This is about selling solutions vs solving problems.
There's an excellent quote about this common phenomenon from Lant Pritchett (Prof @ Harvard Kennedy School) that has stuck with me for years, and helped improve how I begin a working relationship with my clients:
"A lot of the time when we first interact with people and ask them to come up with problems that they want to solve they often name a solution because they have a preset idea of the solution and hence they never really have thought through to the problem to which this that was a solution"
This doesn't seem to be about improvement nor about algorithms. It's about bolting AI on top of everything (because that's the style at the time) for the sake of being able to tell that it has AI.
From marketing pov it kind of makes sense. If I think about it, if there are two otherwise equal toilet brushes, I will pick the one that has AI too.
Cars are a good counterexample. Low-tech cars have much less troubles. Eventually full-featured ones will also be consistently reliable and secure but we're going through growing pains. Safety features that can be enabled/disabled are a good advancement though.
Software today might be more like cars in the 1950s.
Every year, cars needed to feature new, useless styling to be attractive to buyers. Cars were often replaced annually so that the neighbors could see your taller fins and space-age styling.
This also is the era when seatbelts were known to save lives but were not included in most cars.
I completely agree, I think even in some cases eventually AI should be combined with a basic algorithm, for example chatGPT sometimes gives wrong math answers, which makes sense because it is just generating what it "says" based on it's data. If it could take the equation it finds though and plug it into a simple calculator algorithm then the answer would be correct. Just simple example, but you know what I mean..
Yes, indeed. This push for "AI has to do everything" is dumb. It seems to be the new buzzword and koolaid that has to be attached to every context just because.
The post doesn't claim that AI is useless. It just makes the point that sometimes there are much simpler (faster, cheaper) solutions that are just as effective if not more so.
Sure these are often things AI is capable of doing. However, if using AI provides a worse product, building it takes longer and costs more, and the product is error prone, slower, and or more expensive to use then AI is simply a poor solution. The only argument has nothing to do with the product it’s about padding a resume etc.
My friend is a tax accountant. I hope I am wrong, but I am worried for her entire profession. It seems more about memorization of the massive & enormously complex tax code, and less about actual problem solving. This seems very well suited to an AI. I suspect such professions will be decimated in the near future.
Worry about your friend’s human outcomes of her profession being automated out of existence, but not for the profession of tax accounting. The less people that have to waste their lives parsing through tax code documents, the better, IMO.
Fyi decimation means reduced by 10%. The term comes from executing one in ten soldiers in a unit that refuses to follow orders.
Anyway, I'm curious how an AI would look for someone embezzling funds. There's more to the job than just memorizing calculation pipelines. I know I wouldn't use an AI accountant unless the state accepted liability for its failures.
I sort of agree, but this ignores a more realistic problem. In areas where demand for units (things or services) isn’t likely to grow at least proportionally to price drops, the introduction of a tool which makes one person 3x as productive will drive down wages and force people out of the profession.
There's an excellent quote about this common phenomenon from Lant Pritchett (Prof @ Harvard Kennedy School) that has stuck with me for years, and helped improve how I begin a working relationship with my clients:
"A lot of the time when we first interact with people and ask them to come up with problems that they want to solve they often name a solution because they have a preset idea of the solution and hence they never really have thought through to the problem to which this that was a solution"
Here is a great quick explanation of this from him and how he addresses it: https://www.youtube.com/watch?v=--ewJatFeZU