Artificial Intelligence (Al) has made significant strides in recent years, transforming industries, improving efficiency, and enabling new technological advancements. From self-driving cars to intelligent personal assistants, Al has demonstrated remarkable capabilities. However, one fundamental limitation that persists is Al’s inability to have its own opinions. This limitation stems from the very nature of Al, its design, and its reliance on data and programming rather than personal experience or subjective reasoning.

Did that first paragraph make sense to you, the reader?

If it did, then we must congratulate ChatGPT, which I asked to write an article about the limitations of Al. This was the unedited version of the first paragraph of its response.

So, I guess the good news is that if you are an opinionated legend in my own lunchtime such as I, then you are safe for a while… until such time as Al learns how to really think original thoughts.

I’ve often seen marketing hype surrounding Al that gravitates towards a belief in their own publicity. Al is fed a gazillion pieces of information from across the web, your own databases, historical experiences and so on, before being asked to pass judgement on the merits (or otherwise) of a claim. Al makes the subjective become supposedly objective in the sense that ‘opinions’ no longer matter and, instead, outcomes are predetermined by the old but still valid dictum that if you put rubbish in, then you ‘get rubbish out’.

Al Rules OK

This newfound objectivity is not necessarily a bad thing. Indeed, it is the very inconsistency of human judgement when presented with the same, or similar, facts that forms part of the Al attraction.

Nowadays, we can expect there to be a huge degree of conformity in decision making that will relieve hard-pressed claims departments of the drudgery and difficulties presented by the demand from customers who nearly always expect positive decisions and prompt settlement.

But woe betide the set of circumstances that is very nearly the same as any other. Or is so close as to be indiscernible except, perhaps, that the policyholders are very different and need different treatment. Or maybe the final outcomes need subjective decision making by the claimant (cash settlement or repair?) but are subtly encouraged by the Al to go down the preferred route as dictated by the insurer. Or perhaps the customer is naturally suspicious of Al driven decisions. Or maybe the Al finds it difficult to put itself in the emotional shoes of the policyholder facing yet another flood in their property…

Final Thoughts

As Al continues to evolve, it is essential to acknowledge and address its limitations. The inability to form opinions may be seen as a drawback, but it also underscores the importance of human-Al collaboration. By leveraging Al’s analytical capabilities while relying on human judgement and values, we can create a more balanced and ethical approach to technology. The future of Al lies in its ability to augment human potential, not in mimicking human traits such as opinion formation.

(By the way – in case you hadn’t realised – the last paragraph was written by ChatGPT. Do I sense the beginnings of an opinion being formed here?)