4 Things You Should Know About AI


A Hero's Journey

Tuesday, October 29, 2024


Hey there Reader,

I'm going to take a break from PushBrand this week (except for another development screenshot and small development update towards the end of this issue) and instead focus on the title topic of AI, or as I like to say, "AI" (cause the name is perhaps the misnomer of the century).

You've undoubtedly heard a lot about AI in recent months (years), with many predictions of doom and gloom on the near horizon. I'm here to tell you that it's all utter horse-shit.

AI is a technological evolution. It's impressive (in specific circumstances), and it will make a significant impact on the world economy. But it's not the revolution many are making it out to be. Allow me to expound...

4 Things You Should Know About AI

1. AI is not what it sounds like.

I prefer to refer to AI as "AI" because the name is just plain incorrect. AI is not intelligent, or at least not in the way you think when someone says something is intelligent.

AI is an evolution of machine learning, a trend you may have started hearing about ten or even twenty years ago (especially if you work in tech). Machine learning has been affecting our lives through the products we use every day for a long time. It's a powerful technology—when made and applied well—and, despite some bumps in the road (social media algorithms and government spying technology come to mind), I'd say it's been a significant net positive for society.

AI is just machine learning under the hood with a better UX, consumer use cases (aka more widespread), better marketing, and cheaper deployment. It's not intelligence. It's just less esoteric machine learning.

Ten or so years ago, machine learning was expensive to build and deploy (or purchase from a vendor), so it was only available to businesses, as they were the only ones who could afford it. As such, the uses of machine learning were very much business-oriented.

The closest it would come to something consumer-grade would be the tech behind self-driving cars (cars being an expensive enough commodity to shoulder the cost of machine learning tech), but as we all know today, that technology still has never proven itself. And it's only been in the last few years that machine-learning-enabled features like adaptive cruise control, lane assist, and parking assist have become mainstream.

Machine learning, aka AI today, is far from being generally intelligent (able to solve widely varying problems with no or minimal input). Machine learning applications are often very smart, but only in a narrow subset of tasks.

So because of the business-oriented nature of AI, erhm, I mean machine learning, it largely flew under the radar (other than when social media companies got into some hot water over potentially allowing machine learning algorithms to dictate your news feed). And that leads directly into...

2. AI did not come out of nowhere.

As much as it may feel like it, given how much press ChatGPT got and how quickly it took off, AI has been a long time coming now. As I said earlier, AI is just a new iteration of machine learning, targeted at different uses. This technology has been actively developed for at least ten-twenty years.

3. AI is not going to take your job.

AI is no more likely to replace your job than any of the other technological innovations of the last twenty years could have. As I've said, this is just another iteration of machine learning, and machine learning has been around for a long time without putting a ton of people out of work overnight.

Instead, AI is more likely to shift the workforce gradually. As more AI technology gets developed (specific applications of machine learning technology focused on new problems), it will shift how some jobs function, as it will become a new tool available to those who wish to use it.

And like any tool, those who use it well will find new opportunities and thrive, and those who don't risk falling behind their peers.

That may seem harsh, but it's no different than any other considerable technological innovation. In a way, the introduction of AI as a copyediting tool, for instance, is to writers/editors what desktop accounting software was to accountants twenty years ago.

AI will bring significant change, just over a long time. There will be plenty of time for you to adapt, if your job even needs to be adapted (many won't).

4. AI doesn't function without human input.

As much as companies like OpenAI want you to believe that their technology is self-learning and self-growing, that's just not true.

Or at least not in the way we imagine when you say something like that to us.

"Self-learning" AI is *marketing*, not reality.

Humans created AI, and it's dependent on us to function. AI cannot write its own code, and I would be shocked if this iteration of AI would ever be capable of such a thing.

Therefore, the moniker of "self-learning" is very misleading. ChatGPT (and its successor GPT-4) are more or less the market leaders in the AI space, and their tech cannot write its own code. Hence, the development of AI is still limited by how fast humans can improve the code.

There's also a significant bottleneck regarding the data these models need. OpenAI has stated that they need large teams of humans to sift through data and ensure it's free of bias and other less-than-great quirks that would throw off their AI responses.

We can't just throw all the data from the internet at an AI model because then that AI would be training on things like hate speech, depictions of horrible acts of violence, and all the other worst, most grotesque aspects of humanity. We must groom and clean up the data before it enters AI models.

If you've only been hearing/seeing the sentiments that AI will revolutionize the world and would like to see the other side, this video is an excellent place to start. I don't agree with everything Conover says, but he does a pretty good job of showing the anti-AI arguments, and there's a lot in there that's pretty convincing. Hell, a good part of this very issue is inspired by this video.

video preview

Bonus: two AI applications I use, like, and recommend.

Duolingo Max (the new premium subscription level for Duolingo) uses AI to power real-language back-and-forth conversations with an AI companion. It's not perfect by any means (it feels a little rigid in its current implementation), but it's a really cool peek into what could be a rather transformative way to learn a new language.

Github Co-pilot uses AI to power intelligent code completion in your repositories. It's quite impressive. And while I'd say I only use maybe 50% or so of suggestions right now, there are many times where it's given me a complete function implementation right away that works and saved me ten minutes or more that it would have taken to grok it out myself. Again, it's not perfect, but it makes my job of coding significantly easier and allows me to get more done faster.

Journey Update

Whew. I've been brewing on this topic for quite some time, and it feels good to get some of my thoughts out of my brain. As you've probably surmised, I have mixed feelings about AI.

On the one hand, I'm doubtful that humanity will ever be capable of creating an AI that's more generally intelligent than us. We're already capable of creating AI that is more intelligent than humans in very specific tasks. But when you think of just how many tasks our brains can handle (virtually unlimited), you realize that the computation required to handle all of that is also virtually infinite (aka impossible). And even if we invented some brand-new technology that manages to accomplish that level of general intelligence, it has to essentially be as good as what took literally billions of years of evolution to form (aka our fleshy brains).

I just don't think we can beat billions of years of evolution on the order of tens or even thousands of years.

On the other hand, I see a lot of value in machine learning and specialized AI (what we call AI today) as new and valuable tools in helping us continue to grow as a species. There are already so many fascinating uses of current machine learning technology: sequencing DNA, detecting fraud in large data sets, speech recognition (helping many people with disabilities access technology through speech), identifying cancer in medical scans, and so much more.

I'm continuing to explore using AI in my own life and career, and I'm sure I'll write more about it in another issue sometime in the future.

But that's enough AI talk for one week. Let's get to a short PushBrand update before wrapping up.

PushBrand Development Status

As of publishing this issue, here are the stats on PushBrand's development:

  1. 84 tasks have been completed, including 3 since last week.
  2. 24 tasks have been planned for v1.0 (slightly higher than my estimate from last week).
  3. There are no more unplanned features as of this week (mini celebration!).

Estimated completion at the current velocity: mid-September. One-two weeks later than last week's estimate.

This week, I planned out the remaining features left in the v1.0 plan. However, in doing so and testing the app, I found a few additional items that have added a bit to the scope. Hence the completion date is getting pushed back slightly.

However, the biggest unknown with PushBrand continues to be the marketing side, as development is progressing fairly smoothly—more on this in later issues.

That's all for this week's issue. Thanks for reading, and you'll see me in your inbox again next week!


If I can ask you to do just one thing, consider sharing this newsletter with a friend or family member. Here's the link so you can pass it along: https://aherosjourney.co/.

113 Cherry St #92768, Seattle, WA 98104-2205
Unsubscribe · Preferences

Millan Singh

I'm writing about my journey to build, learn, and grow as a Creative Entrepreneur and Software Engineer. Stories about my job (I'm employee #2 at a seed-stage startup), entrepreneurial work, investing (a personal passion), and personal growth—a blend of professional and personal content.

Read more from Millan Singh

A Hero's Journey Tuesday, October 29, 2024 Hey there Reader, Hold on a minute. This can't be right. Am I actually writing another issue of A Hero's Journey one week after the last one??? No, that can't be. Let's see, the last issue was published on—checks notes—July 3rd, 2023...and today is...July 10th, 2023. Huh, imagine that... All jokes aside, it has been far too long since I last published two issues back-to-back, and it feels good to do it again. I don't want to jinx myself, but maybe I...

A Hero's Journey Tuesday, October 29, 2024 Hey there Reader, Not going to lie, I feel guilty. And I don't like it. First of all, to the new subscribers who came from Sam Gabrail, welcome! This issue will start on a little bit of a downbeat, but there's some good reflective stuff in here, I promise. So stick with me for two minutes, and I promise it'll be worth it. Also, there's a sneak-peak at a SaaS app I'm working on at the bottom of the issue. I feel guilty, because I said—nay, I...

A Hero's Journey Tuesday, October 29, 2024 Hey there Reader, So this is a little awkward...I just upped and ghosted ya'll for a little over five months (my last issue was on New Year's Eve). But, I have to say, at least I called it: I can already tell I'll face one critical challenge next year: time and energy bandwidth...So the question remains: will I manage to balance all the things, or will A Hero's Journey be a little quiet next year? -- Me, writing to you on New Year's Eve. And the...