As much as I bemoan all the hype about what we can expect out of AI, I did recently purchase a subscription to one of the big ones. I did so because I desperately needed to organize some information that required a lot of data and a lot of questions.
My experience in getting what I want from AI has alternated between moderately useful/timesaving and underwhelming. It seems to be really good at getting pretty close to results that would normally be produced by a roomful of grad students crunching numbers, or just trying different stuff to see what sticks.
This, BTW, strikes me as one of the wobbliest promises of AI at the corporate level. [Whether the staffing and productivity efficiencies are realized long-term is an open question: my view is in a few years everyone will be massively hiring again].
But just as often, I find it mostly frustrating to expect anything from AI. And the admittedly few instances I've actually spent a lot of time trying to make it useful, I end up close to where I want to be, but never quite there. And when I get closest, it starts fabulating things so that all distinctions between form and substance are lost. I end up feeling like I've just lost an argument with a 4th grader.
@briankrebs Yep. In my experience, AI is almost never a one-stop solution. It's great at enhancing many processes and making them more efficient, but there are exceedingly few jobs it can perform end-to-end.
@LouisIngenthron @briankrebs It doesn't necessarily need to do the whole job. In Brian's example, instead of a roomful of grad students, you would just need 2 students and some AI. In the corporate world, instead of a whole team of writers, data analysts, or programmers, you will only need half as many to check the work of the AI and refine it. Always remember, technology always gets better, never worse, so in a few years it may very well be able to perform jobs end-to-end.