Ah me! how hard a thing it is to say What was this forest savage, rough, and stern, Which in the very thought renews the fear.
I feel the way that Dante talked of entering the dark forest and the Inferno. I think we fear AI’s Inferno but sense that it could bring us to “Paradiso” once we’ve spent some time in the purgatory we feel ourselves in now.
Not to push this literary analogy much further, the Pensions Regulator must know what hell looks like and stand against it. I suspect that we will need a Virgil to show us what could happen and guide us to safety. It is too much to expect TPR do offer us all this!
A sensible voice from a younger generation!
I like this post , I feel the way that Estella Bogira feels about the Pensions Regulator trying to take on AI
I don’t know Estella but I tip my hat for her statement. We are mystified , delighted and terrified that artificial could bring so much but take so much away.
Is it fair to land the Pensions Regulator with this? My hope is that this is one that it can pass on to the FCA who have the resource, the people and the political support to make a difference in the regulation of AI
I hope that the Pensions Regulator can step to the side and , rather than take this helpful but frightening monster on, work with others towards an understanding of AI.
We need guidance on how we manage our management of AI but I suspect there is a limit to a regulator’s capacity to give it. Can this be delegated to Stratford where there is also much going on.
This headed a blog from Freshfields two and a half years ago. It itself states it’s out of date.
As Estella Bogira says in her post, we cannot wait for others to catch up!
My feeble grasp of AI can best be explained by this simple picture!
While this is now 2 years out of date. Stephenson Harwood gave a presentation on AI in pension disputes in a pension playpen coffee morning found via https://henrytapper.com/2024/05/08/video-from-stephenson-harwood-on-how-ai-is-impacting-pension-disputes/
That to me suggests that session suggested the use of large language AI models should be discouraged in pensions. With a DB scheme the knowledge base used by the AI model should be restricted to the Deed (after all it is a DEFINED benefit) plus perhaps UK Pensions Legislation, but being careful to distinguish legislative requirements from regulators’ codes of practice which are not prescriptive. The key requirement of any use of AI in a controlled pension environment is that the AI model will always give consistent results in the same situation that has been or can be humanly verified, now matter how the enquiry has been phrased.
It is the uncontrolled use of unrestricted AI tools by the public that causes the greatest risk. For example someone taking their Pensions Dashboard report and asking ChatGP or similar what should I do with my pensions? They at present are likely to get an answer biased by US pension experiences plus a lot of likely investment performance issues focusing on pot size.
I am sure TPR and FCA would like to regulate large language AI models, but that to me doesn’t seem practical.