.png)
Many companies use technology to solve problems; fewer are solving problems within technology itself. Making it safer, more sustainable, more effective.
“This might be about better hardware, or it might be about new approaches to software,” says Roland Emmans, Head of UK Technology Sector & Growth Lending at HSBC UK. “But in every case, the goal is to solve challenges at scale and extend technology’s reach in ways we haven’t seen before.”
Meet three Trailblazers who are all doing that in different ways. What they have in common is that all of their work is connected to the current gravitational centre of the technology world: artificial intelligence. Might they inspire your next mission, too?
The challenge: Child safety on the internet
One in five 11- to 17-year-olds in Britain has felt pressured into sharing an explicit image of themselves. Once these images are online, they can spread and are extremely difficult to control.
This year, online safety business SafeToNet released HarmBlock, a suite of AI software which can tell when the phone’s camera is pointing towards an underage person in a sexualized context. If a parent is worried that their son or daughter might end up creating child sexual material (CSM) of themselves, they can enable the tool on their child's device.
When HarmBlock detects a sexualized image of a child framed in the phone’s camera, the camera shuts off. This could be transformative, because traditional moderation involves removing CSM from the internet after it has been created. But prevention, SafeToNet founder Richard Pursey says, is much more effective. “Our technology stops the child from being harmed in the first place, not afterwards.”
Crucially, since HarmBlock’s code is embedded into the very core of a smartphone’s operating system, it is all but impossible for a child to disable. HarmBlock software is now being integrated into new Vodafone Three HMD smartphones, and will be available to 7 million consumers this autumn. “It’s safety by design,” says Pursey.
The challenge: AI’s appetite for power
Ever-more-powerful generative AI models come with ballooning energy requirements.
Salience Labs develops photonic switches: computing components which use light instead of electricity.
As AI models grow, the amount of data that needs to move through different CPU and GPU racks in server rooms is overwhelming traditional electronic switches, the hardware that effectively ‘directs’ data between and within servers. Most data travels via fibre optics, but when it hits a conventional switch it must be converted into electrons and back again—adding latency and energy costs.
Salience builds switches that keep data entirely in the optical domain by “bouncing” photons—light particles—through the switch to change its pathway. “At scale, that in turn dramatically improves the efficiency and latency of AI models,” explains Vaysh Kewada, Salience’s CEO. An effective photonic switch, she says, significantly reduces the power used versus traditional electronic switches.
The challenge: Training AI on minimal data
Specialized machine learning models hold huge potential to assist with high-risk tasks, but they usually require masses of training data for the required level of reliability. In many cases, the quantity or quality of information might not exist.
Mind Foundry's AI systems are designed to slot into existing industries with specific, high-stakes use cases involving sensor data, particularly in areas such as defence, national security, and critical infrastructure.
While a typical AI model might need tens of thousands of pieces of labelled training data, if not more, to hit their performance goals, Mind Foundry’s models can reach the same performance with exponentially fewer. The outputs of these models are then used to help humans make better decisions when it matters most. In defence, for example, Mind Foundry SENTRY can detect, track, and classify drones from the sounds they make.
Mind Foundry’s software layers can be described as “sensor agnostic”, says Lindsay Chadwick, the company’s COO. This means they can use whatever sensor data is available, be it from cameras, microphones, or other sources. “Often, our partners have hardware platforms that have advanced really rapidly, but the software and algorithms haven’t kept pace. That’s what we do: power those solutions.”
Learn more about HSBC’s Tech industry expertise: Technology Sector | Business Banking | HSBC
