AI is the New Web

I didn’t really get web version 1.0. I mean, I understood what it was and how people used it, but I think I was too young to really understand why it was useful. To me, going to a website at the time was just like intentionally driving up to a billboard just to get a good look. Static sites providing information that was eaisier accesseed (for me) by other, less complicated, means. Not many people in my world at the time spent their days parked in front of a computer, not even for work. At this point, for me, computers were most useful for playing Doom.

I was starting high school right around the time the web started changing to v2. I remember being pumped that I’d be assigned an email address by the school district. I think I signed in and checked that inbox exactly one time. I still didn’t really have any use for the internet, but part of it was mine. This was also the year that I successfully managed to convince my dad of the nessesity of purchasing a PC for the family. $1500 for a Dell Dimension 4100 and we could do whatever our school could ever dream up as an assignment for us. I think he got a little suspicious when the first thing I did as soon as the computer was setup, was tie up the phoneline using free AOL hours to play Starcraft.

The thing that finally hooked me on the internet though, was Napster. I still remember the day that my friend Jeff came over all excited and insisted that I install a program on my computer. We loaded it up and he told me to search for my favorite song. 12 hours or so later, thanks to a blazing 2.5kb/s download speed, I had my very own “Metallica - Enter Sandman.mp3” file sitting on my harddrive. A post-web 2.0 person says “so what”, but that meant I could now listen to a song, as long as I had the mp3, whenever I wanted to, for free. As a very broke early teenager, this was unbeliveable. After that, I found message boards aka forums, instant messaging, then social media, and now we’re here.

Where are we, in terms of the web, now that we’re here? Well I’d say, after 20 or so years, we’re stepping into web 3.0. Last year, lots of dudes with too much money and not enough real life expereince, tried to sell everyone on the idea that cryptocurrencies and anything associated with blockchain technology is what we’d all come to realize is the new web. I understand their enthusiasm. Who wouldn’t be excited by the idea of creating thier own currency and then reaping the profits of monetizing everything a user does within my app. Unfortunately for them, everything about the idea falls apart if you think about it for longer than 2 minutes. This is where Large Language Model AIs have come in and said “No so fast motherfuckers!” and set the record straight.

Web 3.0 is about AI. ChatGPT blew the top off of what has been building below the surface for quite some time. Web 2.0 was about user generated content, web 3.0 will be about how AI can be used to understand and manipulate that data. The internet is full of content. For a long time, tools like Google, voting system and ranking algorithms on social media sites helped us sort through this content, hopefully surfacing the good and useful content over the trash. What tools like ChaGPT and Github’s CoPilot do for us is not only sort through the content of the internet to find the relavant information we need, but also puts it into the context of our particular use case for us. (Note: there are many uses for these AI models, mine most apply to software development).

With Google, when I’m trying to find the solution to a problem, I have to first think of how I can state the problem in a “general” way, since its search function would be thrown off by things specific to my personal context. Then, once I find what I think is the correct solution to the generally stated problem I’m looking for, I have to convert that solution back to my particular use case - and figure out how to extend the solution further if needed, on my own. This method works well enough, and is a skill that people develop to help them become better developers. AI tools eleminate 90% of this work. The workflow changes to state the problem, then determine correctness of the answer, ask follow ups.

I’m not sure what the impacts of these powerful new tools being in widespread use will be, but I have definitely found them useful in a variety of ways while working. In a following post I’ll explain some of the ways I’ve used chatGPT and CoPilot for tasks like writing bash scripts, adding comments to code, writing test, and review or refactoring code.