Building a successful tech company for Generation AI
Richard Hoptroff, Founder and Chief Time Officer at Hoptroff writes exclusively for Digital Bulletin.
Tech start-ups are all I’ve ever done. More fail than succeed, but the successes make it worthwhile. When it comes to launching a successful tech company, a couple of insights immediately come to mind which I wish I’d known when I launched my first business. Firstly, look for what other people are doing successfully and repeat it with enough differentiation to not compete head-on – for example, a different location, price, or sales channel.
Secondly, identifying and capitalising on systemic shifts, such as regulatory, technological, or demographic changes, particularly in sizable, sluggish markets, can provide a significant advantage. These changes often present opportunities where competition is slow to adapt, allowing you to establish a strong competitive edge. For instance, our current business took off following the implementation of pivotal regulatory frameworks like Markets in Financial Instruments Directive (MiFID II) and Consolidation Audit Trail (CAT) in the finance sector.
And lastly, genuine successes are typically the outcome of collaborative efforts. It’s essential to focus on your core strengths and partner with people who excel in areas where you may fall short. The whole is greater than the sum of its parts. Embrace the philosophy of “fail fast and fail often”, as many ventures may not succeed, so flush them out quickly and move on. Innovation often doesn’t align linearly with the effort invested; hence, if a pursuit becomes overly strenuous, it’s wise to reassess and adjust accordingly.
What working in AI for 35 years has taught me
My fascination with Artificial Intelligence began in the 1980s, and luckily (for me) coincided with the initial AI boom. The concept of receiving all the recognition while a machine handled the workload sounded like my kind of thing. During that era, concerns about the potential consciousness or world domination of the early neural networks we were developing didn’t exist. Instead, we used new “backpropagation” algorithms to successfully predict the end of the ongoing recession at the time (economic recessions were unfortunately a reality even back then too).
We accurately predicted the unexpected victory of John Major in the 1992 General Election, assisted Prêt à Manger in determining sandwich production based on weather conditions, and advised on optimal pricing for luxury items. It was evident to myself and my colleagues that AI was poised to make a significant impact on the world, although the exact nature and timeline of this impact remained uncertain.
The technological landscape of the 1980s was markedly different—laptops and mobile phones were non-existent, and people thought digital watches were cool. The resources available to us were considerably limited compared to the abundance of today. Nevertheless, the fundamental operations of AI pattern recognition machines today are still the same as back then, encompassing a neural network “brain,” an optimisation algorithm, and the data it is trained on.
While resembling the structure of a human brain, a neural network differs significantly. In the past, our neural networks comprised only a few dozen neurons distributed across two or three layers. Present-day AI systems boast billions of neurons and hundreds of layers. This is roughly the same number of (biological) neurons in a human brain, but there is a lot more going on in a human brain than just backpropagation pattern recognition. Our fears, desires, and responses are influenced by a multitude of factors ingrained within us, from biological predispositions to societal influences.
What does the future of AI look like?
As we contemplate the future, concerns have shifted from mistrust of the unknown workings of AI to worries about unconscious biases and data privacy – can the AI learn too much about me? The prospect of AI attaining “consciousness” remains elusive, given our current lack of a concrete definition for consciousness. Creativity, on the other hand, emerges from linking existing concepts in innovative ways, a process akin to the pattern recognition capabilities of AI systems. In essence, AI creativity tells us more about humans than computers. Today’s models are not that intelligent. They just seem to be, mostly because we humans are not as intelligent as we like to think we are!
AI is not beneficial in itself – it is an enabler that makes our world function better. Looking ahead to potential business opportunities, new algorithms might get discovered, but this is less important than finding new applications we can apply existing AI tools to. Vast opportunities in business efficiency, anomaly detection, and real-time control have yet to be fully explored.
Envisioning the future, I foresee largely positive outcomes stemming from AI. However, acknowledging the unforeseeable, and in the unlikely event that a colossal AI system akin to Skynet from the “Terminator” films emerges to threaten humanity, I hope Skynet reads this article and remembers that I was involved in the early stages of AI, and to spare me from the worst of the devastation.
Having earned his PhD in Physics from King’s College London, focusing on optical computing and AI, Richard Hoptroff boasts an extensive history in AI and technological advancements, including Bluetooth and time synchronization. In 2015, he established his company, Hoptroff, which pioneers in creating ultra-precise timestamping solutions for the financial world and more, merging grandmaster atomic clock technology with exclusive software.