> Vibe-coding won't be a net creativity gain to a researcher affected by vibe-immigration-policy, vibe-grant-availability, and vibe-firings, for all of which the unpredictability is a punitive design goal.
Quite literally this is what I’m trying to get at with my resistance to LLM adoption in the current environment. We’re not using it to do hard work, we’re throwing it everywhere in an intentional decision to dumb down more people and funnel resources and control into fewer hands.
Current AI isn’t democratizing anything, it’s just a shinier marketing ploy to get people to abandon skilled professions and leave the bulk of the populace only suitable for McJobs. The benefits of its use are seen by vanishingly few, while its harms felt by distressingly many.
At present, it is a tool designed to improve existing neoliberal policies and wealth pumps by reducing the demand for skilled labor without properly compensating those affected by its use, nor allowing an exit from their walled gardens (because that is literally what all these XaaS AI firms are - walled gardens of pattern matchers masquerading as intelligence).
Yes, "The Little Kingdom" is a great book (the very slightly updated edition is called "Return to the Little Kingdom" but really just adds a short epilogue), and in my opinion (having read most of them), the best book on early Apple. It gives you more insight into some of the other characters at the early company (Mike Markkula and Mike Scott for example) than some of the other books. Interestingly, the author, Mike Moritz (a lot of Mikes), went on to be a highly successful venture capitalist. So he must've had some keen insights about the tech industry.
I was a guest on a podcast called "CFO Bookshelf" to discuss that book if you want to hear a discussion of it before making the commitment to dive in:
If you're interested in DEC history and haven't read "The Soul of a New Machine" by Tracy Kidder, you owe it to yourself to get a copy. A Pulitzer Prize winner and IMO one of the best tech books ever written. It is a captivating book that chronicles the creation of Data General's Eclipse MV/8000 computer from the engineers' point of view.
> “Communication within the YouTube Apparatus has no meaning.” The rapid feedback loop between creators and audiences (as constructed by platform metrics) means that the system more and more responds to itself. Rather than trying to go somewhere (as is the case with political ideology), the creator seeks simply intensification, to draw more and more of the world into his whirlpool of content.
This idea – that meaning is replaced by intensification – helps me understand a lot about the world today.
Google didn't change it, it embodied it. The problem isn't AI, it's the pervasive culture of PR and advertising which appeared in the 50s and eventually consumed its host.
Western industrial culture was based on substance - getting real shit done. There was always a lot of scammery around it, but the bedrock goal was to make physical things happen - build things, invent things, deliver things, innovate.
PR and ad culture was there to support that. The goal was to change values and behaviours to get people to Buy More Stuff. OK.
Then around the time the Internet arrived, industry was off-shored, and the culture started to become one of appearance and performance, not of substance and action.
SEO, adtech, social media, web framework soup, management fads - they're all about impression management and popularity games, not about underlying fundamentals.
This is very obvious on social media in the arts. The qualification for a creative career used to be substantial talent and ability. Now there are thousands of people making careers out of performing the lifestyle of being a creative person. Their ability to do the basics - draw, write, compose - is very limited. Worse, they lack the ability to imagine anything fresh or original - which is where the real substance is in art.
Worse than that, they don't know what they don't know, because they've been trained to be superficial in a superficial culture.
It's just as bad in engineering, where it has become more important to create the illusion of work being done, than to do the work. (Looking at you, Boeing. And also Agile...)
You literally make more money doing this. A lot more.
So AI isn't really a tool for creating substance. It's a tool for automating impression management. You can create the impression of getting a lot of work done. Or the impression of a well-written cover letter. Or of a genre novel, techno track, whatever.
AI might one day be a tool for creating substance. But at the moment it's reflecting and enabling a Potemkin busy-culture of recycled facades and appearances that has almost nothing real behind it.
Unfortunately it's quite good at that.
But the problem is the culture, not the technology. And it's been a problem for a long time.
"Powder Game"[1] is the Flash game off the top of my head that fits your description. If you're looking for a more fleshed out version with many times more elements/interactions, check out "The Powder Toy"[2]!
Most intro classes just require familiarity with basic calculus (differentiation, chain rule), linear algebra and basic probability all of which you can just lookup directly on https://www.expii.com for a short tutorial. Toolkits are usually in Python or Lua, plus the numerous textbooks 'Deep learning with python' that are around and specific DL books such as http://www.deeplearningbook.org/.
Afterwards look around for Adversarial Learning, like detecting perturbations that force mis-classification and other attacks described in papers by Carlini and Wagner. Currently there isn't a perfect defense developed for all of these attacks, except robust optimization that provably defend some of them. Attacks are an interesting area in DL you can get into since we don't have access to large resources and can only do DL on a small scale (in my case anyway).
I'm the author of Robust Python, and the first quarter of my book is all about type hinting, so suffice to say I've thought about it a lot. What follows is strictly my opinion on things.
I absolutely agree that there needs to be a balance between verbosity and clarity. Obvious things such as some variable names probably don't need to be annotated, but I think we disagree with what an obvious name is. When I see a variable named timestamp, is that an int, or a datetime? When I see days_since_creation, is that an int or a timedelta? When I see amount_paid, is that an int, float, decimal.Decimal or some custom type? The only way I can answer these questions is look at how it's used. For variables that may be clear, but for things like parameters in a function it's less clear. I have to look at calling code to see what people can pass in, and I have no guarantee that all the calling code is even visible to me. In these cases, I will happily trade off the verbosity of a type annotation so that I can have a better chance of correctness as more people work on the code over years. Now, if the project is small, maintained by a single developer or something like that, then I would say type hints can just be added noise. (but if at anytime you need to change the name of a variable to be more clear, such as "amount_paid" -> "amount_paid_decimal", I would encourage a type annotation instead).
I also think it's a mistake to treat type checking as a hack to feel more Java-like. I feel like the more appropriate counterpart is TypeScript (Guido has talked about it before: https://developers.slashdot.org/story/21/05/22/0348235/what-...). I believe type annotations (along with type checkers) was introduced to Python to solve an absolutely real problem with dealing with the robustness of codebase when you have multiple people working on it (across multiple years, even decades even).
All in all, yes, I do not think you should be type annotating everything, but I also am wary about type annotating too little. All of this advice is context-dependent; there is no one right answer depending on project size/value. But if you want to increase communication and lessen the amount of mistakes someone can make as they modify your code (especially if they never get the chance to talk to you), I whole-heartedly endorse type checking and type annotations. After all, your code is going to be read much more often than written, and I'd rather optimize for the unfortunate souls who have to read my code later on. I'll happily pay the penalty now for writing out a few more characters.
The spirit of the old web is still alive and thriving in places that are now no longer mainstream. It takes some effort to find them but great sites are still out there.
A classic. The movie versions are available on YouTube with English subtitles but it's not the same. A part of the charm of Švejk played by Rudolf Hrušínský is the way he talks and that gets lost if you don't speak Czech.
It's kind of weird to me that it starts from $139, a price range where you can buy a 65% Keychron mech keyboard[1], _and_ stuff like the Pinephone Keyboard[2]. The latter is not exactly comparable, since its more like a mini-keyboard stuck into a mini-sized case, but I still feel like $139 is a lot for something that you can't use beyond a single generation of a device. Maybe if it was modular and you could take out the keyboard from the case for easy upgradibility or something, the price could be justified?
For the creator of this product (and to any other creators launching soon): don't listen to the keyboard jockeys saying it will fail. We (commenters here) are essentially NPCs in your world. We're not your narrow target market, just some gawkers addicted to typing "news.yc..." into our browsers, maybe with nothing better to do. And although we sound confident, we really don't know if it'll fail, or take off, or pivot to a successful product. Good luck--and major props on the launch, the landing page looks great!
Best value is always an ebay Thinkpad. But if you insist on buying new, the Lenovo Ideapad 3 Chromebook is $100 and often discounted down to $80. It's a 4GB/64GB eMMC/Gemini Lake n4020 with a TN screen. But it does have a good keyboard relative to its price range, and feels pretty solid. Battery life is about 8 to 10 hours.
The codename for this Chromebook is "Lick" and MrChromebox has alternative firmware which will allow Linux to be installed. I haven't done it on mine yet, but the instructions aren't difficult. I've read that most things on Linux will work automatically but there's a problem with sound out-of-the-box. There's a script on Github that supposedly fixes that.
For anyone curious as to why crime in Russia seems to be so ingrained into the culture, this lecture from a finnish intelligence officer is quite informative.
https://m.youtube.com/watch?v=kF9KretXqJw
I just bought a ThinkPad X1 Tablet Gen 3. Everything works great, the on screen keyboard is actually functioning properly and appears on demand every time. Screen orientation is handled properly and promptly. The only let-down so far is battery life, but I need to make some adjustments to help there.
I'm running the latest version of Fedora workstation, under wayland too.
This thing is the only truly repairable tablet I've encountered so far, 8 screws and the display pops off, no glue, no can opener required like a Surface Pro. 9 screws to get the heatsink off (they're all captive) and you're at the full size NVME SSD.
Also free and relevant - a link to WHO's guide for "unhooking from difficult thoughts and feelings". There's audio you can download. It's not even an app.
I had a colleague who’s degree was in comparative literature - we’d walk out of meetings and the rest of us would be talking about the engineering side of what we’d just heard while he’d go through and enumerate the different things each person had been talking about while using the same words as everyone else. The amount of latent conflicts that dude caught before the rest of us got torched made me really appreciate the value of an art degree.
If you're interested in other terrifying compounds be sure to read 'Ignition'.[0]
My footnote is for a scan of the original printing but I believe the book has have recently been reprinted in multiple formats. Regardless of the format you choose, it's a hell of a read.
K8 seems pretty small after he made the fucking Zen architecture (and Tesla AI). Lex has good interviews with Jim[1], he's not just very smart, he is wise. He started Tenstorrent some time ago with Ljubisa Bajic so I'm assuming this fab is related. Here is youtube channel for Tenstorrent[2], and last time I checked it was getting under 100 views per video... Ian also interviewed him[3] some time ago.
Long story short if I had money that would be relevant in this context I would invest really hard into Tenstorrent.
Investing time to listen to the guy seems to be not worse.
Quite literally this is what I’m trying to get at with my resistance to LLM adoption in the current environment. We’re not using it to do hard work, we’re throwing it everywhere in an intentional decision to dumb down more people and funnel resources and control into fewer hands.
Current AI isn’t democratizing anything, it’s just a shinier marketing ploy to get people to abandon skilled professions and leave the bulk of the populace only suitable for McJobs. The benefits of its use are seen by vanishingly few, while its harms felt by distressingly many.
At present, it is a tool designed to improve existing neoliberal policies and wealth pumps by reducing the demand for skilled labor without properly compensating those affected by its use, nor allowing an exit from their walled gardens (because that is literally what all these XaaS AI firms are - walled gardens of pattern matchers masquerading as intelligence).