My top reads from this past week:
- SV founderbros pretend to build what people want
- the rise of Project Maven
- the ups and downs of high-end art theft
- selling your old Slack chats for cash
and more!
Read online or subscribe:
inothernews.complex-machinery.com/archive/024-...
We still see glimmers of what could have been, in every data shop that was built (or influenced) by knowledgeable, experienced professionals who know better and the execs who heed their guidance.
100%. All technology is automation! From machinery to code to ML/AI to genAI.
People in tech tend to skirt around the topic because it's uncomfortable -- you build things that eat away at others' jobs -- but that is the entire game.
There's an all-too-common problem in tech:
The tools have improved so much that it's faster to build something than to think it through.
genAI has now made that problem worse. You can build faster than you can search.
Every line of this piece is quotable.
And it piece is required reading for anyone in tech -- from execs to product to developers.
This article does double duty:
Change a few nouns and it explains how influencers sell _ideas_ to people, not just merch.
(Those direct-to-camera, talk-like-we-have-a-relationship videos are no accident.)
A commerce hellscape: when the marketplace owner is also a participant, and bots go head-to-head over pricing.
Hey NYers, come see me and my dear friend Jesse talk about this moment in AI!
It's next week at the lovely @automattic office in NoHo.
RSVP here: luma.com/jae3jufw
Stellar read about automation and lethal weaponry.
Sounds like the book described here would pair well with Paul Scharre's "Army of None," which presents a much more reasonable view of autonomous weapons.
www.newyorker.com/books/under-...
As a longtime tech professional, I will say the quiet part out loud:
Automation eats work.
Machinery, code, ML, and now genAI all shrink the org chart.
Say what you will about those first three, but a company could at least argue that they work.
genAI (save for a few use cases) not so much.
Adding to the list -- and something I rarely hear in discussions of technology eating jobs -- is that genAI simply does not work as well as advertised.
The double frustration of not only being fired, but being replaced by a vastly subpar substitute, is a lot for someone to handle.
(For anyone who says that editing can cure a lot of ills: this is true! But ... an editor can only work with the raw material you give them.)
Yes! A good show requires host + guest to 1/ have a rapport 2/ be engaging 3/ stay on-topic while also 4/ letting a conversation flow organically.
As a former (and future!) host: this requires a ton of prep work that few are willing to undertake.
Which explains the current podcast scene.
For those who wonder why I hold such strong opinions on this topic:
I'm fast approaching 30yr building automated systems (software, ML/AI, genAI)
with a focus on risk (having worked in finance)
and a longtime student of war history (mostly: asymmetric, insurgencies)
It's all connected.
I'd go as far as to say that "Army of None" is required reading for people building _any_ kind of automated system.
The lessons from war apply just as well to civilian/business pursuits:
www.paulscharre.com/army-of-none
Excerpt from the linked article: >> Manson continues, "Speaking to me years later, Cukor made no bones about it either." What was the point of all this speed if you needed to wait for cumbersome human supervision? If the machines could identify the targets, couldn't they also pull the trigger to rain death from all angles?
A key point in Scharre's "Army of None" is that a machine should pick a target or act on a target (as chosen by a human); it should never do both. The cost of an error is simply too high.
Compare that to Project Maven's advocates:
www.newyorker.com/books/under-...
Stellar read about automation and lethal weaponry.
Sounds like the book described here would pair well with Paul Scharre's "Army of None," which presents a much more reasonable view of autonomous weapons.
www.newyorker.com/books/under-...
The anti-datacenter backlash has been genAI's first real source of friction.
Despite surviving incidents in our personal lives and businesses, genAI's undoing may stem from its push into the physical realm.
My old warning of
"write every email as though it'll be replayed in a courtroom"
needs an update to
"write every Slack post like it'll wind up in a training dataset"
My old warning of
"write every email as though it'll be replayed in a courtroom"
needs an update to
"write every Slack post like it'll wind up in a training dataset"
That not-good-middle-manager bot sounds eerily similar to some failed dev-to-middle-manager transitions (which I may or may not have witnessed ...)
New issues land two or three times a month. You can read Complex Machinery online, or subscribe to get it in your inbox.
newsletter.complex-machinery.com
What you _won't_ see in Complex Machinery:
⛔ code examples or reviews of the latest hot tech tools
⛔ raving pro-genAI boosterism
⛔ performative, rage-click anti-AI rants
newsletter.complex-machinery.com/archive/
Past issues have covered LLM bots gone wild, AI's impact on jobs, the CrowdStrike incident, how the world may look after a genAI crash, and what self-service kiosks can tell us about AI.
Two years of archives are available for your perusal. (No paywall!)
newsletter.complex-machinery.com
What you'll read in Complex Machinery:
✅ plain-language observations + analysis of the AI space, provided by a longtime industry expert who has Been There & Done That
✅ parallels to financial concepts and asset bubbles
✅ the occasional cat pic
Periodic reminder: I publish a newsletter called Complex Machinery.
2-3 times a month I share my thoughts on AI and risk, plus related topics like business models and complex systems.
You can read online, or subscribe to get it in your inbox:
newsletter.complex-machinery.com
The "Miyazaki stressed" meme, from Imgflip. It depicts Hayao Miyazaki with eyes closed and hands on his head, looking frustrated.
tfw you're in the middle of writing (yet another) essay comparing genAI to historical asset bubbles, and the Allbirds news drops.
Anyway, next week's newsletter should be a real treat. There's still time to subscribe:
newsletter.complex-machinery.com
My top reads from this past week:
- Anthropic's Mythos makes news
- the Stanford AI Index report
- books: paperback or hardcover?
inothernews.complex-machinery.com/archive/023-...
Seconded -- this series is well worth a watch.