I finished this video and felt the same way I felt reading Hope and Help for your Nerves: seen.

When I talk to myself, there are times that I say unpleasant things to myself. I’ve spent the better part of 20 years trying to completely silence those thoughts.

When I started listening to them and welcoming them, my depression and anxiety improved almost immediately.

If you feel like you say mean crap to yourself and are looking for a way to stop, start with the advice that Karen Faith gives in this TEDx talk. It’s pretty much spot on, with what I’ve experienced.

Aristotle, who had a few things to say about human nature, once declared, "The greatest thing by far is to have a command of metaphor," but academics studying the personification of tech have long observed that metaphor can just as easily command us. Metaphors shape how we think about a new technology, how we feel about it, what we expect of it, and ultimately how we use it.

I love metaphors. I gotta reflect on this idea a bit more.

There is something kind of pathological going on here. One of the most exciting advances in computer science ever achieved, with so many promising uses, and we can't think beyond the most obvious, least useful application? What, because we want to see ourselves in this technology?

Meanwhile, we are under-investing in more precise, high-value applications of LLMs that treat generative A.I. models not as people but as tools. A powerful wrench to create sense out of unstructured prose. The glue of an application handling messy, real-word data. Or a drafting table for creative brainstorming, where a little randomness is an asset not a liability. If there's a metaphor to be found in today's AI, you're most likely to find it on a workbench.

Bingo! AI is a tool, not a person.

The other day, I made a joke on LinkedIn about the easiest way for me to spot a social media post that was written with generative AI: the phrase “Exciting News!” alongside one of these emojis: 🚀, 🎉, or 🚨.

It’s not that everyone who uses those things certainly used ChatGPT.

It’s more like how I would imagine a talented woodworker would be able to spot a rookie mistake in a novice’s first attempt at a chair.

And here I go, using a metaphor again!

There are an unbelievable amount of points Molly White makes with which I found myself agreeing.

In fact, I feel like this is an exceptionally accurate perspective of the current state of AI and LLMs in particular. If you’re curious about AI, give this article a read.

A lot of my personal fears about the potential power of these tools comes from speculation that the LLM CEOs make about their forthcoming updates.

And I don’t think that fear is completely unfounded. I mean, look at what tools we had available in 2021 compared to April 2024. We’ve come a long way in three years.

But right now, these tools are quite hard to use without spending a ton of time to learn their intricacies.

The best way to fight fear is with knowledge. Knowing how to wield these tools helps me deal with my fears, and I enjoy showing others how to do the same.

One point Molly makes about the generated text got me to laugh out loud:

I particularly like how, when I ask them to try to sound like me, or to at least sound less like a chatbot, they adopt a sort of "cool teacher" persona, as if they're sitting backwards on a chair to have a heart-to-heart. Back when I used to wait tables, the other waitresses and I would joke to each other about our "waitress voice", which were the personas we all subconsciously seemed to slip into when talking to customers. They varied somewhat, but they were all uniformly saccharine, with slightly higher-pitched voices, and with the general demeanor as though you were talking to someone you didn't think was very bright. Every LLM's writing "voice" reminds me of that.

“Waitress voice” is how I will classify this phenomenon from now on.

You know how I can tell when my friends have used AI to make LinkedIn posts?

When all of a sudden, they use emoji and phrases like “Exciting news!”

It’s not even that waitress voice is a negative thing. After all, it’s expected to communicate with our waitress voices in social situations when we don’t intimately know somebody.

Calling a customer support hotline? Shopping in person for something? Meeting your kid’s teacher for the first time? New coworker in their first meeting?

All of these are situations in which I find myself using my own waitress voice.

It’s a safe play for the LLMs to use it as well when they don’t know us.

But I find one common thread among the things AI tools are particularly suited to doing: do we even want to be doing these things? If all you want out of a meeting is the AI-generated summary, maybe that meeting could've been an email. If you're using AI to write your emails, and your recipient is using AI to read them, could you maybe cut out the whole thing entirely? If mediocre, auto-generated reports are passing muster, is anyone actually reading them? Or is it just middle-management busywork?

This is what I often brag about to people when I speak highly of LLMs.

These systems are incredible at the BS work. But they’re currently terrible with the stuff humans are good at.

I would love to live in a world where the technology industry widely valued making incrementally useful tools to improve peoples' lives, and were honest about what those tools could do, while also carefully weighing the technology's costs. But that's not the world we live in. Instead, we need to push back against endless tech manias and overhyped narratives, and oppose the "innovation at any cost" mindset that has infected the tech sector.

Again, thank you Molly White for printing such a poignant manifesto, seeing as I was having trouble articulating one of my own.

Innovation and growth at any cost are concepts which have yet to lead to a markedly better outcome for us all.

Let’s learn how to use these tools to make all our lives better, then let’s go live our lives.

I am grateful — genuinely — for what Google and Apple and others did to make digital life easy over the past two decades. But too much ease carries a cost. I was lulled into the belief that I didn’t have to make decisions. Now my digital life is a series of monuments to the cost of combining maximal storage with minimal intention.

I have thousands of photos of my children but few that I’ve set aside to revisit. I have records of virtually every text I’ve sent since I was in college but no idea how to find the ones that meant something. I spent years blasting my thoughts to millions of people on X and Facebook even as I fell behind on correspondence with dear friends. I have stored everything and saved nothing.

This is an example of what AI, in its most optimistic state, could help us with.

We already see companies doing this. In the Apple ecosystem, the Photos widget is perhaps the best piece of software they’ve produced in years.

Every single day, I am presented with a slideshow of a friend who is celebrating their birthday, a photo of my kids from this day in history, or a memory that fits with an upcoming event.

All of that is powered by rudimentary¹ AI.

Imagine what could be done when you unleash a tuned large language model on our text histories. On our photos. On our app usage.

AI is only as good as the data it is provided. We’ve been trusting our devices with our most intimidate and vulnerable parts of ourselves for two decades.

This is supposed to be the payoff for the last twenty years of surveillance capitalism, I think?

All those secrets we share, all of those activities we’ve done online for the last twenty years, this will be used to somehow make our lives better?

The optimistic take is that we’ll receive better auto suggestions for text responses to messages that sound more like us. We’ll receive tailored traffic suggestions based on the way we drive. We’ll receive a “long lost” photo of our kid from a random trip to the museum.

The pessimistic take is that we’ll give companies the exact words which will cause us to take action. Our own words will be warped to get us to buy something we’ve convinced ourselves we need.

My hunch is that both takes will be true. We need to be smart enough to know how to use these tools to help ourselves and when to put them down.

I haven’t used Gmail as my primary email for years now², but this article is giving me more motivation to finally pull the plug and shrink my digital footprint.

This is not something the corporations did to me. This is something I did to myself. But I am looking now for software that insists I make choices rather than whispers that none are needed. I don’t want my digital life to be one shame closet after another. A new metaphor has taken hold for me: I want it to be a garden I tend, snipping back the weeds and nourishing the plants.

My wife and I spent the last week cleaning out our garage. It reached the point where the clutter accumulated so much that you could only park one car in it, strategically aligned so you could squeeze through a narrow pathway and open a door.

As of this morning, we donated ten boxes of items and are able to comfortably move around the space. While there is more to be done, the garage now feels more livable, useful, and enjoyable to be inside.

I was able to clear off my work bench and mount a pendant above it. The pendant is autographed by the entire starting defensive line of the 1998 Minnesota Vikings.

Every time I walk through my garage, I see it hanging there and it makes me so happy.

Our digital lives should be the same way.

My shame closet is a 4 terabyte hard drive containing every school assignment since sixth grade, every personal webpage I’ve ever built, multiple sporadic backups of various websites I am no longer in charge of, and scans of documents that ostensibly may mean something to me some day.

Scrolling through my drive, I’m presented with a completely chaotic list that is too overwhelming to sort through.

Just like how I cleaned out my garage, I aught to do the same to this junk drawer.

I’ll revert to Ezra’s garden metaphor here: keep a small, curated garden that contains the truly important and meaningful digital items to you. Prune the rest.

(Shout out to my friend Dana for sharing this with me. I think she figured out my brand.)

We tried to hang a pretty picture on a wall, but accidentally opened a small hole. This hole caused the entire building to collapse. While we did not intend to create a hole, and feel terrible for all the people impacted by the collapse, we believe it’s also worth investigating what failures of compliance testing & building design could allow such a small hole to cause such big damage.

Multiple parties involved, myself included, are still students and/or do not code professionally. How could we have been allowed to do this by accident?

It’s certainly no laughing matter, neither to the people who rely on npm nor the kids who did this.

But man, it is comical to see the Law of Unintended Consequences when it decides to rear its ugly head.

I applaud the students who had the original idea and decided to see what would happen if you installed every single npm package at once. It’s a good question, to which the answer is: uncover a fairly significant issue with how npm maintains integrity across all of its packages.

But I guess the main reason I’m sharing this article is as a case study on how hard it is to moderate a system.

I’m still a recovering perfectionist, and the older I get, the more I come across examples (both online like this and also in my real life) where you can do everything right and still end up losing big.

The best thing you can do when you see something like this is to pat your fellow human on the back and say, “man, that really sucks, I’m sorry.”

The worst thing you can do, as evidenced in this story, is to cuss out some teenagers.

I gotta find the video from this and watch it myself, because essentially every single thing mentioned in this article is what I wanna build a podcast around.

Let’s start with this:

As Kwan first explained, modern capitalism only worked because we compelled people to work, rather than forced them to do so.

“We had to change the story we told ourselves and say that ‘your value is your job,” he told the audience. “You are only worth what you can do, and we are no longer beings with an inherent worth. And this is why it’s so hard to find fulfillment in this current system. The system works best when you’re not fulfilled.”

Boy, this cuts to the heart of the depressive conversations I’ve had with myself this past year.

Finding a job sucks because you have to basically find a way to prove to someone that you are worth something. It can be empowering to some, sure, but I am finding the whole process to be extremely demoralizing and dehumanizing.

“Are you trying to use [AI] to create the world you want to live in? Are you trying to use it to increase value in your life and focus on the things that you really care about? Or are you just trying to, like, make some money for the billionaires, you know?”  Scheinert asked the audience. “And if someone tells you, there’s no side effect. It’s totally great, ‘get on board’ — I just want to go on the record and say that’s terrifying bullshit. That’s not true. And we should be talking really deeply about how to carefully, carefully deploy this stuff,” he said.

I’ve literally said the words, “I don’t want to make rich people richer” no fewer than a hundred times since January.

There is so much to unpack around this article, but I think I’m sharing it now as a stand in for a thesis around the podcast I am going to start in the next month.

We need to be having this conversation more often and with as many people as possible. Let’s do our best right now at the precipice of these new technologies to make them useful for ourselves, and not just perpetuate the worst parts of our current systems.

There are still going to be situations where someone wants to write prompts that are used at scale, and, in those cases, structured prompting does matter. Yet we need to acknowledge that this sort of “prompt engineering” is far from an exact science, and not something that should necessarily be left to computer scientists and engineers.

At its best, it often feels more like teaching or managing, applying general principles along with an intuition for other people, to coach the AI to do what you want.

As I have written before, there is no instruction manual, but with good prompts, LLMs are often capable of far more than might be initially apparent.

If you had to guess before reading this article what prompt yields the best performance on mathematic problems, you would almost certainly be wrong.

I love the concept of prompt engineering because I feel like one of my key strengths is being able to articulate my needs to any number of receptive audiences.

I’ve often told people that programming computers is my least favorite part of being a computer engineer, and it’s because writing code is often a frustrating, demoralizing endeavor.

But with LLMs, we are quickly approaching a time where we can simply ask the computer to do something for us, and it will.

Which, I think, is something that gets to the core of my recent mental health struggles: if I’m not the guy who can get computers to do the thing you want them to do, who am I?

And maybe I’m overreacting. Maybe “normal people” will still hate dealing with technology in ten years, and there will still be a market for nerds like me who are willing to do the frustrating work of getting computers to be useful.

But today, I spent three hours rebuilding the backend of this blog from the bottom up using Next.JS, a JavaScript framework I’ve never used before.

In three hours, I was able to have a functioning system. Both front and backend. And it looked better than anything I’ve ever crafted myself.

I was able to do all that with a potent combination of a YouTube tutorial and ChatGPT+.

Soon enough, LLMs and other AGI tools will be able to infer all that from even rudimentary prompts.

So what good can I bring to the world?

But now Gemini 1.5 can hold something like 750,000 words in memory, with near-perfect recall. I fed it all my published academic work prior to 2022 — over 1,000 pages of PDFs spread across 20 papers and books — and Gemini was able to summarize the themes in my work and quote accurately from among the papers. There were no major hallucinations, only minor errors where it attributed a correct quote to the wrong PDF file, or mixed up the order of two phrases in a document.

I’m contemplating what topic I want to pitch for the upcoming Applied AI Conference this spring, and I think I want to pitch “How to Cope with AI.”

Case in point: this pull quote from Ethan Mollick’s excellent newsletter.

Every organization I’ve worked with in the past decade is going to be significantly impacted, if not rendered outright obsolete, by both increasing context windows and speedier large language models which, when combined, just flat out can do your value proposition but better.

I ended my first blog detailing my job hunt with a request for insights or articles that speak to how AI might force us to define our humanity.

This op-ed in yesterday’s New York Times is exactly what I’ve been looking for.

[…] The big question emerging across so many conversations about A.I. and work: What are our core capabilities as humans?

If we answer that question from a place of fear about what’s left for people in the age of A.I., we can end up conceding a diminished view of human capability. Instead, it’s critical for us all to start from a place that imagines what’s possible for humans in the age of A.I. When you do that, you find yourself focusing quickly on people skills that allow us to collaborate and innovate in ways technology can amplify but never replace.

Herein lies the realization I’ve arrived at over the last two years of experimenting with large language models.

The real winners of large language models will be those who understand how to talk to them like you talk to a human.

Math and stats are two languages that most humans have a hard time understanding. The last few hundred years of advancements in those areas have led us to the creation of a tool which anyone can leverage as long as they know how to ask a good question. The logic/math skills are no longer the career differentiator that they have been since the dawn of the twentieth century.¹

The theory I'm working on looks something like this:

1. LLMs will become an important abstraction away from the complex math
2. With an abstraction like this, we will be able to solve problems like never before
3. We need to work together, utilizing all of our unique strengths, to be able to get the most out of these new abstractions

To illustrate what I mean, take the Python programming language as an example. When you write something in Python, that code is interpreted by something like CPython² , which then is compiled into machine/assembly code, which then gets translated to binary code, which finally results in the thing that gets run on those fancy M3 chips in your brand new Macbook Pro.

Programmers back in the day actually did have to write binary code. Those seem like the absolute dark days to me. It must've taken forever to create punch cards to feed into a system to perform the calculations.

Today, you can spin up a Python function in no time to perform incredibly complex calculations with ease.

LLMs, in many ways, provide us with a similar abstraction on top of our own communication methods as humans.

Just like the skills that were needed to write binary are not entirely gone³, LLMs won’t eliminate jobs; they’ll open up an entirely new way to do the work. The work itself is what we need to reimagine, and the training that will be needed is how we interact with these LLMs.

Fortunately⁴, the training here won’t be heavy on the logical/analytical side; rather, the skills we need will be those that we learn in kindergarten and hone throughout our life: how to pursuade and convince others, how to phrase questions clearly, how to provide enough detail (and the right kind of detail) to get a machine to understand your intent.

Really, this pullquote from the article sums it up beautifully:

Almost anticipating this exact moment a few years ago, Minouche Shafik, who is now the president of Columbia University, said: “In the past, jobs were about muscles. Now they’re about brains, but in the future, they’ll be about the heart.”

Crypto failed because its desire was to remove humans. Its biggest failure — or was it a feature? — was that when the technology went awry and you needed somebody to step in, there was nobody.

Ultimately, we all want to appeal to another human to be seen and understood — not to a machine running a model.

Interacting with each other is the whole point.