Skip to main content


Introducing my competitor to LLMs: SKM

The Small Keysmash Model will generate bespoke keysmash with almost no overhead!

It's a neural network I trained on me keysmashing once.



@Hypolite Petovan I am looking to buy a bunch of Lego for my son for his birthday. Is there a good tool for looking at a general view of like, weight to volume, and is there a standard bulk-lego supplier?
in reply to Andreas vom Zwenkauer See

@Andreas vom Zwenkauer See I have a hard time taking these sets with a very low price per part seriously, in part because they're standing on the shoulders of the elephant in the room. For example, I sold today a couple of LEGO sets that were released in 1993, so 30 years ago, and not only the parts I found for these sets in the bin I sorted were in great condition (albeit a little dirty), I was able to replace most missing parts seamlessly with much newer parts. These for me are proofs of an ongoing quality commitment dating back at least 30 years that none of the newer LEGO-compatible brick ventures, however well-intentioned, can ever claim to match.

One of the reasons the LEGO second-hand market has always been very strong is the normalization LEGO introduced in the 70s (!). As a result, LEGO parts always have been a known quantity, and you can build a business based on trading genuine LEGO parts because of this ongoing trust in the product itself. Is LEGO overcharging for new sets? With such an incredible engineering legacy for what was considered like mere children toys until very recently, they damn well can, and I'm absolutely not ready to quit this field because of outlandish price-per-part claims.

Let's talk about the BlueBrixx second-hand market in 30 years, shall we? 😄

in reply to Hypolite Petovan

Let's talk about the BlueBrixx second-hand market in 30 years, shall we?


Yes !

;-)


silverwizard reshared this.


You guys! It's here! It's finally here! Party like it's Y2K38! 🎉

calendar-australia.com/holiday…

in reply to Jan Schaumann

now you can say all you like, but this "date" actually column *is* Y38K compliant. What else would be the reason to state that Easter Sunday falls on a Sunday every year, were it not that after the Y38K Easter Sunday actually is a Thursday!
in reply to Jan Schaumann

The fun bit for me is I actually wrote a holiday calculator for myself which obviously include Easter days so seeing this I had to check.... Mine goes to year 9999 before throwing a ValueError: year 1000 is out of range



The fact that I was in BitCoin in like 2009 (it was before I started dating my wife, and that was May 2010, which leave a very thin window), and did AI stuff even before that, makes brinigng me into Crypto and AI chats incredibly weird.

I've been here since the beginning, and seen every grift and been calling it a grift and a con for longer than many people have been in the space.

in reply to silverwizard

@silverwizard Look, he already ruined electric cars, flame throwers, Mars colonization and Twitter, there's nothing out of his deathly grasp.


Having #UnknownArmies thoughts - thinking about a data mining tech company run by a cabal of Cryptomancers - not selling the data or advertising, just collecting the secrets of millions and using it for charges

silverwizard reshared this.


Either capitalism dies or the planet does

silverwizard reshared this.



My RPG group needed to evoke a dive bar correctly, so I coined the phrase "greasy beer"

This is literally the worst concept for a dive bar

in reply to silverwizard

not a dive bar, but eons ago went to the El Macombo in the middle of the day and got a pitcher of beer that was at least 1/4 foam...and the beer was pretty brutal. Funny for a bar with such a good rating. Haven't been back since, then again it's been ages since I been in Toronto for more than a few hours.
in reply to Chaslinux

@Chaslinux I was at a concert at the El Macombo in like ... 2007, but that was not my experience. Tragic.



silverwizard reshared this.


Police are asking for your help in finding a missing Waterloo person - Pamela, 37 - #Waterloo #Ontario #missingperson #missingpeoplecanada #missingincanada

Regional police are asking for your help in finding a missing person.

Pamela is 37 years old and is around 5'0'' tall and 130 pounds. ...

More Info: missingpeople.ca/waterloo-regi…

reshared this



Thinking about this again web.archive.org/web/2013092322…

That was 4 days after the domain was registered, and the day the phone was released....

This entry was edited (1 year ago)
in reply to silverwizard

When I see people arguing over faceid ve touchid, I find myself considering suggesting that if biometrics are secure enough for their use case then just leaving the phone unlocked should also be up for consideration.
in reply to ⛅ w chance of bears

@⛅ w chance of bears I use fingerprint because I want to make sure I don't pocket dial.

But yeah - seriously - it's... upsetting to me

So much just weird security choices




youtube.com/watch?v=EiZhdpLXZ8…
It's longrunning character Hat Dan! The Dan with a Hat



It seems incredibly impossible to email a tailor and be like "can you tailor me a pair of baggy cargo pants?"

It just...

I'm sure they'd take the commission, but it just seems wrong


silverwizard reshared this.


This blog post seemed very normal until I hit this bit

GITHUB ACCIDENTALLY POSTED THEIR PRIVATE KEYS TO GITHUB

THERE IS LITERALLY NO ONE ON EARTH ABLE TO USE THIS PROGRAM SAFELY

github.blog/2023-03-23-we-upda…

reshared this

in reply to mcc

so THAT'S why i got the known hosts warning! incredible. name a more iconic duo than programmers and checking sensitive data into source control.


As AI art gets better and better at photorealistic art, it gets worse at *art* and better at *deception*. But of course, as it gets worse, it gets worse at *art* and better at *making garbage*.

There is an obvious solution to this dilemma.

in reply to mike

@mike No I agree
I was thinking about this image I saw of someone next to a TV with a person in the TV coming out to hug them.

If an AI made it, it would look bad and dumb

If a human made it, it means something weird and personal

@mike
in reply to silverwizard

Kind of like how sometimes who the artist is matters. The context of a work’s creation changes its meaning. But, a human still made potentially thousands of creative choices in the creation of an AI generated image, in much the same way a photographer or director does. An AI didn’t decide that out of it’s quasi-infinite potential outputs, a certain one should be circulated on social media with a certain presentation.

…yet




I love the idea of Joke Theft and shit on social media.

Someone makes a joke, and then someone else riffs on it. Or tells a similar joke. And people get super up in arms about the joke being "stolen".

Copyright and Clout just rotting their brain until they ignore the idea of culture

in reply to silverwizard

I love how so often the "how dare you steal this joke" posts are about such low-hanging fruit, the most obvious riff a person could make about some current thing jokes. Maybe they saw that joke and reposted it, or maybe dozens of people came up with the same thing independently.
in reply to ⛅ w chance of bears

@⛅ w chance of bears Yeah

And like, maybe someone sees a joke, and would prefer a slightly better one and riffs.

you don't need to cite your Joke Sources

MLA Style Joke Citations


silverwizard reshared this.


If you have an SSH server and worry about disclosing your private key: OpenSSH has a server key rotation feature. Transparently change your server keys without requiring your users re-verify host keys.

It's documented in SSH Mastery, by the way. mwl.io/nonfiction/tools#ssh

Not that this has anything to do with github, of course. Pure coincidence.

in reply to philipp

@eingemaischt SSH Certificate Authorities. Same method used by Google and Yahoo. Also in the book.


I just wanna be an embedded dev so I can port Android Open Source to surplus Google Glass and mount it inside a pair of ski goggles

I am a simple man



Web of Trust For Content is going around, and uh, that feels way more fragile than the other WoT
in reply to silverwizard

yeah, that's a nonstarter on like every level.

I would really love digital signatures to be on more fragments of content, like I feel like the possibility of digital signatures is severely underused. But the truth value isn't something we can ever fix technically.

in reply to серафими многоꙮчитїи

@серафими многоꙮчитїи I dunno - I don't own enough technology that I trust to own a key longterm

I don't have a device I feel like my key wont be exfiltrated from, and I am terrible at keeping reliable hardware tokens.


in reply to vxo

here for the repost. wondering what it would look like if the connector sizes were proportional?


Are you really playing RPGs properly if one of your players is concerned about going on an adventure but says "I wouldn't have my home bees"

silverwizard reshared this.


"ai"

Expanding on my own thoughts a bit here, there used to be limited epistemological value to even search engines that returned disinfo and misinfo. In particular, they generally would still deliver reliable statements of the form "x claims to believe y," which has some value even if y is false.

LL.s give up even that limited epistemology by disconnecting factual claims from their sources.

mathstodon.xyz/@cgranade/11006…


100%. I'm also struck by the entirely strange epistemology of "we see search as a way to discover information, not as a truth provider." Like..... what is the epistemological value of information that *isn't* true, or at least probably approximately true?

silverwizard reshared this.

in reply to Cassandra Granade

"ai"

If I ask a search engine how old the earth is, and I get a link to Answers in Genesis saying "6,000 years," then I've learned a true and correct fact: namely, that AiG claims to believe that the Earth is 6,000 years old.

That's in essence the promise that search engines make when you ask factual questions, that they will deliver you relevant evidence to support "x claims to believe y." Thus, I always learn something from searching even if it's of limited utility.

If I ask an LLM how old the Earth is and get back 6,000 years, I've learned jack shit. In that hypothetical, it just spat back nonsense without providing me any information at all. That also means that I don't learn anything when an LLM is correct, either — there's no distinction I can use to separate those two cases, making LLMs useless to any sort of epistemology.

silverwizard reshared this.

in reply to Cassandra Granade

"ai"
On its own, that's fine! Plenty of useful things aren't useful in an epistemological sense. It does make it very weird to see LLMs used as a source of truth, though, even if a only as a secondary or tertiary one.
in reply to Cassandra Granade

"ai"

because it's making statements in a vacuum and detached from any sources it's also a classic epistemic paradox.

Like Socrates's Meno, unless you know the answer to a question already (or perhaps you are a convinced Bayesian and firmly trust your credences), you cannot recognize a correct answer with 100% certainty based on what stochastic model spits out



The biggest problem with Star Trek Strange New Worlds is that I hear it's really good - but it's related to Discovery and therefore I can't believe it

silverwizard reshared this.


Introducing my new organic, all-natural intelligence search engine: Mark.

It's just this guy Mark. He'll do his best to answer your question, and he'll make up a reasonable-sounding answer real quick if he doesn't have one.

How will we scale? We'll hire more guys named Mark as demand ramps up.

reshared this

in reply to Darius Kazemi

My university sponsored that phone service until 2013. I hoped it would last forever, but clearly missed the mark.
pittnews.com/article/62675/new…


The hardest part of being the only Security person in my org is that it's really hard to document "this is common knowledge in my field, but arcane knowledge outside it", because everyone is really worried that they don't understand the context, but anyone in the intended audience would understand it without the documentation.


I am slowly going mad

I am trying to report to Simplii Financial that they don't have SSL on simplii.com and they just keep telling me that they have SSL on www.simplii.com

This is one of those "type it into SSL labs and see what pops out situations", incredibly boring, and it just breaks my HSTS and it's annoying and bad.

Their support team needs a screenshot of my browser not connecting , and a version number of a browser and the model number of my computer, and it needs to be running Windows or Mac in order to report this. But they did finally send me to BugCrowd.

BugCrowd tells me that it's a false positive, and that this means SSL is working fine ssllabs.com/ssltest/analyze.ht…

I am definitely moving my money, but also - is BugCrowd usually this dumb?! Is there anywhere where you can report a (admittedly incredibly minor) security issue to a Canadian bank where someone who knows what SSL is will read it?

in reply to silverwizard

This is their proof that the bareword domain has SSL on it. Whicjh uh, I am glad BugCrowd hires the best.

in reply to silverwizard Quincy reshared this.

Day 5 of trying to report a vulnerability to a bank >.<




The theme of this week has been
"damn we made a mistake 2 years ago, I guess we can't ever fix it"

and then me stubbornly deciding to force people to fix it

fun fact: the security team can't get a ticket in the sprint. but they can get a high priority issue to override sprint priorities as part of an incident postmortem.

in reply to silverwizard

A couple years ago I was delighted to be in “code freeze” because the exception process was faster than the normal workflow, so we pushed more code than ever. 😀


‎silverwizard‎: One of my skills is giving the *wrong* right answer
becky‎: It is a skill
‎‎becky‎: I don't know how you do it so effectively so often
‎‎EmeraldMagus‎: That is in fact the category this answer falls under...


Fun fact - working as a tech worker for Meta is very rarely ethical
in reply to silverwizard

@silverwizard Microsoft went a step further and made sure of it: dair-community.social/@emilymb…


MSFT lays off its responsible AI team

The thing that strikes me most about this story from @zoeschiffer and @caseynewton is the way in which the MSFT execs describe the urgency to move "AI models into the hands of customers"

platformer.news/p/microsoft-ju…

>>


in reply to Hypolite Petovan

@Hypolite Petovan I was mostly thinking of people telling me to me empathetic to laid off Meta employees

But yeah - Microsoft and Google employees also don't seem very ethical.

And at least Microsoft seems to be doing it because they outsource all their AI (hopefully they are thinking this) and because they have decided it's layoff season. But at least they aren't Google who fired their best AI ethics person because she was like "Maybe we should not burn the environment for bad AI?"


silverwizard reshared this.


I was woken up today by a phone call letting me know that due purely to budgeting changes I’ve been let go from my full-time contracting position I just started in January after being laid off in September.

I am a former #Apple #engineer with 15 years of experience making apps for #iOS and #macOS. I also have experience with #Ruby. Available immediately for full time or contract work.

Please boost. #Swift #RubyOnRails



I wish the RSS Aggregator Without Delusions of Grandeur had a less embarrassing name so I could recommend it to more people
in reply to Spencer

@Spencer The thing is, it's just a really good way to make a stupid simple HTML page out of an RSS feeds group and get a very boring RSS


I just realized that the thing I want most in terms of clothes is "bondage pants designed by Rob Liefeld"
This entry was edited (1 year ago)
in reply to silverwizard

Gimme lots of loops and hooks and dozens of little pouches and pockets!


I am "whenever I see What's New I add With Phil and Dixie" years old

silverwizard reshared this.


Tbh if you were against the #NFT and #blockchain crap because of its energy requirements, you should be against #ChatGPT and most other currently hyped #AI crap.
If you aren't, I'm really curious what you're doing that you think makes the energy expenditure worthwhile.

reshared this

in reply to Csepp 🌢

Along that line, here's a link I shared earlier: limited.systems/articles/clima…
in reply to Csepp 🌢

While I'm not going to argue that AI is great for the environment, the wastefulness of cryptocurrency is next level.

Training Stable Diffusion v2.1 created 15,000 kg of CO2 equivalent, which is the same as roughly 34 Bitcoin transactions. Bitcoin chews through 3.4 million Stable Diffusion trainings in a year.

I run SD at home on my gaming PC and it uses the same electricity as gaming. So after paying that up front cost, it's about as bad as PC gaming.

reshared this

in reply to Urusan alcinnz reshared this.

@urusan Bitcoin did at least promise (but so far failed) to decentralize finance. I haven't seen a decent use case for AI art that justified its energy costs.

And yeah, I'm holding PC gaming and other forms of digital art to the same standard. Games should have a tiny fraction of their system requirements. Same goes for 3D animation.

in reply to Csepp 🌢 alcinnz reshared this.

@urusan The one saving grace for 3D and AI art is that once rendered it's just an image/video, but you don't need either an AI or raytracing to create aesthetic art.
in reply to Csepp 🌢

@urusan That said, I am excited for NeRFs! That should help bring filmmaking within reach of theatre groups! And save some trips to get good scenery.

And we've gotten the performance down to reasonable levels on them...

in reply to alcinnz

@alcinnz @urusan I already dread trying to look that up. How do people choose the least searchable names for their projects...
in reply to Csepp 🌢

@urusan Then try "neural radiance fields"!

Here's the link: matthewtancik.com/nerf
And NeRVs get really interesting: pratulsrinivasan.github.io/ner…

in reply to alcinnz alcinnz reshared this.

@alcinnz @urusan Seems more useful than most other image generation stuff I've seen, but to be honest I think we're gonna have to scale our expectations back of what art should look like if we want to solve the climate crisis.

Which, when put into perspective, is not a huge sacrifice compared to other changes we should be making.

in reply to Csepp 🌢

@urusan More useful & cheaper, make that!

I think there's a middle at which to meet people at, & NeRF/NeRVs could be a low-cost means of finding that middle!

Personally though I'd be happy with no visuals, I'm loving those stories!

in reply to Csepp 🌢

@cnx The framing on HN is pretty silly, but the link that it eventually gets down to (petals.ml/) is really interesting.

It's essentially a Folding @ Home or Bittorent style project for running and (even more notably) training AI.

The project seems to be specific to a specific family of LLMs (BLOOM), but it's a very interesting concept for decentralizing training in particular, without relying on powerful hardware owned by centralized parties.

in reply to Urusan

@urusan @cnx What bothers me is not whether it's interesting, but that no one seems to be asking if it's necessary.