Three Thoughts on AI

Looking at a couple of the best ways to use AI to see the true value.

Reading about Generative AI (or ‘AI’ as we’ve now circled back to using) has been interesting over the past year. Those testing these tools as designed/built (i.e., as general-purpose tools for help) tend to write optimistically about the technology. Whereas those who have been trying to trip AI up or placing precise demands on these tools have been writing about how overhyped this technology is. It’s rather tricky, for journalists alone, to get a true sense of what this technology is and isn’t.

At its current core, Generative AI tools are good at many things and far from being experts at most. But being pretty good at only one thing can generally mean a human has a solid career in front of them. Pretty good at many things — that’s of high value when it’s a human. This is an excellent way to think about the current tools, as they are not likely to run away with any job of considerable substance. Still, these tools will make parts of everyone’s jobs trivially easy — and we will realize these changes rather quickly.

All of this is to say that generative AI from the likes of OpenAI, Anthropic, Mistral, Meta, or Google is some of the most impressive technology I’ve seen to date. Perhaps more impressive than the tools themselves is how fast this technology is evolving and improving — it is hard to keep track of, and this is most of what I have been working on for the past year.

This is why now seems like a good time to stop and talk about what AI is useful for and where it will disappoint you.

(The caveat to all this is that this technology is changing so quickly, that it is unlikely that this article holds true for long — but should be a good snapshot of where we are now.)

Using It

There are three things you can use AI for now, but only two of them are high-value uses:

  1. To do something you already know how to do well;
  2. To do easy but tedious or slow tasks;
  3. To do something which you currently have no or little idea how to do.

Items 2 and 3 above are of exceptionally high value, today. Item 1 is of meager value and, thus, a more problematic use of the technology as it currently exists. Many writers tend to try and use AI for things they know how to do very well, and thus, their realized experience is generally poor. Only when you use AI tooling to do items 2 and 3 is the inherent value so apparent that it is smacking you in the face.

There’s no comprehensive way (without throwing AI at the task) to expand on each of these, so I’ll give a few examples based on how I find value in the tools or not find value.

Something I Already Do Well

I have yet to get a high-value return by using AI to do something I already know how to do well. Asking AI to review an item or have a cranky opinion on something doesn’t do this as well as I could, or even close to that. In the same vein, I write so much that I get meager returns out of AI reviewing my text for clarity/grammar/etc. I spent months using AI as a copy editor, and while it would catch a few grammatical errors, it was a time-consuming process that didn’t help.

You can see this pattern repeating, as most who write that AI is overhyped are doing so because they are trying to use these tools to do something they already do well. Don’t fall into this trap.

If you write code, don’t ask AI to write that code for you — you can ask it for help working through a bug or explaining something you are unfamiliar with. Reviewing code to explain what certain things do — AI does that well. But it will not come in writing a legal contract better than a lawyer with decades of experience. This rule always has exceptions, but such is true with all rules.

Easy But Tedious or Slow

There are two angles to this use. The first is to task AI with doing something you know how to do, but is rather time consuming for a number of reasons. For instance, I know how to sleuth out good information on the internet, but it absolutely takes dedicated time to do this. Using AI powered search, generally, can net me the same result in a few seconds of my time. The return on this is potentially astronomical — and because I already know how to do this, fact checking the result is easy enough for me that I still am saving time.

Unlike using AI to do something I already know how to do well, the context here becomes: regardless of whether I know how to do this well or not, it will be very time consuming for me to do this. Looking at a 5,000 word document to pull out specific references or inferences of something is a task we can mostly all do, but AI can do it about as well, and significantly fast.

Likewise, summarizing something I might otherwise watch or read, becomes a huge time saver for items which really are only of passing interest to me. Compiling lists of things, or even using AI as a sounding board whereas instead of having to get the time of another human, I can simply utilize AI. Things like asking how AI perceives the tone or content of a message. Asking about how someone might refute a particular idea.

These are all areas where the potential for inaccuracy either do not matter, or would otherwise be trivial for me to catch since I can do this work already. A slow or tedious task then isn’t something that only would take you a long time, but perhaps would add the time of others to get an outside perspective.

Thus, if we go back to my copy editing example, rather than asking AI to copy edit my writing — asking AI what it’s takeaways are, where arguments seem weak — those are things which would require another person with knowledge to read over my work and respond, but AI does it in about 10 seconds.

Something I Don’t Currently Know How to Do

Perhaps the most fun, and potentially high return task, you can do with AI right now is to play around with it to do something you do not know how to do. There are any number of things: music creation, image creation, code creation, video creation, story creation. All of those things, if you don’t already know how to do them, become sort magical that you can, in a mere handful of words, have AI do these things for you.

By way of example, if you ask AI to generate a file list, with the content of each file filled out, which is code for a mobile application — and go on to describe the app you want to build in detail, you’ll get code which (with a few back and forth chats) will very much compile and run. Now, will this replace a developer? No. But does it give someone a way to do something they might never have even attempted to do before? Yes.

That’s kind of wild to think about. In a lot of ways this is where the optimists with AI live. While the total value add here might be debatable, there is no doubt that the potential for massive value exists.

AI Tools Right Now

AI feels like the word processor, or Excel/Lotus123 — there’s no going back now. Those tools did not show up to do something which could have never been done before, they simply made what had been done before easier, and more accessible to all.

That’s where we are, where we are going — well who knows, it’s changing drastically week to week.

Note: This site makes use of affiliate links where and when possible. These links may earn this site money when utilized. 

BECOME A MEMBER

Join Today, for Exclusive Access.


Posted

in

by