
[My biggest gripe with AI companies](https://x.com/dltnio/status/1813267142642639201)—and people who seem incapable of thinking about anything *other* than AI—is that they're all building and using products with the implied goal to replace the hard work that's required to do or make something great.

For example, there are a lot of PDF summarization tools being built. After you upload a PDF, these tools will summarize the contents of it and allow you to "chat" with the document. You can say things like, "What's the author's main point?"; "Where does the author talk about supply and demand?"; "Where does the judge define the issue in this opinion?"

 Now, people who get paid to spend a lot of their time reading PDFs to glean unique insights don't have to spend a lot of their time reading PDFs. But if you're rewarded for gleaning unique insights from PDFs, why are you using a tool that attempts to glean unique insights for you?

I like how Readwise [touched on this in a product update email](https://readwise.io/reader/update-july2024) about their AI feature, Ghostreader:

> There might be some ways to complement the reading experience before or after with an AI chatbot, but as far as we're concerned, chatting _while_ reading is a clash. To use a crude analogy, it's like having a conversation while watching a movie or sitting in a lecture. You'd be better off just paying attention in the moment and saving your conversation for after. What's with ​all these "chat with PDF" apps attracting millions of users then? As far as we can gather, none of our users regularly use these products. If there is a use case that retains, it's probably people using those apps as a substitute for reading rather than a reading companion. For example, a kid trying to quickly answer a homework question without reading the textbook or an office worker trying to avoid reading a boring 200-page research report. Those are valid use cases, but not what we're building software for. **Ultimately, we want to maximize quality time spent reading long-form documents, not substitute for it.**

Yes! That's why I love using Reader and supporting the Readwise team[1]. AI tools should maximize the time that's spent doing the hard work of creating something great. "Maximize the time" has two different definitions:

1. *Actually* maximize the time by using AI to accomplish insignificant tasks, thereby giving you more hours in the day to do the hard work that's required to make something great.
   
2. *Metaphorically* maximize the time by giving you different tools to spend every minute of a working block of time that you have productively.

AI tools should *help* you do the hard work that's required to make something great; AI tools should not *replace* the hard work that's required to make something great.

Jack Raines, author of the [Young Money blog](https://www.youngmoney.co/), recently wrote about this too. [In a post](https://www.youngmoney.co/p/the-purpose-of-things-isnt-to-stop), he shared a screenshot from a reader who suggested Jack create a GPT model that "writes in your distinctive style, tone, and format."

Jack didn't like the idea. He explains:

> Anyway, while I appreciate messages from readers, this particular reader missed the point of writing.
> 
> The problem with taking an AI-first approach to tasks is that **it robs you of everything that you would have gained by doing the work yourself.**
> 
> [...]
> 
> I don’t write to simply generate a 1,200 word output. I consider writing to be an extension of my curiosity, and the writing process itself is what turns a rough idea into a finished product. I begin with a vague idea based on some observation of the world, and I put that on paper. As I’m writing that idea, two distant synapses in my brain connect, bridging seemingly-unrelated ideas. Maybe an anecdote from my time playing football relates to risk-taking in financial markets. Maybe a conversation I had at the bar the previous weekend sends me in a new direction entirely. As I continue down this path, the story evolves until it hardly resembles the original idea. **Writing is a metamorphosis that turns vague abstractions into novel ideas, but you have to go through the writing process to connect the various points along the way.**

As I get ready to start law school in the fall, I'm learning how to brief cases and make outlines. Since most law schools teach the same cases and structure their classes in similar ways, there are thousands of pre-made briefs and outlines you can buy online. Though those might be helpful to use as examples, we've been explicitly told *not* to use them and to always make our own. Why? Because the *process* of briefing cases and making outlines is how you learn to read and think like a lawyer. No one is grading you on how perfect your outline for Torts is; they're grading you on whether or not you can memorize and deeply understand the elements required for battery. To memorize and deeply understand the elements required for battery, you have to do the hard work that's required to brief cases and make outlines.

I *could* copy the cases I read into ChatGPT and ask it to make me a case brief for me. But if I want to get great grades (and one day be a great lawyer), I have to learn how to read dense, dry court opinions so I can to identify what's at issue, what the relevant facts are, and so on. The only way I can do that well is if I do the hard work that's required to brief a case.

Creating my case brief *from scratch* and *then* asking AI to help spot problems or over-looked concepts in my briefs is a great strategy to learn the material better. If I have questions about relevant law, I don't have to wait until the next day to ask the professor, I can ask AI to teach it to me. *That's* using AI to maximize the time spent doing the hard work. I'm still actively doing the work that's required to make something great, but I'm asking AI to help me make it even better.

When you use AI to replace the hard work that's required to make something great, you probably will never make something great.

All the benefits of writing blog posts about ideas you're struggling with, or reading abstract PDFs about complicated ideas you want to learn vanish when you outsource writing or reading to something that's not your actual brain.

If you want to make great things, or do great things, *you* have to do the hard work that's required—AI should only help you.

---

1. I think I'm in the minority of people who support Readwise's decision here, because Daniel Doyon, the author of that product update email, [commented on the Tweet that sparked this blog post with](https://x.com/deadly_onion/status/1813679541895979405), "thanks man! glad to hear that resonated with someone 😂"



Do The Hard Work That's Required


One problem I run into a lot while producing *The Knowledge Project Podcast* is making timestamps for multiple versions of an episode.

For example, we have different videos that appear on Spotify and YouTube. The full-length interview is the same, but the YouTube version and Spotify version have different introductions; the Spotify version is much longer because there are ad reads and a longer guest bio.

This proved difficult to manage timestamps because once I had them for the YouTube version, I either had to re-watch the Spotify version for the adjusted times, or do some math in my head to calculate the difference in every timestamp.

But then I realized ChatGPT can help. So here's what I do.

1. I watch the YouTube version of the podcast to create the timestamps.
2. Then I note the difference in timestamps between when the full-length interview starts on both versions. For example, if the interview starts at 01:00 on YouTube and 03:15 on Spotify, the difference is 02:15.
3. I paste the YouTube timestamps into ChatGPT and ask it to add 02:15 to each timestamp and return it to me in the format that's suitable for Spotify: (00:00) Intro).
4. Sometimes I have to prompt it a few more times to get the format right, but the answer it gives are now the timestamps I use for the Spotify version of the podcast.

This seems dead simple—and it is,—but it's been really helpful for me, and when I was trying to figure out the quickest way to do this, I couldn't find much online, so I thought I'd share!

How to make timestamps for multiple versions of a podcast

From a 1955 article written by von Neumann titled, "[Can We Survive Technology?](http://geosci.uchicago.edu/~kite/doc/von_Neumann_1955.pdf)"

> What safeguard remains? Apparently only day-to-day --- or perhaps year-to-year --- opportunistic measures, along sequence of small, correct decisions. And this is not surprising. After all, the crisis is due to the rapidity of progress, to the probable further acceleration thereof, and to the reaching of certain critical relationships. Specifically, the effects that we are now beginning to produce are of the same order of magnitude as that of "the great globe itself." Indeed, they affect the earth as an entity. Hence further acceleration can no longer be absorbed as in the past by an extension of the area of operations. Under present conditions it is unreasonable to expect a novel cure-all. For progress there is no cure. Any attempt to find automatically safe channels for the present explosive variety of progress must lead to frustration. The only safety possible is relative, and it lies in an intelligent exercise of day-to-day judgment [...]
>
> The one solid fact is that the difficulties are due to an evolution that, while useful and constructive, is also dangerous. Can we produce the required adjustments with the necessary speed? The most hopeful answer is that the human species has been subjected to similar tests before and seems to have a congenital ability to come through, after varying amounts of trouble. To ask in advance for a complete recipe would be unreasonable. We can specify only the human qualities required:\
> patience, flexibility, intelligence.

This sounds an awful lot like what AI Alignment has been trying to do. Granted, as [Scott Alexander notes in a book review](https://astralcodexten.substack.com/p/book-review-the-man-from-the-future?utm_medium=email) of von Neumann's biography:

> This sounds suspiciously like the smartest man in the world admitting he's not sure what to do.


technology