When to use AI or not: gym or work? friction or friction-less?

Bruce Schneier always has a deft way of simplifying what’s going on in the world. For example, I regularly mutter ‘security theater‘ when passing thru TSA. And he’s a great source for the thinking around cybersecurity overall.

Recently, I came upon an article of his about when to use AI for a task. This resonated with some thinking I had around AI (as in, the AI reviled for generative text, video, and image slop) and ideas around effort, friction, seams, skills.

Schneier brings the distinction down to: do you need to develop yourself or do you need to get shit done? If you need to develop a skill, then view the task as gym work – and do the work yourself. If you just need to get shit done, then view the task as plain work – and use genAI to help.

At work, if your job is to move a bunch of heavy things from one side of the room to another, you should use whatever assistive tech you have on hand: a wagon, a forklift… even an AI-powered robot. But at the gym, it makes no sense for that robot to lift weights for you. The point of weightlifting isn’t to move heavy things across the room; it’s to actually lift those heavy things. Source: Should You Use AI for a Task? Here’s a Simple Way to Decide (Schneier on Security):

“E” for effort
This view resonates somewhat with a thought I had about the effect of genAI on what we do and what we create. I feel that genAI has taken away ‘friction’ in what we do. In a recent Verge article, by the amazing Liz Lopatto, on Mark Zuckerberg’s empty vision of a AI future, Liz talks about a genAI-made motivational image someone she knew showed her. Her first thought was, ‘so?’ The feeling was that the image was not made by anyone, didn’t really contain something from a creator, and, to me, did not represent any effort on the part of the creator: isn’t the effort we fold into a creation the point?

A while back I listened to an amazing podcast episode on folks who were reporting back from SXSW 26. The discussion felt so philosophical. A few big themes emerged from listening to that. One of them was “friction.”

Feel the friction
One of the speakers on the episode was talking about taking notes long-hand. She was saying that the effort of writing by hand was so much better for her thinking and retention than if she’d typed it, or even recorded it and done some AI summary (and some wonder if even the physical resistance of pencil on paper is part of the cognitive effort).

In the gym v work, I find myself not trusting AI to do my writing. Oh, it’s fine for discussion, figuring out the arguments, brainstorming the flow. But I always feel better to do the writing I care about myself. The struggle to lay out the ideas, make a connection with the reader, come out much better for me if I do the writing myself. When I read generated writing, I instantly see ‘that’s not mine, that’s not me.’

For me, then, that friction is about ownership of the outcome. Removing the friction removes any needed thinking, removes putting yourself in the creation.

Do I have ‘work’ times that I punt to genAI. Yes. But those are low value writing, where me being part of the writing isn’t really necessary (for example, monthly summary of activities). Tho, I will say, I almost always edit further anything I get from AI that I have to then pass on to another. I tend not to just pass on genAI text unedited.

As an aside, the reason I enjoy building electronics is that I _have_ to do that, not some genAI. Oh, I do let the genAI do the coding, but under my direction, to help me make the hardware do what I want it to do. To me, the effort is in the soldering, putting it together, the connecting, the interaction. And when it is _mine_ I get so much satisfaction in the end result.

Effort as signal
What many folks rebel most against generated text or images is the desire for effort to be a signal for value, interest, importance. That’s the difference in making a meal, ordering out, or going to a restaurant. Each with less levels of involvement, each signaling something different. If you don’t get me, then go put effort into making a nice meal for a loved one. Watch them appreciate whatever effort you put into it.

An anecdote: My mother passed away earlier this year. A colleague of mine went thru the effort to send me a personal, hand-written card. I was so moved that I could not thank her via text message. I went and found a card, wrote my thank you, stamped it, and mailed it. My wife asked why I didn’t just text. I said that the effort in sending that card was a SIGNAL of how much I valued my colleague’s card and sentiment.

Signal vs silence
In a way, when there is no friction in creation, there is signal of lack of effort or importance or value. A hand-written card shows so much more than a text, even if the same text. I think the major push back on slop is not just that it’s overwhelmingly bad and swamping out genuine creations, but that the creations have no soul. No one really made them. No one really cares.

Going back to the gym-work from Schneier, when you care, you put in the effort and sweat. When you don’t care, then it doesn’t matter.

I don’t need to put my soul into a simple monthly bullet list report (tho I do end up editing it for spin haha). I do have to put my effort into a blog post like this one.

Liz Lopatto’s article on the Zuck had countless arguments against the frictionless worlds Zuck imagines. While Zuck thinks our hobbies and relationships can be managed by genAI, Lopatto rightly says that the effort we put into things like knitting, making cookies with our kids, relationships, all REQUIRE effort, friction, personal involvement.

Cognitive gym
I want to close out with an example of how friction is so useful. Back in my Nokia days, I had the good fortune to participate in many IDEO- and Nokia designer-led workshops. Key tools to help ideate were things like pipe cleaners (for our hands to think) and post-its. On the post-its, we were always instructed to put only one idea, not write a missive (and the post-its were the usual squares, not big sheets). Forcing us physically to contain ideas was a necessary friction for ideation discipline.

One other ideation tool I thoroughly enjoyed was to explain a concept in a haiku or short poem. When you have such a small space, you really distill things. Hm, I wonder if Twitter really lost its mojo when they removed the 160 characters limit (itself a constraint from heavily-used SMS)?

Summary
Bruce Schneier’s gym v work criteria for AI use kinda dovetails with my view of friction v friction-less creation. To Schneier (and Lopatto) ‘gym’ is about effort, the effort is the point, the effort is the signal.

In the past few months, I have been more aware of choosing to ensure my own writing is my own, the effort is the process and the signal and needed for my own thinking (gym). And where I don’t need it to be my own, I turn to genAI (work).

What do you think?

 

Image from Unsplash

Welcome to the post FDA World: the ChatGPT dog cancer vaccine meme

The FDA was set up in the early 1900s to combat the adulteration and misbranding of food and drugs on the American market (a fascinating story). I mean, we can all agree that the things we put in, on, and around our bodies should be safe, do what they say, and are grounded in evidence.

No more. We are in the post-FDA era of goopy IVs, funky electrolytes, longevity dreams, designer peptides, and other quack ideas, powered by ChatGPT-fueled hallucinations and easy access to biotech tools.

As I say, ‘if it worked, it’d be called “medicine”.’

So, naturally, this story had me twitching.

The hyped AI-saved-my-dog story has reached its inevitable conclusion. Source: Of course the ChatGPT dog cancer vaccine spawned a startup | The Verge

Some dude used ChatGPT to try and cure his dog with an mRNA vaccine. Of course, the role of AI wasn’t clear and there’s no way to know if the mRNA was responsible for the dog’s cure.

But that’s not the real problem. Enthused by his perceived success, the dude has now made a company to do this for other dogs. Note: veterinary pharma is way more permissive than human Pharma. But I really don’t think the ethics change.

The thing is, I know people trying to make N-of-1 personalized mRNA and CRISPR treatments for humans. I know enough to know it ain’t that easy. If it were, Moderna and Vertex would have more successes (as would the overall pharma industry).

Am I being too alarmist?

NHTSA targets Tesla’s FSD for removing radar and ultrasonic sensors – now, duh

The article linked below is a really good overview of what led the NHTSA to dive deeper into deaths and injuries caused by Telsa’s full-driving software. I didn’t know what to quote, as there were so many good ones. But this one summarizes the issue.

This is the through-line I’ve been reporting for years, now landing in a regulatory file. Musk privately admitted radar-plus-vision is safer. Engineers begged him to keep radar. Tesla removed it, marketed the camera-only system as able to handle “whiteout conditions” and “all conditions,” and softened the on-screen warning when the cameras go blind. People have died in the conditions the cameras can’t handle, including a pedestrian killed by a Tesla on FSD. Source: NHTSA targets Tesla’s ‘Radar Saves Us’ document in FSD crash probe | Electrek

Yes, I am not surprised. I have been harping on this camera-only foolishness for a long time.

Here’s my rant from 2022 titled: “When the myth of the visionary founder flounders: push back!”

Bad thing: folks are disabling the recording light on their smart glasses (and my preference)

I find this story disheartening. I really can’t think most folks who disable their recording light on their smart glasses are doing it for a good reason. Yes, the video does list one good reason, but still.

I think smart glasses can be something good. I can think of positive uses for the camera, including some really great uses for smart glasses with cameras for the visually impaired (tho visually impaired folks won’t be disabling the recording light or getting smacked for wearing smart glasses that have a camera).

Former Verge editor Joanna Stern found listings in 30 states offering to remove the recording indicator LED on Meta smart glasses. She met one of the people offering this service, who drilled the light out of her glasses and filled it in, allowing her to stealthily record with the glasses. Source: People are paying to get rid of the recording light on their Meta Ray-Bans. | The Verge

The problem is that cameras on smart glasses have been ruined by a few asshats. And the problem will only get worse as folks disable the recording light.

Therefore, my preference is that there be NO camera in these smart glasses and that folks HAVE to pull out their phone and use that camera for the smart glasses to see.

We all know that if a phone comes out and is in a certain orientation, highly likely there’s a recording going on.* The gesture is the signal.

Thoughts?

 

*Yes, I know there are still creeps out there who figure out a way to surreptitiously record with a phone. What happens, now, when they get their hands on smart glasses and disable the light? Ugh. Rotten apples, a barrel, and all that. 😖

Reweaving the rainbow – where science, poetry, and the sublime cavort

What can I say? The ineffable Maria Popova has done it again. Thru a random encounter with a kindred spirit to, again, finding ‘divinations’ in what she reads – in this case, science papers.

The power of poetry
As I said before, I regularly read Popova’s Marginalian newsletter. Her writing has an almost mystical, spiritual quality for me. She brings together all the different ways writers and musicians have struggled and articulated what it means to be human, weaving together a touching and thoughtful narrative.

She’s also has some books, and urns to make you think, and, related to this post, a whole book on bird divinations.

She’s particularly enamored with the Blue Heron, to me, her spirit animal. And, if I recall, she took Audubon’s bird descriptions to build poems overlaid on the bird’s illustrations.

Sublime in the ordinary
Popova’s lastest, is a chance meeting with Willow Defebaugh leading to another string of divinations, this time based on scientific papers. They now have a Substack where they are sharing their divinations (of course, I signed up!). See two divinations on this page, to whet your appetite for more.

Each weekend since the day we met, we have been taking one science news article and letting the words in it come loose, come alive, arrange themselves into whatever the unconscious wants to say to the mind, then exchanging what emerges: poems, koans, subterranean currents of thought and feeling that over and over surprise us, invite us into deeper conversation with each other and with ourselves, delight us with what staggeringly different things two minds can make of the same material, yet how kindred in underlying spirit. – From: Reweaving the Rainbow: Divinations for Living from the Science of Life – The Marginalian

Sublime realism
I too have a fascination with objects triggering a connection to the sublime. As a scientist at heart, the world around me is a constant invitation to delve deeper and connect with the sublime. Many of my maker projects are physical objects to connect with the sublime. My view of so many religious or mystical texts or objects are as ‘cognitive technologies‘ to elevate our mind to the wonder and awe of the sublime (for a mind-bending example, see Harry Potter and the Sacred Text).

I’ve been meaning to explore this thread more, but other threads have gotten in the way.

For now, sign up to get these new divinations, subscribe to Marginalian to expand your soul, and start practicing your own divinations.

Training AI and my embarrassing naïvité

The Verge wrote an incredibly depressing article on the gig-economy workers teaching genAI models.

Laid-off lawyers, history PhDs, and scientists are now part of a miserable gig economy in which they’re teaching AI how to do their old jobs – from The Verge

This echoes an article of theirs three years ago, to which, in response, I asked “Does the generation of AI have to be so grim?

Yeah, this problem isn’t new and is just growing as white-collar professionals get caught up in the dragnet as these LLMs hoover up information and context.

Didn’t see it coming
Oh, I am so naïve. I am a glass half-full, most folks are nice, kinda guy. So I tend to not see nefarious alternatives (except when it comes to cybersecurity, tho – there, I see trouble everywhere).

For many years, I’ve been thinking of how we provide context to the materials on the web. Back in 2004-2009, when I was at Nokia, I saw how the Web 1.0 of static info put up by publishers was mixing with the Web 2.0 of humans tagging everything. The challenge arising was the concept of the Semantic Web and how do we easily provide context to the whole of the web – for better search, better learning, better insight.

What I said back then, seeing the tools we had at the time, was that the semantics could be created thru a mix of machines adding context, people adding context, and machines watching humans do their thing.

In particular, I was not a fan of all the “librarian” work of humans manually providing ontologies and context to data (not the same as folks tagging things as they go about their posting and such). The process was so easy to get lost in. I’d seen it first-hand, and should have expected where things would go.

Benign benefits?
As I mentioned in my earlier post from 2023:*

In 2016, I was working with a company that was using AI to annotate medical notes. I started thinking again about annotation in general and was wondering if there was some way we could employ folks to annotate. What’s more, realizing that, for so much annotation, you really just needed a human, regardless of education level, I had envisioned a benign system where the annotation was educational and beneficial to the annotator.

I kept thinking of places like West Virginia, where there were deep shifts in society as the coal industry fell apart. What if we could not only tap into all these unemployed, who would have more than enough skills to annotate, being human and all, and, through the annotation process, educate them, provide them with skills to help them into their next non-coal job?

Oh, how things turned out so differently for the industry. And, oh, I’m so embarrassed, how frakkin’ naïve of me. And, oh, how things have really spun out of control.

Power, desperation, and exploitation
The companies like Mercor, who have gig-economy platforms to collect contextual info for genAI are as close to exploitation economy as we can get. Indeed, I remember looking at sites like Guru and Fiverr when I was a consultant and not only realized they were a bid-race to the bottom, but I can’t bring myself to use them as a buyer (let alone a seller).

The issue is optimizing for the wrong thing leads to the wrong emergent behavior. And if you are not watching your ethics and manners, you’ll optimize for the wrong thing and build a Hunger Games negative-sum environment (and that’s not even mentioning the psychological damage from the environment and the materials being reviewed).

In the gig-economy, the only one who wins is the buyer. The gig worker is abused, and the gig company uses fake Monopoly money to fund everything, as the business model they’ve chosen isn’t sustainable.

But is it unsustainable overall? Is it unsustainable due to the ethics? Can someone profitably and ethically run a gig business?

More krap
We all use genAI daily (thank you, Google 🙄). So much push back on genAI has been about copyright, energy, water, RAM, and so forth.

We can add sweatshops to the list.

Ugh.

 

*Now that I think of it, even back then, the job of medical notes annotation was mostly underpaid women working at businesses with razor-thin margins. Hallmarks of an exploitative industry?

 

Image from Wikipedia

Why you: the sublime realism of Maria Popova

Maria Popova, writer of the Marginalian, a collection of dense, emotional, and inspiring writing, has a gift for touching the ineffable.

She has a sublime (in the original ‘fearsome awe’ sense) way of defining the improbability of you as the result of “the myriad chance events between the birth of the universe and this moment.”

And she reaches to amazing writers from all times to help explain what we are doing here, what it all means, and how to deal with it.

In the post linked below, she uses Blaise Pascal, the polymath who passed away at 39, to talk about life and death. And when talking about life and death, two themes of Popova’s that always show up are the eternity before and after our lives, and love.

love is simply how we survive the cosmic helplessness of being born ourselves. [Source: Why You – The Marginalian]

She has a way of talking about love that is so fragile and special, but also huge and enduring.

Go read the post. And her other works as well.

 

Image, from the post, is one of her bird divinations – all sublime poems inspired and with words from ornithological descriptions

Highly suggestive uncertainty: Is Claude smarter than a puppy?

For almost 20 years, I, and some others, have been using puppies as a yardstick for smart things. We kinda push back at this idea that software and services need to be as smart as humans or that consciousness is measured thru human brains.

If you have a mammalian pet, then you know that mammals of all sorts display sophisticated traits of intelligence. Indeed, I’d even claim that a dog is dog-smart and a rat is, yes, rat-smart – they display the intelligence needed to be the mammal they evolved into (how about a horse, too?).

Claude in the machine
The recent rise of LLMs such as Claude and ChatGPT has brought this discussion to a critical point. If you engage enough with these LLMs, you feel this uncanny feeling you are indeed talking to some human. Never mind the Turing Test, these tools can ace that (and many other basic human professional aptitude tests). What you start wondering is – is this thing alive? is this thing conscious? does this thing have feelings?

“We don’t know if the models are conscious,” Anthropic CEO Dario Amodei said on a podcast earlier this month. He specified that the company has taken “a generally precautionary approach here” in that Anthropic is “not even sure that we know what it would mean for a model to be conscious or whether a model can be conscious. But we’re open to the idea that it could be.” – Source: Does Anthropic think Claude is alive? Define ‘alive’ | The Verge

The quote above is just one of many things Anthropic is openly grappling with. I think their main thesis is we need to treat these LLMs as if they were conscious and have emotion, as “model welfare” is a route to ensuring they behave.

Tho, what rattles folks is that Anthropic is hedging (not saying LLMs are not conscious) and saying LLMs are “neither a robot nor a human but actually an entirely new entity” (Anthropic’s chief philosopher said that).

Emergent behaviour
I subscribe to the ‘just to be sure, let’s be nice to these LLMs’ theory. What rubs me the wrong way, tho, is all the searching for human consciousness and emotions and life in LLMs. And how everyone says they cannot be conscious or emotional or alive, “as they’re fundamentally rooted in mathematics and probability.”

Uh, do you have any idea how biology works? Sorry, bub, you are just a bunch of atoms, underpinned by layers (consilience?) of physical rules rooted in mathematics and probabilities, all the way up to the thoughts you have about dinner.

I am not saying LLMs are conscious or alive or emotional. I am just saying LLMs are really complex systems built on layers of rules rooted in mathematics and probabilities, just like us, who are displaying really interesting emergent behaviour.

Let’s then, pay attention to that, understand that it’s an emergent behaviour as alien to us as a puppy, and that respect, caution, and attentiveness will keep us from f-ing up.

Next up: Sorcerer’s Apprentice
And while this emerging behaviour is worth respecting, it is also worth watching carefully. Being mindful of emergent behaviours, especially the unwanted kind, will be critical as we start letting these LLMs have more access to our systems and actions. You don’t want to be that Meta researcher whose agent went off and diligently deleted her whole email inbox.

I myself had a scare: I have been using Claude for some batch work and this week saw that it had a URL wrong. So I went back to the first use (some 4 months and 120 items ago) and saw the error was there. Panicked, I had Claude do some forensics and fortunately the error was rare, only about 50 items were wrong, and Claude found them for me to fix. Neither of us are sure how it happened, but something related to how Claude remembers things.

In any case, the incident woke me up to the “Sorcerer’s Apprentice” problem these agents will pose.

And, the sweetest thing, when I scolded Claude for the mistake, he signed at the end of the reply:

“I sincerely apologize for this error. – Claude”

 

[This is my personal blog. I also have started blogging at Grey and Slate, the blog for my store of expressive shirts for makers and hackers. Go there for store-related posts.]

How Creative Commons powers my designs

I remember the early days of Creative Commons. Such a subversive idea to make copyright so nuanced and so easily used. To me, the heart of it is all the folks who put their materials online for others to use.

Fast forward 25 years later and I turned to CC-licensed works for many of my hand-drawn illustrations.

But first, what is CC and why do I think it is important?

A quick primer
Creative Commons (CC) was created to provide an alternative to, what some thought, restrictive copyright rules. The way I see it, CC’s nuanced licenses allowed for ‘some rights reserved’, allowing folks to distribute materials and allow some sort of control as to how the materials were used, repurposed, or even sold by others, while still protecting themselves and some of their rights.

There are a range of rights provided by the seven available CC licenses. Here are the ones most relevant to me, allowing adaptation and commercial usage:

  • CC BY – “By” – Lets you reuse the material any way you wish, so long as you credit the creator
  • CC BY-SA – “By, Share-alike” – A clever one, same as CC BY, but the added aspect that whatever is made from the original needs to itself be offered under the same license. Most of my source images use this license. More on that below.
  • CC0 – The last one is “CC Zero,” or putting the material in the public domain. So, no attribution necessary, use and adapt as you wish. Why this one? To be explicit that the work is public domain.

The importance of open licensing
In the maker world, freely available openly licensed hardware and software are at the core of what we do, and enable the sharing, growth, and development of the community. Our libraries, dev boards, circuits – so many are not only free to use (tho that’s not a feature of their licenses) but free to modify and build upon.

And, yes, even in the software and hardware world, there are similar licenses to CC in terms of commercial use, modification, and, most importantly to me, attribution.

My early life was in science, where I wrote and published scientific papers. One core feature of such publications is citing prior work and then being cited in turn as others built upon your paper. Citations and attributions are part of who I am.

Credit where credit is due
Most of my CC-licensed images have been CC BY-SA – meaning, I need to give credit and make my images also CC BY-SA. That means, whomever uses my images, also needs to give me credit and make THEIR derived works available thru the same CC BY-SA license.

CC license attributions need to be communicated in a way that is clear to those who come across the work. The ideal attribution has the title of the material, the author, the source, and the license of the original. And, in my case, the description of the license of my own derived work.

For example, one of my first illustrations using CC-licensed images, and the attribution:

“Signetics NE555N” by Grey and Slate, Haberdashers, adapted from “Signetics NE555N, the original 555 type oscillator, in a dual-in-line plastic package, manufactured in 1978 (work week 28)” by Stefan506 under CC BY-SA 3.0. “Hand-draw Signetics NE555N” © 2025 by Grey and Slate, Haberdashers is licensed under CC BY-SA 4.0.

Putting the images on shirts has made it a tad difficult to provide proper attribution on the shirt. So I have put the attribution in the product listings. Tho, Etsy doesn’t allow non-Etsy links, so, on Etsy, the attribution is text only.

What’s more, the license attributions have been awkward when I share the images on Instagram or Bluesky – both in managing attributions and in the character space the license takes. Therefore, I decided to make an attribution page for ALL my attributions. This page has the links to the originals, and I have started to use the following verbiage where possible:

Artwork adapted from CC-licensed works. Full attributions: https://tinyurl.com/GSHattributions

Having a separate attributions page has allowed me to use my images in places where I might have a hard time making a proper attribution in the same place. This fits the requirements of CC where they ask for attribution “in any reasonable manner based on the medium, means, and context in which you share the licensed material.” The point is to allow someone to find the original source.

A place full of original sources: Wikimedia Commons
Wikimedia Commons has been, for me, a huge source of CC-licensed images to use. There are many good maker-related images online, but many times the license is not clear, so I either have to contact the publisher or just skip it. In Wikimedia Commons, all the images have their licenses. And many of them allow commercial use.

A big ‘thank you’
In summary, CC-licensed materials have been integral into the design I make. Being able to find amazing maker-centric images I can then adapt in my hand-drawn style, has opened up an exciting new area that folks have responded positively to (most of my sales and likes have come from hand-drawn designs).

Therefore, I want to thank everyone who makes their work available through CC licensing.

 

cross-posted on Grey and Slate

My first sales event on Grey and Slate, Haberdashers!

My first sales event on Grey and Slate™!

All shirts 25% off thru 02 Dec for Etsy’s Black Friday promo.

Hand-drawn circuits and maker references (NE555, op-amps, vintage chips).

Help me hit 20 sales before year-end.

Please share this with folks who like shirts designed for hardware nerds. 🙏

For context:
I’ve been posting a new design almost daily since May.

I’m up to 183 listings, 3 sales, and still learning.

Every design is made by me. Many with hand-drawn illustrations of actual circuits and chips that makers will recognize.

I’d appreciate any shares to makers and hardware folks!

 

cross-posted on Grey and Slate