The tech industry is never one to shy away from grand declarations and the ushering in of new eras.
When the launch of ChatGPT opened the world up to the possibilities of artificial intelligence, many proclaimed society was approaching its next industrial revolution, a fundamental change in the way humans live and work.
That was over two years ago and while today the lives of the average person don’t yet feel upended by access to clunky customer service chatbots and hit-or-miss generative tools, governments and CEOs maintain that the supercharged economic growth brought on by AI is just around the corner.
This week, France unveiled a €109bn AI investment during the Paris AI Action Summit in hopes of keeping pace with international competition including the US, which recently unveiled its $500bn Stargate AI investment strategy.
In the UK, the institutional confidence in the transformative power of AI, whether affected or not, has become so strong that the Labour government seems to be counting largely on the technology coming good for its key mission of growth within the next five years.
For those struggling to see past the gap in the current use cases of generative AI and the prosperity-bringing,era-defining power lauded in speeches, a recent breakthrough may offer confidence that AI-driven growth is possible.
While the word is thrown a lot in quantum circles, last week, Anglo-American tech company Quantinuum announced a “breakthrough” new system to train generative AI models using data from quantum computers.
Large language models (LLMs) as we know them today generally get more powerful and accurate the more data is provided to them.
Developers feed these LLMs masses of data to generate outputs like summarising documents, writing copy and stitching together images.
Not bad, but perhaps not moving the global economic needle all that much. No matter how much information is fed into these models, they will always be limited to the categories of data that are available to them.
But those categories become infinitely broader when quantum computers are brought into the mix.
Qunatinuum CEO Raj Hazra offered the example of estimating the ground state energy – or most stable state at the lowest energy levels – of the interaction between two complex molecules.
“There is no known classical chemistry technique to estimate what the ground state energy is,” Hazra said.
The upshot is that while generative AI models can make educated guesses about the interactions of molecules, with quantum-generated data, they can accurately characterise this molecular behaviour.
This can translate into some very specific uses – Hazra described using quantum generative AI to calculate an effective and cheaper alternative to platinum as a catalyst in automobiles – but the implications are wide-reaching.
A quantum-backed AI model could accurately predict what can happen to individual molecules when variables such as temperature are changed, translating into incredibly powerful capabilities.
“If you look at most uses of ChatGPT, or things like that, it is still the construction of language,” Hazra said.
“I’m not saying these are not useful, they have good productivity tools. But what quantum-generated data can do is open up multi-billion dollar use cases, like the design of a new drug, like the design of a new chemical for refrigerants that is more green, the transformation of hydrogen as a fuel from an idea to reality at scale.”
Looking at the areas the government hopes will be improved by AI – in particular, health and finance – a major barrier has been that conversational LLMs are not reliable enough to be part of a regulated industry.
A bank could feasibly use a generative AI chatbot for customer service, but it is not going to trust that same technology to manage a hefty portfolio.
“Until we get to this quantum-inspired and quantum-based representation of knowledge in a network, we won’t probably get there,” Hazra said.
Quantinuum was formed in 2021 through a merger between British company Cambridge Quantum Computing and American company Honeywell Quantum Solutions and remains jointly headquartered in Cambridge and Colorado.
It has become a titan of the quantum sector, with a 2024 investment of $300m from JP Morgan Chase and Mitsui Global Investment among others, valuing the firm at $5.3bn.
There are, of course, plenty of companies innovating in quantum computing and AI, but Quantinuum internally feels well-placed to lead.
And why is that? Ask Hazra himself and he’ll tell you that the machinery and software his company has are the most powerful of their kind in the world.
“That’s not just us saying it. It’s been proven by benchmarking, by neutral third parties.”
When asked if Quantinuum could leverage its position at the helm of the incorporation of quantum computing into generative AI to join the exclusive club of global companies worth over $1tn, Hazra said it was possible.
Pointing to the success of Ozempic, the weight-loss drug that put Novo Nordisk in the running for Europe’s most valuable company (worth over $300bn), Hazra argued that if one drug can do that for a company, imagine the value of the provider of technology that could support the discovery of thousands of new drugs as well as hundreds of other use cases.
The post Quantum generative AI: The next ChatGPT moment? appeared first on UKTN.
