jarfil

jarfil , 1 day ago

Could mean FOSS but they keep the trademark.

jarfil , 1 day ago

The real solution is, a heuristic analysis of the phone's gyroscope and accelerometer data.

Marketing calls that, "AI".

jarfil , 1 day ago

It isn't clear which of these features use Google servers and which ones don't. The "Find My Device" definitely does, and has no place in AOSP. If they're actually using AI to compare phone state with some tracked "habitual" behavior, it may also have no place in AOSP, but who knows.

jarfil , 1 day ago

It's kind of both Google's and manufacturers responsibility. Google has made available a Dynamic System Updates feature:

https://source.android.com/docs/core/ota/dynamic-system-updates

https://developer.android.com/topic/dsu

...but it requires manufacturer support to allow adding custom keys.

jarfil , 19 hours ago

"AI" used to mean "whatever we don't fully understand yet". A lot of processes have walked the path from "fantasy" to "AI" to "algorithm". Doesn't need to be non-deterministic, the original tic-tac-toe playing software was "AI" at the time.

Until we get some AGI, the term "AI" will remain a moving technological target, and a static marketing target.

jarfil , 1 day ago

AI can also understand extra weights for hand picked sources of truth. Whether you then agree with the choices of whoever is doing the hand picking, is a separate matter.

jarfil , 1 day ago

Is coming, and more.
Very good video, with good points. Slightly outdated already, with AutGPT being a thing.
What's coming, is going to be orders of magnitude more than what they predicted in that video.

jarfil , 1 day ago

Yup.

They also correctly identify the function of LLMs as a glue between siloed AIs. We have barely seen the beginning of that, but as the AI race continues, it seems likely that some LLM models will be created that will have less human language, and more "interop language". Where nowadays LLMs can be somewhat probed for words and relationships, we'll have zero chance to probe an LLM using tokens that are part of some made up (by the AIs) interop language. Black boxes inside black boxes.

A naive approach will be to "democratize AI", and that will surely be better than centralized AIs responding to every query... but won't solve the deepening of inscrutability.

One point made me chuckle: when they showed the graphs for cualitative jumps above certain network size. Recently someone commented about the "diminishing returns" and "asymptotic growth" of making a LLM larger and larger... but also so it was in these models: diminishing returns all the way up to a point... followed by a sudden exponential jump until the next asymptote. The truth is we don't know where the asymptotes and exponential jumps lie, we don't even have a remote hypothesis about it.

jarfil , 19 hours ago

Yeah, they call it A-Z testing, because A/B wasn't enough, and AIs can fill A to Z cases with ease.

jarfil , 1 day ago

Was that on the Insider Dev channel, a Preview channel, or did it stop asking after the first time?

I haven't seen a single "ad" of those since I booted Windows 11 Pro for the first time.

jarfil , 3 days ago

Senior talent tends to be "T people", while juniors tend to be "I people". Removing that T scaffolding, is how corporations end up like Boeing or the Titan.

jarfil , 3 days ago

How to survive Windows in 2024:

• Windows 10/11 "Pro" is codeword for "not a joke". Always use Pro or better for anything serious.
• DO NOT ever install any "Preview" updates, unless you want to become a betatester (comes with config reset and telemetry).
• DO NOT join the "Windows Insider Program", same reason as above.
• WSL2 is your friend, so are Cygwin and WinGetUI/UniGetUI.
• ONLY install official software, and FOSS software.
• DO NOT install anything cracked, or run any cracks, serial generators, or anything that promises you a free gain, ever.
• Windows Sandbox, is also your friend.

Sandbox is particularly nice, since it may run a "Preview" version of Windows, but sandboxed, without polluting your main install.

If you have an Intel 12th gen or higher, you may be able to run it with TME-MK for extra isolation between the systems.

jarfil , 3 days ago

Rule of thumb:

• Linux: you set a system from bottoms up, adding/enabling layers and features.
• Windows: you set a system from top down, locking/disabling features that shouldn't be available.

Windows has tools to control and restrict updates/installs, with a full centralized logging system. You will rarely find log files, it all goes into the centralized log (think systemd log). Some failed installs may leave behind log files in temp, so be quick at getting them or they'll be gone.

If you have non-Windows programs and tools installed, you can use either the msstore, winget, or 3rd party choco and scoop, or all of them via WinGetUI.

Many admin tools are only available on the CLI via PowerShell. Some may or may not conflict with how their GUI counterparts work.

Additional built-in tools... are "secret". You may get a glimpse at some of them by checking the "install fix" tool or similar, but it may use non-backwards-compatible calls, which is why they're not made official.

jarfil , 3 days ago

Hm, from a privacy point of view, what's the difference between this, vs. a private/incognito tab with a pi-hole?

I mean:

• pi-hole eats the ads
• private/incognito eats the cookies
• the sites can still do basic tracking via the video feed itself

As for functional differences.... right now, the player seems to be missing the skip ±10sec feature, and the full-screen with rotate.

On the other hand... multi-site private bookmarking, would be interesting.

jarfil , 5 days ago

Chromium, Firefox... if you open the dev tools, you can edit everything, with it showing in the browser in real-time (WYSIWYG).

Firefox Developer Edition has some extra tools and debugging modes, but some are redundant if you're using VS Code.

If you're looking for a Dreamweaver-like thing, where you could drop elements with minimum HTML writing... you may want to check Seamonkey Composer.

For a simple personal website though, I'd recommend using a markdown editor, then either export it through a template, or have a template interpreter on the site, like GitHub Pages.

jarfil , 5 days ago

BlueGriffon has been discontinued in 2019, not sure whether there is a fork.

jarfil , 5 days ago

I didn't know that Slashdot added something like this... just googled for WYSIWYG FOSS editors to make sure, saw that BlueGriffon's site had a farewell message, and checked the GitHub.

FF Dev and Seamonkey, I have installed along VS Code 🤷

jarfil , 4 days ago

Seems like federation filtered it out, it shows on your instance but not on this one 🤷

jarfil , 5 days ago

Neuralink has a technology that specifically addresses two of the main issues with BCI: data density, and implant effective duration.

There are more issues, but it addresses those two in particular, which is something quite interesting to see, and can be turned into patents that can be sold to other BCI initiatives.

The rest of Musk is... well, he's kind of an "unstable genius", with enough money to blow on random moonshots, marketing stunts, and random publicity. Honestly, if I had his money, I'd probably do the same: build a few core businesses, then go on tangents to see what sticks to the wall. It can all still be seen under the general theme of "colonizing Mars" though, which is a guiding starshot as good as any, with Hyperloop and Boring company having kind of exhausted what can be done on Earth, Tesla being a borderline failure, SpaceX, StarLink, or indoor farming working pretty well, and X being an experiment at social manipulation.

jarfil , 5 days ago

TikTok is a weird beast. It can at the same time show the destruction in Gaza, Israeli soldiers poking fun at it, ASMR videos, mindless looping footage, fake AI idols, underage girls asking for payment from strangers, and siphon engagement data to train Chinese propaganda bots.

It's the closest thing to "shove everything into a bowl, then shake"...

jarfil , 5 days ago

It's not just to destabilize "American" politics, it's a series of worldwide campaigns to destabilize all information flow, to sow doubt and confusion among everyone, then out of the blue present an aligned front to push a certain narrative.

If people are kept in a "flux state of distrust", they're easier to convince when suddenly a bunch of their sources agree on some point, "it must be true if conflicting sources suddenly say the same".

jarfil , 4 days ago

The biggest problem with Ukraine... is that they aren't fully detached from Nazis:

• During WW2, Ukraine was allied with Nazis and fascists, helping them exterminate Poles
• 21st century Ukraine, still uses Nazi symbology, the fascist salute, a fascist hymn, has set national support for WW2 Nazi combatants, and even their national shield is a fascist remnant.

All of that has nothing to do with the Russian invasion... but it does give Russia's propaganda machine an awesome excuse. It's just too easy to get people hooked up with some actual facts, then get them to do a leap of faith and fall straight into full propaganda... and Russia knows it.

Israel and Palestine is a particularly juicy case, where there are really shitty groups coming from both sides, ending up like an "all you can eat" buffet for every propaganda machine out there. No matter what narrative one wants to spin, chances are they'll find a latch point in the Israel vs. Palestine conflict, even contradictory ones for different audiences.

jarfil , 5 days ago (edited 5 days ago)

That's a very long article to decry some echo chambers and manipulations... from inside a different echo chamber and manipulation.

Yes, China, Russia, and friends, are playing dirty to subvert the international influence of the USA.

No, Ukraine is not free of Nazis, and Israel is not free of genocidal fanatics.

Unfortunately, insisting on calling "democratic" or "free", countries that are not, does a gross disservice to democracy and freedom as a concept. Acting like there is no more work to do, gives a wedge for unfriendly regimes to pull quasi-democracies apart.

jarfil , 5 days ago (edited 4 days ago)

Copyright can only be waived in the US by dedicating the work to the Public Domain. In most other countries, it can only be assigned or licensed to someone.

The "standard practice in all forum software since practically forever", has been to include a very broad use license on the work, without switching the copyright holder, in order to protect the forum owner from liability.

The GDPR is about a very broad take on "privacy", where the rights of "access, modification, and removal" get extended to any "personal information", no matter whether it's "personally identifiable" or not.

Kind of a two birds with one stone situation.

jarfil , 5 days ago

Now do Reddit 😈

jarfil , 5 days ago

This seems to have been addressed by the judge:

By attempting to exclude Bright Data from accessing public X posts owned by X users, X also nearly "obliterated" the "fair use" provision of the Copyright Act, "flouting" Congress' intent in passing the law, Alsup wrote.

jarfil , 7 days ago

Impostor syndrome is feeling like you aren't qualified for a post that you are.

The opposite, would be feeling like you are qualified for a post you are not.

What you describe, seems like over-competence, rather than either of those. Congrats 🎉

jarfil , 7 days ago

No.

Whatever the next big thing is, money will pull in the scammers who will turn it into the next cancer.

It's always been like that.

jarfil , 7 days ago

You will need an LLM to tell that apart, so... 🤷

jarfil , 7 days ago

Neuromorphic hardware is going to jump many orders of magnitude over classic hardware. When we get a RAM that can execute multiple layers in parallel at once, per clock tick, we'll see whole AI ecosystems cooperating to get a solution in a fraction of the time a single modern NN would take.

jarfil , 7 days ago

Rule of headlines? 🙄

No, it's not peaked out.

• A simple path forward, is to go from classifying single elements of training data, to classifying multiple elements and their relationship in the training data.
• Slightly less simple, is to gather orders of magnitude more data, by just hooking the input to an IRL robot.
• Another step, is for the NN to control the robot and decide which parts of the data require refinement, and focus on that.

There is a lot of ways to improve data acquisition still on the table, it isn't going to stop at creating large corpora and having humans to fine-tune them.

jarfil , 5 days ago

The orders of magnitude will come from the RAM running a whole layer at once in "a single clock", without the need for a processor to execute any of it. It's conceivable that multiple layers could be written/"programmed" into neuromorphic RAM, then a processor could just write the inputs, send an execute, move data from outputs to the next inputs, and repeat for all layers.

For example, an nVidia A100 goes up to 1,200 INT8 TOPS with 80GB of RAM at 1500MHz... but if the RAM could execute a neural network directly, that could raise it up to 80G*1.5G=120,000,000 INT8 TOPS, or 5 orders of magnitude.

jarfil , 4 days ago

Current research points to memristors, which can work both as memory cells, and as weights in a n×m grid representing a fully connected n->m layer that executes in 1 clock. I forgot which company was showing prototypes since pre-covid... and now Google is so full of wannabes that I can't seem to find it, oh well.

Cerebras is at the limit of SRAM, that's true.

Spintronics could be the next step, but seems to be way less ready for production.

Higher dimensionality would be nice, but even at 2D, being able to push multiple processes at once, through multiple n×m layers, would already give those 5 orders of magnitude, at least for inference. Since training also involves an inference step, it would speed that too, just not as much.

Self-training would be the next step after that... I don't think I've seen research in that regard, but maybe I've just missed it.

jarfil , 4 days ago

It's a "push as much data as a baby gets to train its NN" step, which is several orders of magnitude more, and more focused, than any training dataset in existence right now.

Even with diminishing returns, it's bound to get better results.

jarfil , 4 days ago

That's not how watching the video or reading the paper works either.

Whatever.

jarfil , 7 days ago

There is a basic misunderstanding in OP's formulation: a "firewall for" is something one needed with Windows XP and earlier, as in a piece of software that acted as a firewall; nowadays, both Windows 7+ and Linux come with a built-in firewall, that one might want a "GUI for {}'s firewall".

Whether people feel more inclined to explain the misunderstanding, or to just spew a "you can't" that's technically correct but unhelpful... YMMV, different people are different and may be of different mood at different times 🤷

jarfil , 7 days ago

There seems to be a misunderstanding:

• A "firewall for" is something one needed with Windows XP and earlier, as in "a piece of software that acted as a firewall".
• Nowadays, both Windows 7+ and Linux come with a built-in firewall, that one might want a "GUI for {}'s firewall".

One of such GUIs, is TinyWall, that is also FOSS (GPLv3). I see people have suggested some more.

To be precise, all these options are inferior in functionality to firewalls like ZoneAlarm... but since you're asking for a non-tech friendly solution, they should be adequate.

jarfil , 7 days ago

Because spending years setting up a system using nothing but open source from the start, you'd still not approach what windows can do out the box with far less effort.

This is a flawed argument, the opposite of:

Because spending years stripping(*) a system from adware and bloat, you'd still not approach how slim Linux can be out of the box with far less effort.

Just pick a target, then use whichever tool gets you closer to it... and I think you know it, no need for a rant.

(* there are actual tools to strip and reset the tracking and ads in Windows... obviously for people who accepted to get early updates, install the "Preview" versions, and haven't read that it means they're now betatesters with telemetry enabled 🙄)

PS: settling on a "single GUI", is kind of ironic given the multiple GUI versions of the control panel in modern Windows.

jarfil , 7 days ago

TinyWall is a simplified GUI for the Windows firewall... some may like it, some may not.

jarfil , 7 days ago

TinyWall doesn't change the firewall, it's just an alternative GUI... like setting it from PowerShell.

jarfil , 7 days ago

Does Suricata or Snort allow the user to block per-process outgoing traffic?

jarfil , 4 days ago

Process-level filtering is to avoid exfiltration from environments where "all processes run as the same user, with full access to all other processes"... which, unfortunately, are still most of them.

DPI is nice to stop incoming attacks, and to detect suspicious outgoing traffic, but it's kind of late when the data is already on the wire, and you won't be able to stop all possible kinds of traffic that way.

jarfil , 9 days ago

automatically train one that can "learn" to generate similar-looking data by just being fed a bunch of files to emulate

Sounds like a job for a "compression prompt" for ChatGPT... [and thus, the AI wars began]

jarfil , 9 days ago (edited 9 days ago)

An interesting video about the whole project:

Neom - The Line - The Rise and Fall of Saudi Arabia's Linear City

TL;DR: It's impractical, unfeasible, and the closer you look, the more it looks like a scam.

jarfil , 9 days ago

In the previous comment, you put it marked down as "code". Not sure what client you and @apis are using, it should show as inline code snippet.

jarfil , 5 days ago

From the video, it looks like part green washing, part fearing a post-oil word, part corruption, and part scam.

If SA was serious about their own future, they'd do like Norway with its Oil Fund. Instead, they've been squandering their oil income, and it seems like these projects are aimed both at scamming international partners as well as their own funds, with green washing as a facade.

jarfil , 10 days ago

Tuvix was made, Tuvok and Neelix were recovered. Such are the paths of the universe, choices are just choices, right and wrong lie in the eye of the beholder.

jarfil , 10 days ago

Redefining identity in terms of cell organization, would definitely solve some ethical issues like human cloning: different structures, different individuals.

Now, the remaining question would be, how to "read" the structure. We can sequence DNA from a tiny sample, but disassembling people wouldn't be... practical.