learn from variation.”

Different tasks have different requirements for judging affinity and diversity. Recognizing a specific human face is different than distinguishing it from that of a dog or a cat, with each task having unique data requirements. Therefore, in systemically assessing the quality of synthetic data, the team applied a task-specific approach.

“We applied our method to downstream tasks and evaluated the existing work of synthesizing data,” said Gao. “We found that most synthetic data achieved good diversity, but some had problems satisfying affinity, especially wireless signals.”

The challenge of synthetic wireless data

Today, wireless signals are used in technologies such as home and sleep monitoring, interactive gaming, and virtual reality. Cell phone and Wi-Fi signals, as radio waves, hit objects and bounce back toward their source. These signals can be interpreted to indicate everything from sleep patterns to the shape of a person sitting on a couch.

To advance this technology, researchers need more wireless data to train models to recognize human behaviors in the signal patterns. However, as a waveform, the signals are difficult for humans to evaluate.

It’s not like human faces, which can be clearly defined. “Our research found that current synthetic wireless data is limited in its affinity,” said Gao. “This leads to mislabeled data and degraded task performance.”

To improve affinity in wireless signals, the researchers took a semi-supervised learning approach. “We used a small amount of labeled synthetic data, which was verified as legitimate,” Gao said. “We used this data to teach the model what is and isn’t legitimate.”

Gao and his collaborators developed SynCheck, a framework that filters out synthetic wireless samples with low affinity and labels the remaining samples during iterative training of a model.

“We found that our system improves performance by 4.3% whereas a nonselective use of synthetic wireless data degrades performance by 13.4%,” Gao noted.

This research takes an important first step toward ensuring not just an endless stream of data, but of quality data that scientists can use to train more sophisticated AI models.