An Interesting Trend with AI

Sam Altman said something that got my attention

I’ve been using ChatGPT and/or GPT-4 for about a year now. I wouldn’t say I use it frequently but it has some great applications. I mostly use it to generate text such as a marketing blurb or a press release. So far we’ve issued two AI generated press releases which took a total of 30 minutes from concept to finish to get ready. That saved a ton of time and headache. It’s become my go-to for research and even learning new programming libraries. It still doesn’t write code well enough to the point I use it but it can give me a peek at how different python libraries work and point me in the right direction. It’s also been integrated into my email client which produces very useful summaries of emails and threads. I’ve used other AI systems to generate cool images including face swapping me out of some famous photos, which was great fun. I don’t use it but if my Slack was busier the new Slack AI looks great too for catching up on conversations.

The thing is, if I thought about it more, there are probably a lot more times I could use AI. Writing emails and letters, helping me code, replying to texts, making travel plans, etc.. I actually rather like the Bumble founder’s proposal that dating happen by having an AI based upon your personality talk to another person’s AI instead of the endlessly frustrating current messaging back and forth. (However, I’m in a very happy relationship and hope I never have to deal with that again). I also am really fascinated with the AI capabilities on Samsung’s latest Galaxy phones. The image editing features are world class and a huge step above what’s available on the iPhone. I’m sure Apple is racing to add similar features as this was the first time in a decade I almost considered switching devices seriously (damn FaceTime you’re addictive).

So, as I do, Sam Altman, the CEO of OpenAI, sees uses for AI everywhere.

What’s interesting though is apart from a few layoffs here and there, AI doesn’t seem to be replacing people yet though. In fact, Sam Altman mentioned in a recent interview that he was actually shocked that “AI did not have a noticeable or significant impact on the economy yet.” I would actually agree with him. Used right, tools like ChatGPT can both dramatically increase productivity or replace some employees outright.

But so far, with the hundreds of billions spent and NVIDIA on track to become the most valuable company in the world, AI hasn’t made a dent.

Like Sam, I find that very interesting.

Many AI experts can’t quite figure out where we are on the AI timeline since there are so many unknowns. AI is going to grow much like the internet did. One recently wondered about whether ChatGPT is the “Napster” moment for AI or if we were already much further along, since we can’t predict the future.

GPT-4o is a massive upgrade I’ve already integrated into some products. It’s significantly faster and the voice conversations you can have with it are amazing. But I went to try the voice feature for the first time and all I did was have a boring conversation about planning my day. I probably won’t revisit it until it’s available to integrate into other platforms.

So what’s going on? With all this amazing new technology, how come it’s not having a larger economic impact?

Well, the first answer is of course it’s too new. People don’t really know what to use it for or even how to properly use it (or don’t know about it). Most of my time with AI has actually been spent creating apps for a client that services needs via text message (it’s actually really cool and people flip at the demos). So I think companies and people are still figuring out where AI fits into everything. So in that sense, I do believe we are still at the Napster stage of AI development and the next few years we will really see it come into its own, probably most importantly with the launch of the iPhone 16 and iOS 18 later this year which will bring AI into the hands of billions of people.

But there’s an even more significant fact. Once people get used to doing something one way, it’s very, very hard to get them to change. I learned that with my very first app. Way back in 2011, I launched SmartFuel, an iOS app with realtime gas prices pulled from every credit card swipe at the pump in the US. We had taken a long time to design a great UI/UX and the reviews were all 5 stars. We even got featured on the front page as the app of the day for the App Store.

Yet our metrics showed people would use it a few times and then forget about it.

Why? Simple.

When you need gas, you don’t think. You just pull into the nearest gas station or maybe a local one you are comfortable at. Even saving 25 cents a gallon was not enough to entice people to remember to pull out their phone and check prices along their route. The behavior was simply too ingrained.

I look at my workflows today. For almost 12 years, I’ve built every piece of software on a single tech stack (Python, Flask, Lambda, Zappa, Algolia, Firebase). I can think of only one time where I started a project on something else entirely and that was because of security concerns (I built banking software). Why do I do this? Because it’s familiar and I can do it very quickly without bumping into lots of unknown issues. That behavior is deeply ingrained in me and I imagine most people are the same with their preferences.

Inserting AI into that process is not something that comes natural. I don’t stop and think, “I should ask ChatGPT about this” while I am working. I have only used it for a few gimmicky things despite it being capable of a lot more.

Changing people’s behavior either takes a long time or is basically impossible and inserting a new tool into workflows is insanely difficult. I suspect this is the main reason we aren’t seeing more people (certainly there are many) adopting AI to streamline their jobs or otherwise impact the economy.

And lastly comes a question regarding a topic that’s very close to my heart. Why would you? If you are assigned 5 tasks to finish by the end of the week at a standard 9-5 job, the smart thing to do is get only those 5 done and make sure the last one is done at 4:59PM on Friday.

Having frequently gone above and beyond the call at many jobs, every time I got work done faster, more work was simply heaped onto me and I never got compensated more. I see discussions and memes about this all over IG and reddit. Being more productive only flows up to your employer’s bottom line and doesn’t help you these days. It’s a serious problem with corporate America and its own huge topic. So why would I use a tool that only benefits my employer? Another thing to think about.

In any case, it is quite interesting that AI has not had a significant economic impact yet. The future is unknown though but it gives some hope we might avoid the mass culling of jobs many AI experts have predicted.

Only time will tell.