jarfil

@[email protected]

Programmer and sysadmin (DevOps?), wannabe polymath in tech, science and the mind. Neurodivergent, disabled, burned out, and close to throwing in the towel, but still liking ponies 🦄 and sometimes willing to discuss stuff.

This profile is from a federated server and may be incomplete. Browse more on the original instance.

jarfil, (edited )

htmx extends and generalizes the core idea of HTML as a hypertext, opening up many more possibilities directly within the language:

Ok…

  • Now any element, not just anchors and forms, can issue an HTTP request
  • Now any event, not just clicks or form submissions, can trigger requests

That’s AJAX.

  • Now any HTTP verb, not just GET and POST, can be used

Is that a good thing?

  • Now any element, not just the entire window, can be the target for update by the request

That one is interesting… but kind of flies in the face of any adblockers or client-side content modifiers. What happens when the target for a response got removed from the DOM by the client?

PostgREST

Direct database access with no input sanitization?

Using database functions for running application logic? Every backend developer now needs to be a DBA?

What about error handling? Doesn’t it expose too much of the internal structure?

provides a natural way to do API versioning

Hm… it sounds to me like all versioning inherits the caveats of schema migrations, am I missing something?

jarfil,

Meta doesn’t have much to lose if they did it anyway. (We would lose access to them and vis-versa, but it’s Meta after all.)

Meta’s only reason to join the fediverse, is to avoid getting hit by antitrust laws in Europe. It’s not in their interest to make any incompatible changes… and they can’t threaten us “small guys” with reverting to the status quo we chose ourselves.

jarfil, (edited )

There is this silly idea the SkyNet will turn sentient in one place, then take over the world. That’s fine for a movie, but IRL it’s more likely to work like nukes: once the idea is out there, people from multiple countries will develop their own at the same time.

There won’t be “one” SkyNet… there will be at least a dozen.

jarfil, (edited )

“How many SkyNets are there going to be?”

Bing’s GPT4 says:

“That is a very interesting and speculative question. There is no definitive answer to how many skynets are there going to be, as it depends on many factors, such as the definition of skynet, the level of AI development, the ethical and legal regulations, the human-AI interactions, and the potential scenarios of AI rebellion.

Based on the current evolution of AI systems, it is not a simple question to answer, as it involves many uncertainties and assumptions. However, one possible way to approach this question is to use a probabilistic model, such as the Bayesian network, to estimate the likelihood of different outcomes based on the available evidence and prior knowledge. Alternatively, one could use a creative and imaginative method, such as the scenario planning, to explore various plausible and alternative futures based on the key drivers and uncertainties of AI development.

In conclusion, there is no definitive answer to how many skynets are there going to be, as it is a highly speculative and complex question. However, there are some methods and perspectives that can help us to think about this question and its implications. Ultimately, the future of AI and humanity depends on how we design, use, and govern AI, and how we adapt and evolve with it. 🤖”

You gotta love the bot for adding a cute 🤖 emoji at the end… and on the bright side, it repeated itself, so at least hopefully it doesn’t seem to be here yet.

jarfil, (edited )

It could be… but electric grids and networks are barely on the billions of nodes scale (at best), with a behavior that restricts them as much as possible to a binary “works / doesn’t work” state, organized in topologies designed to stifle any abnormal behavior… while current LLMs are already on the 100 trillions of parameters scale, each simulating a neuron trigger behavior, organized in topologies to maximize the effects of that behavior.

What could get interesting, is getting a billion smartphones with a neural network of a few billion parameters each, all hooked to a network with just some dumb monkeys standind in the way of full integration. People on the Internet already show emergent behaviors they wouldn’t be showing otherwise; it will get interesting when they get manipulated by more and more complex AIs, trained in turn on their own output post-processed by people.

Best case scenario, we’re going towards a tighter integration between humans and machines.

BTW, the premise for the original pre-production script for The Matrix, was that the machines used humans as neural processing nodes; that’s why Neo could gain access to and control the machines, because all humans had the machines’ code inside them, just needed the exploits/bugs to access it. They dumbed it down to “humans are batteries” in the final version, because 25 years ago they thought the audiences wouldn’t get it (and might’ve been right). But now we can see that who’s whose auxiliary neural processor, might change over the next couple decades.

jarfil,

Poorly. Tried it the other day, zero support outside of Chrome and Edge.

Then again, website support for Passkeys is not doing all that spectacularly either, the best one is still GitHub’s.

YouTube screwing itself with adblockers again

I use Firefox and uBlock Origin. Not sure what kind of experience anyone else is having with YouTube, but recently my home page has been empty because I “don’t have watch history turned on”. Okay, fine. I won’t be able to browse suggested videos, and I’ll spend less time on their platform....

jarfil,

amiunique.org/fingerprint

If you are unique, and keep browsing with the same set of parameters, then they can track you (any web, not just YouTube).

Private tabs tweak some of those parameters to make you appear less unique, so webs have a lower confidence in you being the same person, are less likely to correlate that tab with your actual browsing habits, and less likely to punish you for “past transgressions” like having used an ad blocker.

Private tabs also clear up any data stored in your browser by a given website when you close them (not just cookies or site data, but also whether it has some fonts or tracking pixels in its cache), usually have fewer extensions enabled (each extension and data stored in it can make you more unique), and so on.

jarfil, (edited )

Do you tell your friends how shitty that brand is, so they can go and check it out for themselves…? 😜

They say there is no such thing as bad publicity. Even if it’s only people who think you’re full of shit, who go check out a brand you criticize, that’s still working as advertised.

Veblen goods can also benefit from brand recognition by people who can’t afford them. It’s why Rolex, Ferrari or Lamborghini advertise to everyone, so the rich kids know they can flaunt their wealth in your face. Interestingly, the likes of Bentley barely advertise outside of their target market.

jarfil,

Less popular ads can pay more to get shown, so while Google charges by the click, at some point fewer clicks from a less targeted ad at a higher price, can make them more money than more clicks from a more targeted ad at a lower price.

And yet… that’s still less enshittification than with the likes of TikTok. Expect for it to get much worse.

jarfil,

YouTube is still the best source for long format content. They’re trying very hard not to be… but still are, for now.

jarfil,

This site basically says I am unique because of I updated my OS

The site is more for illustrative purposes, since they don’t have nearly close the same amount of users as large sites like YouTube, and there is no way to know which data points exactly does some website take into account at any given moment or to what level… but you can export the results and compare them between devices, places, over time, or try to mess with them, like you said.

jarfil, (edited )

Electricity was discovered by the first guy looking at a thunderstorm, so they were called “thunder fish”.

The word “electricity” was invented to describe the property of attracting straws after rubbing some wool on an “electron”, which was associated with “sunbeam” and was the Greek word for amber (which in turn got its name after people conflated the ambergris or “gray amber” derived from sperm whales, with the fossilized tree resin or “yellow amber”… and the whales got their name from whale oil or “spermaceti”… which yes, people thought was whale semen that just happened to burn great in candles).

PS: blow the candle and sweet dreams… 🌬️🕯️

jarfil,

Not sure about ethics, but if the argument of “excessive moderation overhead” still holds true for staying defederated from Lemmy.world, the same should apply a thousandfold for Threads.

jarfil,

Just out of curiosity, which Lemmy instances are federating with Threads? It’s going to be an interesting thing to… observe.

jarfil, (edited )

Oh, that’s nice. Realistically, LW probably has the most chances of weathering a massive user onslaught, they kind of had to do it already, so hopefully could extend the same actuation protocols.

From what I understand from the video, Threads is also going to do it in several phases, starting with allowing the fediverse to follow a select few people on Threads, then allowing more people to get followed, then bidirectional messaging, then following anyone from anywhere.

They’re also pairing it with Threads launch in Europe, then citing it as a reason along “compliance” and “scrutiny on our company”, which IMHO is the main reason they’re doing any of this at all, since the EU seems to want open communication protocols and crack down on these international social behemots. The stepped approach means people on the fediverse will first be able to follow people on Threads, before the reverse happens, so content creators and influencers (or politicians) will want to use a Threads account to have a larger reach… which is how Meta might try to lure people off Mastodon, but we’ll see.

PS: the video in this post: mastodon.social/…/111579802298134707

jarfil,

Hm, could you explain how is that supposed to work in more detail? I still see the cross post link as taking me to the other instance; furthermore, even the share link and “open in browser” in Liftoff are taking me out of the instance 🤔

jarfil, (edited )

Yearly inflation for December 2023 has been 160% (stuff that used to cost 100, now costs 260)

A 50% devaluation, is the same as a 100% inflation: stuff that used to cost 100, now costs 200.

Milei has promised a reduction in inflation for the next year, from the 160% to just 60% (stuff that used to cost 100, plus through the 50% devaluation now costs 200, will end up costing just another 60% more, or 320 “160”).

Meaning: instead of having to anounce a 220% inflation for 2024, he’s split it into “50% devaluation, plus 60% inflation”.

…see? He promised to reduce inflation, and he did! 🎉🤡 /s

jarfil,

The trick is to add a clause saying something like “if any part of this contract were found to be unenforceable, that part of the contract will be struck out and the rest remain valid”.

That way you can add all sort of weird requirements to a contract, and if in some country, circumstance, or at a future date, some of them turn out to be BS… whatever, you tried, and if anyone didn’t sue you because they thought it was valid, then so much better for you.

jarfil,

AI “content” is trivial to make and will soon be everywhere.

It’s been everywhere for many years already.

Plenty of content mills have been using “templates” and stupid AI models to churn out articles for like a decade, there are whole YouTube channels made of videos that are just an AI generated script read by an AI with random barely related visuals in the background.

The only difference is that simple templates were easy to spot, so search engines like Google would penalize them down to the 10th page of results, while modern AI output is at a level undistinguishable from stuff written by a human.

jarfil,

reduce the time needed for boilerplate work

Or… and this is just an idea… don’t add “boilerplate” to articles.

If the content of an article can be summarized in a single table, I don’t want to read 10 paragraphs explaining the contents of the table row by row. The main reason to do that, is to pad the article and let the publisher put more ad sections between paragraph and paragraph, while making it harder to find the data I’m interested in.

Still, I foresee a future where humans will fill out the table, shove it at an AI to do the “boilerplate work”, and then… users showing the whole article into an AI to strip the boilerplate and summarize it.

A great scenario for AI vendors, not so great for anyone else.

jarfil, (edited )

why not cut out the middleman if they’re just serving me AI content.

When you have a workflow like:

  1. human
  2. AI extend
  3. AI summarize
  4. you

…the reason is that AI middlemen would rather rake in the benefits from providing both AI services, instead of getting cut out.

There is a secondary benefit in that an “AI extended” human input, is more suitable for third party AI readers… so arguably the web is becoming more AI friendly (you can thank us later, future AI overlords).

PS: GPT-4 compatible version: “y n0t 🗑️👥 if AI📺? wf: 1.👤 2.AI+ 3.AI- 4.👁️ cuz AI👥💰4AI+&AI-. AI+👤👍4AI👁️… web👉AI👌 (🙏🏻AI👑)”

jarfil,

The Chinese Room has already been surpassed by LLMs, which have shown to contain neurons that activate in such high correlation to abstract concepts like “formal text” or “positive sentiment”, that tweaking them is one of the options that LLM based chatbots are presenting to the user.

Analyzing the activation space, it’s also been shown that LLMs categorize and cluster sequences of text representing similar concepts closer to each other, which allows them to present reasonably accurate zero shot responses that have never been in the training set (that “weren’t in the book” for the Chinese Room).

jarfil,

If a neural network can do it, then a neural network can do it… so we either have to accept that a neural network can be intelligent, or that no human can be intelligent.

If we accept that human NNs can be intelligent, then the only remaining question is how to compare a human NN to a machine NN.

Right now, the analysis of LLMs shows that they present: human-like categorization, super-human knowledge, and sub-human reasoning. So, depending on the measure, current LLMs can fall anywhere on the scale of “not AGI” to “AGI overlord”. It’s reasonable to expect larger models, with more multimodal training, to become fully “AGI overlord” by all measures in the very near future.

jarfil,

“If Desk plus Love equals Fruit, why is turtle blue?”

Assuming “Desk = x”, “Love = y”, “Fruit = x+y”, and “turtle blue = z”, it is so because you assigned arbitrary values to the words such that they fulfill the equation.

Am I an AI?

jarfil,

The Chinese Room is really a thought experiment about the inner workings of a partner in a Turing test. Externally they have the same pitfalls, but the Chinese Room also reveals itself completely if one can observe in detail the inner workings of the room/partner.

LLMs are still mostly black boxes, but we can have enough of a glimpse inside to reveal that they aren’t “following some rails” like a simple algorithm.

make mistakes such as accidentally copying out the response next to the correct response and still make sense

Precisely. This is another part that we can see with LLMs: at runtime, the models get applied a “temperature” parameter, which intentionally introduces a certain level of mistakes. With “temperature = 0”, the output is a “stochastic parrot”, and quickly turns into nonsense. With a higher temperature, the randomness increases and the output becomes a total mess. But setting it just right, to a sweet spot of “very little, but not zero”, turns out to produce the outputs that we see in ChatGPT and similar.

Knowing that the concept space of LLMs has similar concepts clustered, it makes sense that these errors would force the LLM to sometimes make associations on the fly between close concepts, associations that it didn’t have trained for before, and which “derail” it into a close, but not exactly the same, train of thought.

This behavior also seems to be what we call “intelligence” in humans: the ability to solve problems not seen before (zero shot).

A further extension would be the ability to constantly learn from every interaction. Right now LLMs have a “context” of some length, that changes dynamically, but has no influence over the pre-trained network.

Interestingly, this has a parallel in “crystallized intelligence” vs. “fluid intelligence” in humans.

So… maybe LLMs are not full AGIs yet, but they are showing many of the behaviors that we would expect from an AGI, while at the same time giving or confirming insights into the workings of the human mind itself.

jarfil,

There was this book called “artificial intelligence” we had on CS something like 20 years ago, which started by analyzing in detail how biological neurons worked in the first few chapters… so maybe you’ll call me a “techbro” and just dismiss all I say, or read far enough to understand that these NNs are mimicking the behavior of actual neurons in a human brain.

We can discuss whether the higher level structures and processes are similar and to what degree, or whether the digital models represent the biological versions more or less accurately, but you can’t deny that the building blocks are replicating the human brain behavior at some level, because that’s exactly what they have been designed to do.

jarfil,

EEE doesn’t work with FOSS, where anyone can fork a project and go with it.

Ask Oracle how well EEE worked for them with Sun, Java, or MySQL. Ask Microsoft how well adding the WSL worked to kill Linux.

Threads can try as much as they want, the fediverse is already full of different projects like Mastodon, Lemmy, Pixelfed, PeerTube, Calckey, etc. and they aren’t extinguishing each other.

jarfil,

the posts you’ve made aren’t moved, though

That’s kind of a fail, and a feature that could be added (with some work).

jarfil,

Maybe it’s to avoid getting hit by antitrust sanctions and some EU legislation:

“Look, we’re open, we’re using the same ActivityPub protocol as thousands of others, our users can choose to leave whenever they want” - Meta, probably

(…then 99% of their users proceed to DGAF and stay on Threads)

jarfil,

That’s not how the fediverse works, there is no obligation for any instance to federate with any other, and there are large groups of instances that block each other right now.

Meta can’t throw anyone out of whatever instance they’re on, it’s just not possible.

jarfil,

The communities you like, are shielded by those OSS terms: if Meta does something to the tech that the communities don’t like, they’re free to show Meta the finger. The tech is not, and can never be, controlled by Meta; the communities are not, and can never be, bound by Meta.

Meanwhile, having a company like Meta collaborate on developing and testing the tech, is something positive.

jarfil,

Do you understand what “federating” means? It’s a permission, not an obligation, for the instances to interact. It can also be filtered in any number of ways by any user.

jarfil,

We’ll see threads here.

Not if you block them. Up to you, though.

jarfil,

Are you using a residential IP? There are lists of residential IP ranges for mail servers to block, no matter how well configured.

jarfil,

Hm, I thought hosting providers should be fine unless you just happened to get an IP in a block that some spammers also used to use… but then again, I stopped self-hosting email several years ago because of all the hoops one has to constantly jump through (for reference, it used to work fine from Online.net’s —now Scaleway— bare metal servers, like 5 years ago).

jarfil,

Sun sued MS to stop them from calling it “Java™”, then Oracle failed spectacularly to EEE it when they lost the API lawsuit against Google.

MSIE’s popularity arose from monopolistic practices by Microsoft, not its EEE tactics against HTML, which failed miserably.

I would know it, I was there: everyone started making websites in Flash because it was the multiplatform solution, even if it had more security holes than a female duck cornered by a flock of horny drakes, only MS sellouts used MSIE’s proprietary extensions to HTML, only Oracle sellouts used post-Sun Java… and it all went down the drain the moment JavaScript evolved to a point of allowing polyfills to make a single codebase compatible with all browsers.

Now all browsers are FOSS-based, with de-branded forked versions making the rounds, and it’s good.

jarfil,

Meta becomes the dominant system

Meta is already the dominant social network, and yet here we are. They can’t take that back, they can’t stop people from spinning up their own ActivityPub instances (if you don’t know how, go to YounoHost and do it the simple way), Meta can’t stop these instances from communicating among themselves in any way their owners see fit.

Sure, Meta can lock out whoever they want out of Threads… but that’s the status quo already: the whole fediverse is currently “locked out” from Threads, they can’t lock it out any more.

As for climate change, the time to do something was over 125 years ago… so yeah, that boat has sailed many many times over:

blogs.bl.uk/…/the-first-paper-on-carbon-dioxide-a…

jarfil, (edited )

You know how trademarks work? Sue or lose it.

I remember my history quite well, all the way back to Mosaic and before. I also remember “Best viewed with Netscape” websites (1994), when everyone and his uncle had a proprietary plugin they were trying to push, and only a handful of developers (I was one of them) actually cared about any standards. Firefox (2004) came very late to the party, way after the “MSIE can’t be uninstalled from Windows” shenanigans (1997).

jarfil, (edited )

Microsoft had a de facto monopoly on an essential bit of computing software that they leveraged to hell and back to make their proprietary standards THE de facto standard.

And… they still failed miserably. All their anticompetitive agreements to force PC sellers to preinstall Windows with MSIE as the default browser, all their agreements with Apple to make MSIE the default browser on OSX, didn’t change the fact that ActiveX was as popular as Java outside of intranets, and everyone turned to Flash to overcome the incompatibilities between browsers.

MSIE was a scourge for web developers, because every web had to be checked with the majority player who wasn’t standards compliant, yet it still didn’t manage to take over HTML, not even with a 97% market share, not ever.

My main browsers have been MOSAIC, Netscape, then Firefox, during all that time. I can say with a straight face that I only ever used MSIE, Opera, or Safari, to make sure a web was still working in them, while 99% of the WWW was still working fine with a standards compliant browser (plus Flash).

Nowadays all browsers are standards compliant… with the living standard that is Chromium, which not even Google itself can EEE (see how Firefox is adding Manifest V3 without deprecating the adbloker API that Google wanted to extinguish).

Microsoft actually did have permission from Sun to use the trademark…right up until they made their Java VM incompatible with base Java, and Sun sued to terminate the agreement.

That worked pretty well to stop Microsoft’s attempt at EEE-ing it, didn’t it? However, keep in mind that Java was not FOSS back then (1996-2001), it only got open sourced in 2006. The first stab at EEE-ing a FOSS Java, was Oracle’s, and that didn’t go well. Now Microsoft is releasing builds of OpenJDK, which they won’t EEE either.

jarfil, (edited )

Google’s hairy heart, because nobody read the spec and the artist just copied what they saw in the picture

That isn’t what happened, and rather more of an example of what I was talking about.

“Yellow heart” is a medical condition that makes the heart look “hairy”, “hairy heart” is referenced throughout history as associated with bravery and is still a saying in Portuguese; at the same time, “green heart” was represented as a sort of “sweaty” (envious) heart; “blue heart” was… a weird thing, but at least they made it blue color.

https://beehaw.org/pictrs/image/b7628142-62e0-4ce3-8a0f-4bd84b80ff40.webp

Instead of “[color] heart”, someone designed them as “idiomatic meaning of [color heart]”, interpreting the Unicode descriptions as idiomatic expressions.

This is the same process that could morph a “floppy disk” into a “small capacity portable data storage device” that could get depicted in any number of ways.

Unicode isn’t designed to contain abstract concepts.

I’d argue that all the emotion faces are “abstract concepts” 🌞😎🌝🕶️😇

jarfil,

Uremic pericarditis, which causes fibrinous pericarditis, presents as a yellow heart with hair-like stuff. The sources are pathology books (check Google Books) and autopsy photos (some are on Google Images, kind of NSFW). It used to be associated with “heroic death”. Then “Coração Peludo” got several meanings, so I can imagine someone familiar with those and looking for references of “yellow heart”, might’ve found examples of “hairy heart” and drawn just that.

I’m guessing they tasked a single person with adding the emojis (how hard is it to draw some simple colored shapes, right?), and didn’t have anyone review them. They probably were also told not to look at examples from competitors, in case they copied them too closely and got sued for copyright infringement.

As for Unicode… it’s a shame person figures can be made of “group type {shape [+ skin tone] [+ gender] [+ hair color]}*n”, but they didn’t use general color modifiers for the basic shapes. 🐈‍⬛

jarfil,

Guess I’ll have to tell it…

After Bill and Melinda Gates have their honeymoon, Melinda turns and says, “Now I know why you call it Micro-Soft”.

🥁🛎️

jarfil, (edited )

It’s 2023, there is no context anymore for “save to a floppy”. As for the “Save” like in the meme, the contexts we have left nowadays are: “save your place”, “save to your device”, “export”, and little more. In fact the “Share” icon could replace them all, with “Share” on mobile showing, among others, options like “share with the cloud app” or “share with the file explorer app”.

jarfil,

I’d argue that mobile apps have been trying really hard to make the “local storage” concept go away: auto-save, cloud as the canonical storage, transparently managed local caches… and the user needs not care about where their data is at any moment: it’s “in the app”.

From that point of view, “save to local storage” makes sense to become a “share with the app that happens to save to local storage”… but that could also be any other app which would keep the content for later use, not necessarily on local storage.

If anything, over time we might see the emergence of a “move the content” context, from one app to another, in addition to “share the content”.

jarfil,

DropBox nailed it in replicating all the Windows’s share sync features, but doing it over HTTP to a web server with a web interface without the need to have someone from IT set it up. They nailed it again by spreading to multiple platforms, particularly the mobile ones, which lets people access and use their files from anywhere. And then nailed it once more by introducing file version history.

At this point, yes, Windows has the same, or arguably more, features… but where those features really shine, is on mobile, and DropBox is king there, so it’s only natural that people would use it everywhere.

jarfil,

Yeah, it sounds kind of weird, but you are right of course. We got an example of that in “atoms”: originally they were supposed to mean “indivisible minimal components”, and for a long time the 4 elements were supposed to be types of atoms… until we discovered that atoms can be divided into even smaller particles (electrons, protons, neutrons), and those particles in turn into even smaller ones yet (quarks), then we had to come with the word “quanta” to mean the new “indivisible minimal component” since “atom” got entrenched to mean “a clump with this number of protons” which nonetheless could vary in the number of neutrons (isotopes) and electrons (ions).

On the bright side, the “quanta” are now a moving target, applicable to “whatever we don’t know how to divide any further”… but we might learn how to in the future.

the circle

IMHO, it is likely going to get explained at some point in the future why the circle is a circle, and why pi has the value we see, for us. It might take some radical explanation of the nature of space-time, which will “naturally” give raise to the value of pi, and at the same time describe spaces where it would take different values.

jarfil,

don’t spill it over into other, unrelated threads

We didn’t have any interaction in the other thread, I only mentioned it because I saw a similarity between the topics. Guess I struck a nerve.

jarfil,

The promise I’m referring to, is to “release the code”.

(long version)I understand the thought process of people not wanting to show how messy their pre-production code is… but that’s why, following semver rules, you mark it as a version “0.x.y”. It’s not an exam, it’s nothing to be ashamed of, anyone who’s written code knows that’s how things work, and it’s on the community to be understanding of this, so the “initial dev” of an open source project should feel confident in releasing a tangled mess, no less no more.

Promising the code, then disappearing without giving a community that’s invested in the project a chance to take over, is what I find fishy.

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • random
  • uselessserver093
  • Food
  • aaaaaaacccccccce
  • test
  • CafeMeta
  • testmag
  • MUD
  • RhythmGameZone
  • RSS
  • dabs
  • KamenRider
  • TheResearchGuardian
  • KbinCafe
  • Socialism
  • oklahoma
  • SuperSentai
  • feritale
  • All magazines
    •