There's this really great quote from Steve Jobs that I suspect most people reading this blog have heard, but I'll repeat it anyway since it's so good:

For me, a computer has always been a bicycle for the mind, something that takes us far beyond our inherent abilities.

It's really one of the all-time great quotes about computers, isn't it? Computers have been able to amplify and accelerate humans' ability to do so much. I don't think it's too controversial to say that computers have radically changed our world in a remarkably small amount of time. As a very pedestrian example, when I was in grade school, I was told I would have to write cursive because I wouldn't always have a computer with me to type on. And I was told I had to know advanced multiplication because I wouldn't have a calculator with me at all times. Of course, knowing how to write freehand is useful, as is the ability to understand math. But I do 99.999% of my writing on a keyboard, and if a math problem is too hard for me to solve in my head, I can easily use a calculator more powerful than anything any of my teachers dreamed of in the palm of my hand.

I just read this very good piece called How to Think by Kevin Wammer today, and while I don't agree with everything in it, I did think it was a good read from someone grappling with how much of their thinking they're willing to hand off to a computer. From that piece:

Since the rise of Generative AI, what I caught myself doing is using tools like ChatGPT or Claude to go through problems. Not as a help, but instead had it spit out an answer that I then (at times blindly) adopted as my own solution.

People who despise LLMs with all of their being will nod along in agreement that, "of course using an LLM is a rejection of thought." And to a point I appreciate that perspective. I've heard people suggest that the obvious future of marketing and commerce is for AIs to generate marketing materials that will be viewed by AI bots that will buy products for you. Some box will just show up on your doorstep and I guess I hope you like it!

But on the other hand, so much of computing is about making us think less about stuff. Can I do $67,425 times 3 minus 4%? Of course, give me some paper and I can figure this out for you in a minute. Give me two minutes and I could probably work it out in my head (3 minutes 14 seconds, actually, the percentage took most of the time). But it's very easy for me to just type that into my computer and get the answer right away. I didn't have to think, it just "spit out an answer that I then (at times blindly) adopted as my own solution."

My point here isn't to say that chat GPT and a calculator are the exact same thing, but I do think there is an argument to be made that LLMs are calculators for words. We still teach kids math and there's value in having a base understanding of mathematics. But a calculator is very handy in figuring out things in your daily life. Similarly, I still think people need to know how to think and how to research and how to understand the world around them, but I think that there are benefits to using LLMs to accomplish things as well. We have a calculator for a new category that we've never had before from computers.

Here's a silly example from just today. I was talking to someone at work and I was recalling a certain sort of style trend that I noticed on magazine covers about 20 years ago and I couldn't really describe it well or recall the name of what this was but I could describe it in plain language and I used ChatGPT's search functionality to find this. Here's what I entered into ChatGPT:

What was that design trend from like 20 years ago where you would want to have a slight imperfection in your graphic design? Like a red dot or something on top of an otherwise pristine design? It was a trend in magazine covers for a while

That's not exactly a traditional Google search, but it works with a large language model, and it did get me the result that I was looking for. If you're curious, I think the trend started with Wired Magazine. Here's an article from 2014 explaining this origin. For what is worth I did type this into Google as well. And of course it generated one of their AI overviews, which had a decent answer. The normal search results didn't quite have what I wanted. I tried Kagi for a more purely traditional search engine and struggled there too.

I'll wrap it up here, but I just wanted to write this piece because I think it's important to recognize that there is a vast middle ground between "I use this computer to do all of my thinking for me" and "I will never use this computer because I want to think." I appreciate that tools like ChatGPT create new opportunities for people to offload more thinking than they maybe should, but in a world of Excel, spellcheck, calculators and everything else that lets us offload thought to computers, I don't think it's fair to say that LLMs are unique in this regard. To bring it back to calculators, people were very concerned with graphing calculators in the 90s, with some statements around thinking that sound a hell of a lot like what people are saying today with LLMs:

''There is no work involved, you just run the program, plug in the known values, and watch your answers appear,'' says their Web site (members.aol.com/highersat). The brothers are charging $35 for the programs and guarantee that a student's score will rise by 160 points or more.

And:

But once he is convinced that he understands the math, he simply programs his calculator to do the number-crunching for him. ''In math class a lot of the homework is routine, you can do it over and over,'' he said. ''If you know it, you can whip out your calculator, write a program and press Enter a couple of times, and you've got your answer.''

Again, this isn't me saying anything goes, this is just me saying that there are tools that can be useful and that can be misused to offload too much thinking and we've been navigating that balancing act for years.