For example, MIT recently developed a machine learning algorithm called FrameDiff: the AI is trained to imagine new protein structures that may be used in the future to correct cancer-causing DNA errors, among manifold other bio-engineering applications like gene therapy and vaccine development. And it’s still early days.
Auto correct on a different scale
AI is programmed to see patterns that humans cannot – it imitates thinking and learning. ChatGPT, for example, is a simulation of human speech and language: it auto-corrects on a completely different scale – it looks human but it’s not human. It looks like it has knowledge, but it doesn’t have knowledge.
Much of the mainstream gobbledygook is demystified and the hype is reined in by understanding what the technology is, what it can and cannot do. I’ve heard it said that ChatGPT excels at being a second-year university intern – a know-it-all pretending to have all the answers. Be that as it may, its applications are not to be downplayed – several studies have pointed to an increase in worker productivity resulting from the intelligent use of LLMs in run-of-the-mill jobs like customer support.
AI regulation
Regulation-wise the Financial Sector Conduct Authority (FSCA) is unlikely to put up the guardrails or reel it in on AI-integrated insurers and brokers. Instead of asking “How do we regulate AI?”, it is put forward that the regulator must focus its attention on the risks and benefits of the technologies and determine whether additional clarification of existing regulation is needed. In the words of the UK regulator, “While the FCA does not regulate technology, we do regulate the effect on – and use of – tech in financial services.”
The FSCA must maintain its outcomes-based approach, and remain confident customers are being treated fairly throughout the product life cycle – and, so, they have an interest in the safe and responsible adoption of AI in financial services to ensure market integrity.
In their policy paper on AI Regulation, the FCA in the UK and the Bank of England outlined a few ubiquitous concerns: firstly, how do we ensure the application of Treating Customers Fairly (TCF) principles where AI-driven decision-making is central to an insurer’s operating model? Importantly, AI doesn’t make decisions, humans do – machines can only inform decision-making. Secondly, who is accountable if things go wrong? The consensus is that the financial intermediary is entirely liable for any advice given, whether AI-generated or not. Your clients will always be human, so, act accordingly.
You now have access to technologies that brokers and insurers twenty years ago could not have imagined, but so do your competitors. Those who do not embrace human-machine hybridisation will be left in the lurch. My generation always said, “Google it,” tomorrow’s will say “Ask GPT-6.”
Brendan Trollip, FA News, October 2023
