Understanding the transformative potential of GenAI

Rarely does a new technology come onto the scene with as much fanfare as generative artificial intelligence (GenAI). The launch of ChatGPT, in late 2022, was a landmark moment that reignited interest in everything related to AI. Despite the popularity of ChatGPT and similar solutions, we are still in the early stages of adoption, and there is uncertainty how GenAI will change the way we live and work.

The promise of GenAI was the focus of a panel discussion at HSBC’s Global Investment Summit’s Private Sessions, which brought together a panel of PhD-level experts to share their vision on a technology that can understand and generate data like no other solution in the past.

“Generative AI creates a very powerful interface that allows you to converse with it in a human way,” said Dr. Sarah Bird, Chief Product Officer of Responsible AI, Microsoft. “We see the potential of the technology to be a bridge to all other technologies.”

While many people are familiar with ChatGPT and how can be used to answer questions, there is more uncertainty surrounding the broader use cases of GenAI. The hope is that the technology will deliver a wide range of productivity improvements – from everything to data modelling to robotics.

“There is great potential for productivity improvements, but to be realised they normally require a change in workflows, and these can take time to come through,” said Dr. Mark McDonald, Head of Data Science and Analytics of HSBC Global Research.

He compares the rise of GenAI with the early days of the Internet, another technology that changed the world. It took a decade before the Internet reached a mature format and companies found business models that allowed them to successfully make money.

Robotics is another area where GenAI could have a dramatic impact. Not only could it create smarter robots for use in manufacturing, it could also drive improvements in the field known as “social robotics”, where robots are used to interact with people in a companionable manner. Robots that can converse like human beings could help assuage the widespread loneliness felt in many societies.

“I’ve seen a lot of very creative solutions to specific challenges in robotics happen behind the scenes,” said Dr. Kate Darling, Expert in Social Robotics, and MIT Media Lab Research Scientist. “And I do think that’s going to build up to something bigger and really push the state of the art forward in robotics.”

The panel discussed the wide range of concerns that are commonly associated with GenAI. Some of the risks come from having too high expectations of what the technology can achieve, said Dr. McDonald.

For example, GenAI can analyse data much quicker than a human being. But at the same time, the computer is not very good at critically assessing its own work, which means that mistakes or unusual results might not get caught without the involvement of a human being.

There are also concerns related to the way that GenAI could be used by malicious actors to create misinformation, harmful code, or make attacks on other computers.

In the cybersecurity, there is already an understanding that GenAI could create a range of new threats. But it also important to acknowledge that GenAI is also a very powerful tool for defence.
Dr. Bird talked about Microsoft’s Security Copilot, which is using AI to help security analysts better understand the threats they face and respond to attacks more effectively.

Another major concern about GenAI is how it will affect the world of work. There are widespread fears that there could an increase in unemployment as people are replaced by intelligent machines.

The panel talked about how employment outcomes will depend on AI is used. But they also made the point that technology is often at its most powerful when it augments, rather than replaces human capabilities – such as when GenAI is used to help a developer programme faster by predicting routine code.

Finally, the panel discussed the emerging regulatory landscape. The tech industry is playing a proactive role in calling for consistency and regulation in the regulations that will govern GenAI, as different geographies different geographies take a variety of approaches.

Regulatory diversity could create complications for global organisations using the technology. Should a company adapt its models to adapt to individual jurisdiction? In this case, a model will act differently in different geographics. Or should it apply the regulations of multiple jurisdictions on its models? But here, the limitations imposed by one jurisdiction will also be in effect in more relaxed jurisdictions.

There are also questions about how regulations are designed for different use cases. For example, a medical regulator will likely have a better understanding of the implications of GenAI on healthcare than a more general regulator. But there may well be the need for rules that apply to all applications of GenAI. It is unlikely that there will be a one-size fits all approach.

“We are looking at our own internal practices and the different shapes of regulation that are needed for the AI supply chain,” said Dr. Bird. “And we are working on sharing this with different governments and organisations our experience of what works and what doesn’t work.”