Sean McKinven's Newsround: 28 September

Hello.

Much has been written (or not) about AI and writing. In this industry, the mechanics of it has been largely simple: whether or not AI writing is a good or a bad thing, or whether any of it can be resisted. I’ll be quite frank, I have come across many pieces – not including LinkedIn, which is almost too ridiculous to even spend time on – that bear the hallmarks of AI, with greater and lesser (read clinical, over full stopped drivel) success.

Much has also been written about AI and financial advice. From should we use it, to when should we use it, to somebody mentioning the human touch, to won’t someone think of the paraplanners, etc.

The most philosophically interesting bit, however, is something that sort of chugs along in the background – who, exactly, is accountable for it. Beyond even regulated use cases, can someone identify who’s at fault if client Joe Bloggs gets screwed by AI use at some point in the ‘advice’ process.

A piece in the Times Literary Supplement titled ‘AI and the Crisis of the Author’ is one of the most interesting items I have read on the subject of AI for several reasons.

Firstly, it intersects between notions of accountability and authorship in a way that most writing on the subject rarely does. Secondly, due to the background of the author, Anchit Sathi, who could otherwise be described as an academic in comparative literature and former tech executive at the likes of Amazon, presumably making him some degree of evil.

The article begins with an example from his own life. Due to an inheritance dispute he was helping his sister with, he had to deliver a document which synthesised complex laws from two legal jurisdictions. He sent a document over to her husband, who was pleased with it. Then, a few hours later, he enquired how much was written by AI.

Sathi contends that his brother-in-law was not asking out of hostility but highlights the question does expose the central tension that concerns authorship in AI. Initially admired, in the age of AI there is now a doubt about to whom (or what) he should send the admiration, or if he should admire Sathi for producing it at all.

Later, Sathi argues the development of LLMs has put the focus onto the absence of an author and a desire to assign one, citing the case of a New York lawyer who generated arguments using Chat GPT. In the absence of clear authorship, the institutional mechanism made sure the named lawyer and the firm were found responsible, not the chatbot who generated the original text.

He also cites examples from his time in the corporate world. Apparently, Jeff Bezos requires multiple pages of narrative for important meetings which are read in silence for half an hour ahead of discussion. Sathi suggests this is an example of writing that engages cognition, as preconceptions you had are challenged as you compose. Hence, you work harder to iron them out and make your arguments more robust.

He contended that when the meeting begins with a presentation with slides, it allows for a smooth veneer, making it easier to bypass substance.

AI does something similar in this respect. Whereas before, the labour of writing might force you to confront any flaws, an LLM will give you highest probable way to argue and compare the thought, while not necessarily challenging the original thesis. Without adequate rigour, preconceptions or lack of substance are now baked into the final product.

For Sathi, we combat this by recognising the value of thinking that can be lost, and by making sure that, if AI is being used, it cannot bypass that principle.

Applying this to our world, we need to make sure we protect mechanisms for the full development of innovation and debate, not just place the generation of ideas on a pedestal and have AIs talking to each other through various articles.

The News:

LANG CAT & CLIENT MENTIONS

How six providers are launching low-cost advice services

Mike popping up here.

Source: Citywire NMA

Will the incoming stocks and shares ISA cash interest levy impact platforms?

Mark P in this one.

Source: Professional Adviser

Losses from AI giving wrong advice could be ‘just as severe’ as scams

This piece from Saturn.

Source: FT Adviser

Crypto went mainstream. But it didn’t get less risky