a.i. a.i. a.i.

don’t read that superb owl prediction!

i must admit upfront that i am a generative a.i. skeptic, both from a is-this-the-end perspective (it’s not) and from a let’s-replace-humans perspective (let’s not). we are undoubtedly seeing companies cut extra jobs right now because they think generative a.i. will 1.) continue to improve as a piece of technology and 2.) serve as a decent stopgap for certain roles in the short term while they try to please shareholders. it seems like a good time to remind everyone that no company ever cut its way to growth.

from a small business perspective, however, the experience has been positive. for me, the question to ask the a.i. machine for help with is related to what i don’t know much about, or things i know a decent amount about and want to get faster at getting to an answer. for instance, i have spent lots of time in my career in excel doing data analysis. robust data analysis is an important part of my pricing strategy consultancy, and generative a.i. has improved my efficiency because it makes it far easier for me to leverage python and r coding languages for better and faster execution. once i understood the basics of these languages, it has been easier to use an a.i. prompt to quickly get code for the questions i am asking of the data. i’ve been threatening to learn these coding languages for years now, and generative a.i. created a shortcut so i could get there.

there are other people who have had similar experiences, but on the writing side. one’s use of generative a.i. almost certainly varies by how you see your own strengths and weaknesses. the interesting question is when the stakes become too high to trust outputs from l.l.m.s. there’s no risk in leveraging these tools to get quick answers to mundane questions, write four lines of code, or to help you quickly outline a few slides for a presentation. how many millions of dollars need to be on the line for a company to make sure a human steps in? or are we comfortable with a human that has less expertise just supervising the model?

there are still lots of questions to be answered.