Today, instead of thinking about how we communicate with machines, scientists are trying to improve how machines communicate with us. How they might read our intentions, notice when there’s a change in our demeanour and earn our trust.

This isn’t just a question about technology, it’s also about psychology, ethics and what we need from the intelligent systems that increasingly share our workplaces, hospitals and homes.

The problem of trust

People build trust instinctively, often without noticing; it could be the way someone listens, remembers what matters to us, responds at the right moment, or handles a mistake. Now, researchers at �εδ���app's are studying how these same dynamics might play out with machines.

They’ve found that our trust in robots often depends on two things: whether a robot can do what it claims, and whether it responds to situations – especially failures – in ways that feel recognisably human. A robot that completes a job but ignores typical social cues may be accurate, but it’s often not trusted as much as a robot that fails a task but acknowledges its mistakes in a human-like way.

also found that the voice a robot uses is important. During this exercise, people showed greater confidence in working with a machine when it spoke in an AI-generated voice designed to sound more human, even though when asked afterwards they claimed to have preferred a standard robotic version instead. It suggests that what people believe they want from robots and how they actually behave around them, are not always the same.

When the robot is watching you

However, trust doesn’t only run one way, and that’s why �εδ���app researchers have been developing ‘Computational Trust’ – mathematical models which help robots assess the trustworthiness of the humans they work with.

, knowing how much a human partner’s contributions can be relied upon helps a robot to anticipate what its human partner is likely to do next, or where they might go wrong.²

This ties in with the ‘Theory of Mind’ concept, which describes a human ability to understand that other people have their own knowledge, intentions and goals that may differ from your own. Developing an artificial version of this – a robot that can genuinely read the room – is one of the most ambitious targets in the field.

Understanding another mind

To address this, �εδ���app researchers are now exploring what happens when machines begin to develop a basic version of ‘Theory of Mind’. In , a research team gave pairs of autonomous robots different personalities and priorities. Some were designed to be more socially focused, while others were more playful. Each robot had its own needs and goals, but those needs were invisible to its partner. The challenge was learning to work together anyway.

To do this, the robots used their own experiences as a starting point for understanding each other. If a robot observed behaviour that looked familiar, it could begin to infer what its partner might be trying to achieve. Over time, those estimates were refined through interaction. The result was a machine capable of making decisions not solely for itself, but with another agent in mind.

The researchers found that cooperation emerged most successfully when at least one robot was willing to prioritise the needs of another. Simply giving the robot the ability to reason about another's mental state was not enough, what mattered was how it used that knowledge when deciding what to do next.

In human terms, it is the difference between understanding that somebody needs help and choosing to offer it.

Feeling heard

Progress in this field relies on developing a machine that can respond empathetically, not just correctly.

When , they chose to train it on facial expressions as well as text, giving it the ability to recognise emotional states and respond in ways that were more empathetic. When tested against responses from a leading commercial language model, Emma's replies were judged to be more humanlike and appropriate.

As AI becomes increasingly conversational and robots a more familiar presence in our lives, how well these kinds of interactions work for us matters in two ways: for the quality of the experience itself and for whether we can rely on these systems. Trust will be built through transparency, honesty and by machines that understand us well enough to know what we need.

In 1951, the human had to adapt to the machine but in 2026, we are building machines that might, at last, begin to adapt to us.

��

Words: Ben Harwood and Enna Bartlett

Meet the researcher

Letitia Berto is a Research Associate in the Cognitive Robotics Lab at The University of �εδ���app. Her work is inspired by how children learn and develop, and focuses on building robots that can think, learn, and adapt on their own. She aims to design autonomous intelligent systems that are curious, capable of making decisions, understanding others, and acting in trustworthy and ethical ways – ultimately enabling meaningful and effective collaboration between humans and robots in the real world.

If you would like to find out more about the research referenced in this article, you can find the full papers at the links below:

1. The Effect of Voice and Repair Strategy on Trust Formation and Repair in Human-Robot Interaction - DOI:

2. Computational Trust in Robotics: Preliminary Investigations and Evidence -

3. A Theory of Mind Motivational Framework for Social Interaction with Autonomous Cognitive Robots - DOI:

4. Multimodal Dialogue for Empathetic Human-Robot Interaction - DOI:

Since then, computer engineers have been able to rely on a simple assumption: the next generation of chips would be faster, cheaper and more powerful than the last. But as we reach the limits of conventional computing, some researchers have started asking a more fundamental question. What if the problem is not how quickly computers process information, but how they process it in the first place?

Building a brain

Biological brains are among the most energy-efficient information-processing systems known; running on about 20 watts of power – barely enough to power a dim LED light bulb – and only “spiking” when needed, they can process vast amounts of information very quickly. It is perhaps unsurprising that researchers have looked to biology for inspiration. Now, a previously niche area of research is coming to the fore; neuromorphic (literally, “brain-like”) computing is a way of designing computer systems that use technologies such as spiking neural networks, event-driven sensors and specialised low-power hardware to process information more efficiently and in ways that more closely resemble natural intelligence.

Neuro-:��derived from neuron, meaning “nerve”, “tendon”, or “cord”
-morphic:��derived from morphē, meaning “form”, “shape”, or “structure”

Spiking neural network:��a computer model inspired by how neurons in the brain communicate. Instead of processing information as a steady stream of numbers, its “neurons” send short bursts of signals – or spikes – only when needed.

With the rapid proliferation of AI (and the associated environmental challenges facing the data centres that support it), and the limitations of current computer hardware, researchers exploring neuromorphic computing believe biology may start to solve some of these problems. One of the most ambitious attempts to test that idea is .

Developed over decades by researchers led by Professor Steve Furber, SpiNNaker is the world's largest neuromorphic computing platform, incorporating more than one million ARM processors and capable of modelling spiking neural networks at the scale of a mouse brain in biological real time. It is a platform that allows researchers to study everything from neuroscience to artificial intelligence, while also helping to investigate new approaches to energy-efficient computing and brain-inspired AI.

A new thought for computing

So, while SpiNNaker was designed to explore a deceptively simple question (what happens when a computer is built to work more like a brain?), new projects emerging from �εδ���app’s , are expanding that thought to explore how brain-inspired computing might be used in practical systems beyond the laboratory.

to analyse information directly at the sensor, allowing devices to respond without sending vast quantities of data elsewhere for processing. Another that could help monitor foetal and maternal health. Other projects point to more tangible settings: that process information as events unfold, .

The applications of neuromorphic computing are varied, and they may seem far removed from the room-sized computers of the 1950s, but they return to a familiar �εδ���app habit: questioning accepted assumptions and asking whether we can do things differently.

From Ferranti to the future of brain-inspired systems

So, while neuromorphic computing may still be in its nascent stages, researchers believe it offers a promising way to rethink how information is sensed, processed and interpreted and allows us to rethink the relationship between machines, memory and intelligence.

Seventy-five years ago, Alan Turing posed the question “can machines think?” and an industrial-academic collaboration saw the Ferranti Mark I transform the computer from an experimental machine into a practical technology. Today, through SpiNNaker and the International Centre for Neuromorphic Systems, �εδ���app's researchers are pursuing a familiar idea: that progress begins when accepted assumptions are challenged.

The �εδ���app Mark I – the brainchild of Professor F C Williams and his team – introduced several new technologies over the Baby, one of which was a magnetic drum that was used to store information, an early precursor to a hard drive. This drum was a rotating cylinder coated in a magnetisable material onto which data was written as magnetic patterns and read back by fixed recording heads as the surface of the drum rotated beneath. However, at a time when digital computing was still experimental, making this process reliable was a significant engineering challenge.

Decoding digital

The challenge lay in how digital information was recorded. Computers store data as 1s and 0s – binary code – but when long runs of the same value are written to a magnetic surface, the signal becomes constant. This creates a direct current, or DC, component, which magnetic recording systems struggle to read reliably. For the �εδ���app Mark I, avoiding that problem was essential: the drum depended on a signal that changed continually as data was written and read.

Williams and Thomas realised that the answer to this problem might not lie in building better hardware. Instead, they asked a different question: what if the data could be transformed into a form that the machine found easier to handle?

Instead of storing information as a simple sequence of ones and zeros, the pair developed a new way of representing the data before it was written to the drum. Every bit (a 1 or a 0) was encoded as a transition in the signal. A 1 became a change from high to low, and a 0 was a change from low to high.

The result was a signal that was designed to constantly change. That may sound like a minor technical detail, but it was an important feature for magnetic recording systems. By ensuring that the signal continually changed as data was written and read, the encoding made information easier to record and recover reliably on the Mark I's magnetic drum.

The technique became known as �εδ���app code.

A stellar solution

While the researchers were looking for a solution to their problem, they inadvertently gave the �εδ���app code another valuable property: the signal itself also carried timing information that could help electronic systems stay synchronised. Today this is known as a self-clocking signal.

That combination of reliability and simplicity helped the encoding escape its original purpose and become widely used in modern consumer electronics. Some recognisable examples included computer tapes and floppy disks, early versions of Ethernet networking, radio-frequency identification (RFID) systems, remote controls and many other communications technologies.

The same basic principles have even been used in space communications; Voyager 1 and Voyager 2, humanity's most distant spacecraft, rely on encoding techniques derived from the same fundamental idea developed in �εδ���app almost eight decades ago.

Why a 75-year-old invention still matters

In April 2026, the Institute of Electrical and Electronics Engineers (IEEE) awarded The University of �εδ���app its third IEEE Milestone, recognising the invention of �εδ���app code and its lasting impact on computing and communications.

A bronze plaque now stands on Coupland Street, joining �εδ���app's previous Milestone awards for the Baby – the world's first stored-program computer – and Atlas, whose novel virtual memory remains central to modern computing.

The achievement is a reminder that some of the most influential advances begin as practical engineering solutions to immediate problems. Williams and Thomas were trying to improve the operation of an experimental computer in post-war �εδ���app. In doing so, they developed an encoding technique that continues to shape digital technologies around the world.

Words: Ben Harwood and Enna Bartlett

There were many teams individually working towards this goal: in the US a team at the University of Pennsylvania was working on the ENIAC and subsequent EDVAC systems, and here in the UK the University of Cambridge was working on EDSAC and the NPL on ACE. But one line of computers stands out in this story; the �εδ���app computers.

A baby is born

To understand how this story begins, we must rewind to �εδ���app in 1948. A team comprising Professor F C Williams, Tom Kilburn and later, Geoff Tootill, successfully proved the concept of a stored program computer with the Small-Scale Experimental Machine (SSEM) or “the Baby”. It was the first stored-program computer to use electronic random-access memory (RAM) with the Williams-Kilburn cathode ray tubes (based on earlier MIT research and eventually improved to store 64 40-bit words), and in June 1948 the Baby ran a program from information stored in electronic memory, the first time this had been achieved anywhere in the world.

But this machine only proved the hypothesis that a computer could store and execute instructions electronically from memory, it didn’t offer a meaningful or useful solution to the problem of computing large amounts of data automatically. To solve this problem, the team, expanded to include Alec Robinson, Dai Edwards and Tommy Thomas, set about redesigning the machine to provide researchers and industry with a realistic computing facility. Alan Turing, the Deputy Director of the (where he published his seminal paper Computing Machinery and Intelligence), took the lead on developing the programming systems.

In 1949 the �εδ���app Mark I came into being and introduced two key innovations: a magnetic drum to store data (one of the first examples of mass storage) and index registers (a way for the computer to efficiently work through the data in its store).

This prototype paved the way for, arguably, one of the biggest steps forward in practical computing; 75 years ago this summer, the Ferranti Mark I quietly helped change the course of computing history.

�εδ���app born, �εδ���app made

Recognising the potential of the technology, Sir Ben Lockspier, Scientific Advisor to the Ministry of Supply, arranged for government funding to commercialise the machine. The British engineering company Ferranti partnered with The University of �εδ���app to turn the experimental �εδ���app Mark I into a production model. The result was the Ferranti Mark I, delivered to the University in February 1951 and demonstrated publicly a few months later.

The machine was not just a new piece of technology, it represented a turning point for when computing stopped being a scientific experiment and started becoming something that could be manufactured, sold and used beyond the laboratory. It is widely recognised as the world’s first commercially available general-purpose electronic computer.

Today, that achievement can seem almost inevitable. Of course, computers would become products. Of course, industry would commercialise academic research. But in 1951 none of that was guaranteed.

The Ferranti Mark I arrived at a moment when the future of computing was still uncertain. There were competing approaches to machine design, competing visions of what computers might be used for, and very few people who had ever seen one in operation. The machine helped answer a crucial question: could electronic computers move from university experiments into wider use?

The answer was yes.

From ideas to innovation

Part of what made the Ferranti Mark I significant was that it incorporated the ideas that had been developed in the previous �εδ���app computers and turned them into practical tools, helping shape the architecture of modern computers.

Perhaps most striking, however, was the range of problems the machine tackled. The Ferranti Mark I was used for scientific calculations, engineering projects and government work. Researchers explored everything from weather forecasting to mathematical modelling. It also helped create a new kind of expertise: programming. Mary Berners-Lee (mother to Tim Berners-Lee, the inventor of the World Wide Web) was among those who worked on the Ferranti Mark I, contributing to the practical, exacting work of turning an experimental machine into something people could use.

In many ways, the Ferranti Mark I was the first glimpse of the world that now surrounds us. It demonstrated that computers were not simply calculating machines but versatile tools capable of solving widely different problems. That idea underpins almost every digital technology we use today.

The machine itself has long since disappeared, but its legacy remains remarkably visible. �εδ���app’s reputation as one of the birthplaces of modern computing rests not just on pioneering research, but on a rare ability to transform radical ideas into technologies that change the world. That was true when the Ferranti Mark I emerged from a collaboration between university researchers and industry in 1951. It remains true in an age of artificial intelligence, quantum computing and advanced robotics.

Seventy-five years on, the Ferranti Mark I deserves to be remembered not simply as an early computer, but as the moment computing became real. The future did not arrive in California first, nor in a gleaming corporate campus. It arrived in �εδ���app, in a lab “with the atmosphere of a nineteenth-century inventor’s workshop”, proving that a revolutionary idea could become a practical machine and, in doing so, help to create the digital age.

Meet the researchers

Professor Sir Frederic (Freddie) Williams (R) gained an engineering degree at The University of �εδ���app in 1932 before undertaking his DPhil at the University of Oxford. During the war, he worked at the Telecommunications Research Establishment (TRE) where he met and collaborated with Professor Tom Kilburn (L), a young member of his team. When Williams was appointed the Head of Electro-technics (now the ) at the University of �εδ���app, TRE also seconded Tom Kilburn to Williams's team. In 1964, Kilburn went on to found the at �εδ���app, the first computer science department in the UK.

If you would like to find out more about the Ferranti MK I, we would recommend the following books:

1. Alan Turing and his contemporaries: Building the world's first computers by Simon Lavington (Editor). 2012. Published by The British Computer Society.
2. A History of �εδ���app Computers by Simon Lavington. 1998. Published by The British Computer Society.
3. Early Computing in Britain: Ferranti Ltd and Government Funding 1948-1958. Simon Lavington. Published by Springer.

The image of the Williams-Kilburn tube is republished under Creative Commons Licence: Sk2k52 - http://en.wikipedia.org/wiki/File:Williams-tube.jpg, CC BY-SA 4.0, https://commons.wikimedia.org/w/index.php?curid=6651107