Daniel Miessler's Blog, page 35

How Good is DALL·E at Creating NFT Artwork?

If you’ve not heard, there are these things called NFTs. I think they’re simultaneously the future of digital signaling and currently mostly hype. But whatever—that’s not what this post is about.

Most NFTs rotate around a piece of collectible art in a baseball card-like format. So you look at something like the Bored Ape Yacht Club, and it’s a bunch of personalized apes with stylization.

An actual bored ape

So the art is a big piece of it. You still need to hype it, get people to believe in its value, etc. But the art is the starting point, and many artists are finding new life in creating these art collections.

So I’ve been playing with DALL-E 2 for a few weeks now—which is an AI project from OpenAI that generates customized art based on prompts. I’ve been telling everyone who’ll listen about how this is the clearest example we’ve ever seen of AI cutting into a previously-human domain of creativity. It’s one thing to do spreadsheets faster, but DALL·E 2 can make you multiple examples of “A Synthwave Dog Riding a Rocket Skateboard” in a matter of seconds.

So I’ve been thinking a lot about how this will directly affect human jobs in the artistic space, but I was just jostled awake by an idea of telling DALL·E 2 to make NFT art. Like, specifically NFT art.

So I started with:

an nft of an 80’s giraffe wearing headphones, digital art

Which was decent:

My first attempt at NFT art using ‘digital art’ as a prompt

The trick is prompt it with the text you would likely read in the caption describing it.

But within a couple of iterations I got to this:

an nft of a steampunk giraffe wearing headphones, digital art

At a glance those look identical to high-quality NFT art. Importantly, they’ve captured the “NFT-ness” element to the artwork, which you would have trouble even trying to describe. I’m guessing it has something to do with the gradient background combined with a type of quasi-reality.

Anyway, the point is—that’s precisely what DALL·E 2 figures out and does effortlessly! It figures out the Je Ne Sais Quoi of NFT-ness—whatever that is—and applies it to content prompt you give it. Let’s try robots with a similar prompt:

an nft of a steampunk robot wearing a turtleneck and an eye patch, digital art

The ‘digital art’ prompt at the end is key to the realistic look and feel.

Join the Unsupervised Learning CommunityI read 20+ hours a week and send the best stuff to ~50,000 people every Monday morning.

NFT robots, Attempt 1

Here’s another pass at something similar:

an nft of a steampunk robot with an eye patch, digital art

Attempt 2 for Robot NFTs

I’m not an expert in the space, but I have to think you could crank through a few iterations (I’ve spent like 6 minutes doing this so far) and come up with some truly stunning results.

Thoughts

So I think the short version of this is that—yes—DALL·E 2 can already generate pretty convincing NFT art. And that feels weird.

You have this moment with NFTs where quirky artists suddenly have a spotlight shown on them. And out of nowhere DALL·E can suddenly—and casually—do a lot of what they can do.

Some think we’re in an AI winter still, but I feel it’s more like outer space where a piece of metal facing the sun will be super-heated, while the backside is deeply frozen.

GPT-3, DALL·E, and Google’s Gato represent the sunlight where the theory collides with practicality. Exciting for sure—but in a scary way.

Anyway, if you find any prompts that create even better NFTs, let me know.

NotesThere’s one aspect of DALL·E where humans still have an advantage for things like NFT collections, and that’s in making a full set of related art. The problem with DALL·E is that if you iterate 20 times you’ll get widely different results, and it won’t be easy to sell the idea that they’re part of the same collection.Also note that creating NFT artwork to sell is specifically prohibited by the beta program’s terms and conditions.

How Good is DALL·E 2 at Creating NFT Artwork?

If you’ve not heard, there are these things called NFTs. I think they’re simultaneously the future of digital signaling and currently mostly hype. But whatever—that’s not what this post is about.

Most NFTs rotate around a piece of collectible art in a baseball card-like format. So you look at something like the Bored Ape Yacht Club, and it’s a bunch of personalized apes with stylization.

A typical example of bored ape artwork

So the art is a big piece of it. You still need to hype it, get people to believe in its value, etc. But the art is the starting point, and many artists are finding new life in creating these art collections.

So I’ve been playing with DALL-E 2 for a few weeks now—which is an AI project from OpenAI that generates customized art based on prompts. I’ve been telling everyone who’ll listen about how this is the clearest example we’ve ever seen of AI cutting into a previously-human domain of creativity. It’s one thing to do spreadsheets faster, but DALL·E 2 can make you multiple examples of “A Synthwave Dog Riding a Rocket Skateboard” in a matter of seconds.

So I’ve been thinking a lot about how this will directly affect human jobs in the artistic space, but I was just jostled awake by an idea of telling DALL·E 2 to make NFT art. Like, specifically NFT art.

So I started with:

an nft of an 80’s giraffe wearing headphones, digital art

Which was decent:

My first attempt at NFT art using ‘digital art’ as a prompt

The trick is prompt it with the text you would likely read in the caption describing it.

But within a couple of iterations I got to this:

an nft of a steampunk giraffe wearing headphones, digital art

At a glance those look identical to high-quality NFT art. Importantly, they’ve captured the “NFT-ness” element to the artwork, which you would have trouble even trying to describe. I’m guessing it has something to do with the gradient background combined with a type of quasi-reality.

Anyway, the point is—that’s precisely what DALL·E 2 figures out and does effortlessly! It figures out the Je Ne Sais Quoi of NFT-ness—whatever that is—and applies it to content prompt you give it. Let’s try robots with a similar prompt:

an nft of a steampunk robot wearing a turtleneck and an eye patch, digital art

The ‘digital art’ prompt at the end is key to the realistic look and feel.

Join the Unsupervised Learning CommunityI read 20+ hours a week and send the best stuff to ~50,000 people every Monday morning.

NFT robots, Attempt 1

Here’s another pass at something similar:

an nft of a steampunk robot with an eye patch, digital art

Attempt 2 for Robot NFTs

I’m not an expert in the space, but I have to think you could crank through a few iterations (I’ve spent like 6 minutes doing this so far) and come up with some truly stunning results.

Thoughts

So I think the short version of this is that—yes—DALL·E 2 can already generate pretty convincing NFT art. And that feels weird.

You have this moment with NFTs where quirky artists suddenly have a spotlight shown on them. And out of nowhere DALL·E can suddenly—and casually—do a lot of what they can do.

Some think we’re in an AI winter still, but I feel it’s more like outer space where a piece of metal facing the sun will be super-heated, while the backside is deeply frozen.

GPT-3, DALL·E, and Google’s Gato represent the sunlight where the theory collides with practicality. Exciting for sure—but in a scary way.

Anyway, if you find any prompts that create even better NFTs, let me know.

NotesThere’s one aspect of DALL·E where humans still have an advantage for things like NFT collections, and that’s in making a full set of related art. The problem with DALL·E is that if you iterate 20 times you’ll get widely different results, and it won’t be easy to sell the idea that they’re part of the same collection.

News & Analysis | NO. 334

This is UL Member Content

Subscribe

Already a member? Login

No related posts.

How Good is DALL·E 2 at Creating NFT Artwork

If you’ve not heard, there are these things called NFTs. I think they’re simultaneously the future of digital signaling and currently mostly hype. But whatever—that’s not what this post is about.

Most NFTs rotate around a piece of collectible art in a baseball card-like format. So you look at something like the Bored Ape Yacht Club, and it’s a bunch of personalized apes with stylization.

A typical example of bored ape artwork

So the art is a big piece of it. You still need to hype it, get people to believe in its value, etc. But the art is the starting point, and many artists are finding new life in creating these art collections.

So I’ve been playing with DALL-E 2 for a few weeks now—which is an AI project from OpenAI that generates customized art based on prompts. I’ve been telling everyone who’ll listen about how this is the clearest example we’ve ever seen of AI cutting into a previously-human domain of creativity. It’s one thing to do spreadsheets faster, but DALL·E 2 can make you multiple examples of “A Synthwave Dog Riding a Rocket Skateboard” in a matter of seconds.

So I’ve been thinking a lot about how this will directly affect human jobs in the artistic space, but I was just jostled awake by an idea of telling DALL·E 2 to make NFT art. Like, specifically NFT art.

So I started with:

an nft of an 80’s giraffe wearing headphones, digital art

Which was decent:

My first attempt at NFT art using ‘digital art’ as a prompt

The trick is prompt it with the text you would likely read in the caption describing it.

But within a couple of iterations I got to this:

an nft of a steampunk giraffe wearing headphones, digital art

At a glance those look identical to high-quality NFT art. Importantly, they’ve captured the “NFT-ness” element to the artwork, which you would have trouble even trying to describe. I’m guessing it has something to do with the gradient background combined with a type of quasi-reality.

Anyway, the point is—that’s precisely what DALL·E 2 figures out and does effortlessly! It figures out the Je Ne Sais Quoi of NFT-ness—whatever that is—and applies it to content prompt you give it. Let’s try robots with a similar prompt:

an nft of a steampunk robot wearing a turtleneck and an eye patch, digital art

The ‘digital art’ prompt at the end is key to the realistic look and feel.

Join the Unsupervised Learning CommunityI read 20+ hours a week and send the best stuff to ~50,000 people every Monday morning.

NFT robots, Attempt 1

Here’s another pass at something similar:

an nft of a steampunk robot with an eye patch, digital art

Attempt 2 for Robot NFTs

I’m not an expert in the space, but I have to think you could crank through a few iterations (I’ve spent like 6 minutes doing this so far) and come up with some truly stunning results.

Thoughts

So I think the short version of this is that—yes—DALL·E 2 can already generate pretty convincing NFT art. And that feels weird.

You have this moment with NFTs where quirky artists suddenly have a spotlight shown on them. And out of nowhere DALL·E can suddenly—and casually—do a lot of what they can do.

Some think we’re in an AI winter still, but I feel it’s more like outer space where a piece of metal facing the sun will be super-heated, while the backside is deeply frozen.

GPT-3, DALL·E, and Google’s Gato represent the sunlight where the theory collides with practicality. Exciting for sure—but in a scary way.

Anyway, if you find any prompts that create even better NFTs, let me know.

NotesThere’s one aspect of DALL·E where humans still have an advantage for things like NFT collections, and that’s in making a full set of related art. The problem with DALL·E is that if you iterate 20 times you’ll get widely different results, and it won’t be easy to sell the idea that they’re part of the same collection.

Just Copy What Works

I’ve had an idea lingering for years about habits and behaviors and outcomes. If we accept that peoples’ output usually comes from their inputs, what if we just completely copied their inputs?

For example, I’m a heavy guy because I eat too much. I have a friend who eats way less. He’s very thin.

So here’s the crazy part: What if I just ate what he ate?

No magic. No plan. No philosophy. Just copying what demonstrably works for someone I know directly.

Every thin person I know eats in a similar way, and same with heavy people.

In this case, he eats like 6 times a day, with lots of salad and cereal, and some occasional meat added in. It’s also very small amounts. Like when we eat out he orders less and only eats half. He always takes half to go.

I’m quite sure that if I just did that, I’d be thin too. Or at least not heavy.

Or let’s say someone is a total badass at making hacking videos on YouTube. And let’s say we want to be like them.

Well maybe we don’t ask for ideas, or tips, or strategies. What if we instead ask for their schedule, and their calendar. In other words, their daily routine. And what if we simply mapped our day to copy their exact activities?

I bet we’d make some cool shit.

Now, this is probably sitting wrong in your skull. You might be thinking:

Well if it were that fucking easy everyone would be doing it, andBollocks, that sounds too easy

Yes, and yes. And that’s where I found the actual value in this whole thought exercise.

It’s not hard to know what to do. What’s hard is actually doing it.

Let’s think about that, because it sounds like fortune cookie seminar bullshit. The new idea here is realizing which part exactly is hard.

When we look at someone doing something successfully we tend not to think about their daily routine, which they’ve been doing for days or weeks or decades. What we look at instead is the output, and we then look at our own lives and say, “Wow, I could never do that.”

But what if we don’t need to think about their output. What if we only think about their input? And what if it were trivially easy to figure out what those inputs are? Why? Because they talk about them. Or they wrote them down. Or you can just ask them.

Here’s a great example. Writing. Let’s say a writer named Stephan Kring captures his daily routine on his blog, or on a podcast, and it says:

Start with coffee at 7:30 AM.Go for a long walk with no technology on me whatsoever.Eat 14 peanuts and have another cup of coffee.Sit down and write at 8:30 AM, no matter what. Keep writing until 11:30 AM.Have lunch at 11:30 AM and don’t worry about writing for the rest of the day, unless I’m in a groove.The key to my success is to write every day, no matter what, between 8:30am and 11:30am. No exceptions.I’ve done that for 38 years and I’ve written 39 books.

Cool. What are we to take from this?

The old me took from this that, “Hard work pays off.” “Consistency matters.” “99% is showing up.” And a whole bunch of other cliches.

Fuck all that. The only thing that matters is what Stephan yelled at the summit of his voice (which many other successful writers also echo) that you have to set time aside every day to write. Period. That’s it.

Join the Unsupervised Learning CommunityI read 20+ hours a week and send the best stuff to ~50,000 people every Monday morning.

So let me ask you this. How many people do you know, who are unsuccessful at something, have copied the actual inputs from someone who is successful? Here are some examples:

WritingPracticing an instrumentLosing weightInstilling discipline in a household with kidsLearning to paintGetting 90%+ of a high school to go to college

What percentage of people who are trying to make those things happen have actually outright copied the behaviors of those who are doing it well?

In my experience the answer is virtually none.

They might copy some parts, as an add-on, as an experiment, with very little follow-through. But they don’t just say, “We’re going to do things that way from now on because it clearly works.”

Now, you might be thinking, “Well, that’s because it’s not always easy—or even possible—to copy certain things.” Sure. Absolutely. Like if you work nights you can’t wake up at 6:00am and do yoga. Or if you live in a bad part of town, and the schools are crap, and you have three jobs trying to feed your kids, it’s hard to get your 2 hours of solitary meditation in.

But here’s the thing, and this is why I think this idea is valuable: Let’s call out those obstacles as the actual obstacles.

Let’s acknowledge that the primary way to succeed is to perfectly emulate those who are succeeding. Copy everything. We don’t know what works. Use the same damn toothbrush. Walk the same on the sidewalk. Copy what works. And if life prohibits us from doing that, then that’s what we need to change.

For personal lifestyle changes, like someone with a good job who wants to become a writer, it’s all about making the required lifestyle sacrifices to write every day. Does that mean you give up watching TV? Playing video games? Working too much at your 9-5 job? You might need to make some hard choices.

But for bigger issues, like schools and school districts not graduating kids who can read, write, and do math—we need to break it into two pieces:

Do you have a model of what works in schools where the kids are literate and numerate when they graduate? And are you willing to put those policies and standards and expectations into effect in your schools? If so, you are set up for success.Do you have the resources to make that happen? Do you have the books, the teachers, the public transit, the school lunches, etc., that would allow you to actually implement those policies? If not, that’s what we go to war for. That’s what society owes those kids.

But what we can’t do—either as an individual or as a system—is start by looking at different outcomes and assume it’s due to something outside of your control. You have to start with the inputs. You have to start with the routine. You have to start with the behaviors.

Just copy what works.

How Much Incel Terrorism Can We Prevent With Kindness?

When I think about violence caused by young, socially-rejected males, I often wonder how much bullying and mistreatment cause their violent behavior. That doesn’t mean excuse the behavior; I am talking about the proximate cause—or exacerbation—that contributes to the act.

I have a model in my mind that captures this, which goes something like this:

This doesn’t imply a “normal” family will make a well-adjusted kid.

Family structure in the US is highly broken, with many children living in single-parent or grandparent-raised households that didn’t convey enough stable love during upbringingThis causes social problems in a lot of kids raised in those situations.Those kids tend to suffer economically as wellThis combination of lack of socialization and economic status often leads to relentless bullying by kids all throughout school.The combination and bullying often causes trauma that limits their attractiveness and options to potential romantic partners.This often results in the kid then having few friends, no romantic options, and basically living life in hell from childhood to early adulthood.This then results in some percentage of those boys/young men becoming angry and attracted to negative and even violent narratives about women, e.g., the “incel” phenomenon.Becoming involved in that type of culture sometimes leads to the commission of violent acts, like the one described in this Hot Yoga Incel Violence Report by the Secret ServiceSince it’s really hard to fix the broken families part of this equation, our best opportunity to address this might be through marketing/education campaigns with students and young people, administered through schools.

In short, our best way to fix this might be teaching kids to be kind to shy and anti-social kids—similar to how we teach them to get under their desks for earthquakes, or—unironically—how to react to school shooters.

What if we had a mandatory, country-wide curriculum that basically said:

Some kids aren’t as lucky as you. They’re not as good looking, or athletic, or maybe they grew up with trouble in the family, or with less money.Those kids need more love from their fellow students than othersIf you see someone being bullied, or just being extremely quiet, be niceBeing kind can not only prevent someone from doing something that hurts themselves or others (which is very rare), but more importantly it’s just THE RIGHT THING TO DOBasically, kindness toward everyone is good for everyoneLet’s treat bullying and mean behavior similar to the way we treat terrorism: if you see it, report itAnd if you see someone being treated like that, go out of your own way to be kind to them

This is crazy, I know. But fuck—I’m ready to try anything, and I can think of worse ideas.

Curious what you all think.

News & Analysis | NO. 333

window.location.replace("https://mailchi.mp/danielmiessler/uns...

No related posts.

Newsletter Analysis: What My Favorite Newsletters Have in Common

I read a lot of newsletters as part of my content consumption workflow, and since I have my own newsletter as well (started in 2015 before it was cool!), I’m hyper-curious about what works and what doesn’t.

More specifically, I see lots of similar tricks being used across the 20-or-so that I subscribe to. These are things like having a personal intro, a custom subject line, asking people to share, using links within the copy, running a referral program, etc.

Attribute breakdowns

So I decided to take my top 10+ newsletters, break down those tricks as features, and put them in a spreadsheet.

The main two purposes I had were:

Learn what the common tropes/memes that are spreading within top newslettersLook for things that my favorite people are doing with their newsletters that I am not

…and here’s the full list of attributes I looked at:

See the raw data.

Click to enhance

Use of imagesA personal introA table of contentsInline linksNewsletter sizeA newsletter value propA custom title per episodeA referral programA call to shareA link to subscribeHow much humor is usedThe type of ads usedThe tech platformUse of an inspirational quoteMy takeaways

I also just wanted to play with the raw data Google Sheets and Google Data Studio.

Most Top Newsletters Have a Value Proposition: This means a short blurb at the top that reminds readers (especially new subscribers) what value the newsletter is bringing them.

All Three Top-tier Newsletters Include an Intro: This is a decent signal that those who don’t (including me) might want to include one.All Three Top-tier Newsletters (and the majority) Use Custom Email Subjects and/or Headers: This is where each episode has a tagline that tells you what’s in it. It takes more time but the top offerings almost universally have it.Only Super-long Newsletters Tend to Have a Table of Contents: Of the biggest three that I follow, only Trends.vc has a table of contents. And it’s basically a giant index of content so that makes sense.

Very Few Newsletters Have Referral Programs: Only two out of the 12 have them, but I expect this to change shortly. I’m adding a program to mine very soon as well.Most Don’t have a Sign-up Link, Probably Because They Assume Someone Reading is Already Signed Up: Two of the three top don’t have it, but 1440 does. The idea is to catch people who had the newsletter forwarded to them.Most Have a Read-on-Web Option: This still seems important for some people, since the majority still include it.

Three of My Top 4 Have No Images: While there are a number of newsletters that have images now, it seems like the top ones are still mostly text.My Top 4 Have Virtually No Humor: It seems like mass-appeal skews towards being neutral, and humor isn’t neutral.Substack Had the Highest Platform Representation: This was followed by Mailchimp.

To make this prescriptive, if you run a newsletter I recommend you do the following.

Unless you’re over a million subscribers, include a value propositionConsider opening with a brief, personal introUse a custom email subject that describes the episodeConsider getting a referral program to jumpstart growthMy top 4 recommendations

You should obviously consider following every newsletter I mentioned here, but here are my must-haves:

1440—Unbiased, broad-spectrum news. Subscribe3-2-1 Thursdays—A concise, inspirational newsletter by the author of Atomic Habits. SubscribeTLDRSec—An ultra high-quality newsletter focused on application and cloud security. SubscribeUnsupervised Learning—My own newsletter that combines news, analysis, and original ideas in security, tech, and society. Subscribe

Thanks for reading, and I hope this helps you on your journey.

And if I missed any newsletters or attributes you think I should add, please let me know.

News & Analysis | NO. 331

window.location.replace("https://mailchi.mp/danielmiessler/uns...

No related posts.

News & Analysis | NO. 330

This is UL Member Content

Subscribe

Already a member? Login

No related posts.