Home Creators Posts Import Register Favorites Logout

Downloads

Content

We discuss Marissa Mayer’s innovations in building digital wives for wives, add another few hundred thousand to the Kirk fund, and reflect on Lula’s victory. Then, Eva Galperin, director of cybersecurity at the Electronic Frontier Foundation joins Riley to discuss why every website in the world is suddenly being infinitely hacked all the time by everyone.

RILEY ALERT Check out No Gods, No Mayors here!

HUSSEIN ALERT Check out 10k Posts here!

MILO ALERT Check out Milo's tour dates here:https://www.miloedwards.co.uk/liveshows

NATE ALERT Nate's band Second Homes has just released their debut album and you can stream it for free here!

Files

Comments

etienne

- Big Brother, Jar Jar Well 1984

cole d

That post from stacy sumpter is a quote from Django Unchained, that Christoph Waltz says when they are leaving Leo's character's plantation

James Rule

AI supporters write articles admitting to things that CIA torturers couldn't drag out of me.

Angel-of-Deadlifts

The thing I'm worried about is Lula is quite old, what happens when he dies? Like the right has Bolsonaro's clone son, but Lula doesn't have a clone son! Could Brazil elect Erika Hilton as president?? She's been the most voted for deputy!!

Hovis

As someone who already lost one career to AI I am just loving seeing it show up in my current job as the last line of defence against a threat that it helped create like the Terminator in the psychiatric hospital.

YungCereza

"AI Wife for your wife"

Carley They/Themsen

Marisa Myer feels like a Kelly Slaughter Philosophy Tube bit gone rogue

etienne

us old Tumblr denizes remember Marissa Mayer as the person who bought Blue Hellsite for more than a billion dollar and pawned it off a few years later for pocket change

Frankie

Considering Marissa Mayer is also the woman who brought us 'Tumblr should be the next PDF', digital wives for your wives isn't the most bizarre thing this woman has ever come up with. With any luck, she'll come up with even stranger ideas in the future

Hovis

Apparently the phrase is triggering for a certain kind of racist. I don't think the guy in Django took it well either.

Ester Daniel Ytterbrink

I would like to know what Nicole Perlroth thinks about this. I found the book “This Is How They Tell Me the World Ends: The Cyberweapons Arms Race” very interesting and can’t help wondering how many security holes that are left in place will now be found and patched.

Huronbikes

what the fuck is this interview. The PHP vuln was found by a fuzzer 6 years ago, vulnerabilities found by AI are often crap, AI cannot patch vulns reliably, and the centrist view of "will we just need skilled devs using AI" is bullshit because it's all slop.

Huronbikes

The larger problem is no one wants to bother with maintenance and you already know this because that is one of the core reasons the future is becoming garbage.

etienne

yeah I haven't listened to the episode yet but someone pointed out that the vulnerability exists on paper but no exploit is in the wild yet and I was like huh

Paul Natsuo Kishimoto

She named her kids MacAllister Mayer and Marielle Mayer? Bonkers

Nemo

I can't believe this one girlboss mom invented the TV remote from Click

Keith

It is very cute that the guest thinks it is reasonable to expect that tech companies will provide for their customers' internet security. I am inclined to believe that she has never worked for a corporation.

Paul Natsuo Kishimoto

Reliable AI opponents I follow on Mastodon have been grumbling for months about EFF both-sidesing stuff.

Mike Kangal

The Boys From Brazilfuture

Al

I had successfully forgot that movie exists until now. Lol

Al

Unsexy risks

Nemo

Those who do not learn from Click (2006) are doomed to repeat it

kiwi

That lady having the AI handle her life for her is extremely close to that "computer, do" animation. I hope she made sure to add masterpiece at the end of her prompts

Josephine

I know that capital has to constantly strive for new frontiers and new revolutionary technologies to exploit so will nessicarly big its stuff up but also what the fuck does that even mean?

Frankie

Your guess is as good as anyone else's - afaik no one every figured out what the fuck she was talking about

skedazzle

Old heads know there were warnings about having critical appliances connected to infrastructure using protocols designed so academics could talk to one another. Building essentially Dubai on the foundations of a rickety old west set will have consequences. And then a bunch of vulnerabilities were found after high profile hacking incidents. Because people in the know are aware of the rather frayed security situation they were a bit jittery and leaked the patch before it was complete. Many sites were taken down to do the update out of an abundance of caution. Ai companies jumped into the conversation because they can't have tech conversations not be about them, and took advantage of the fact that lots of people aren't talking about this honestly for a number of reasons. That is my understanding of what happened. If eff is saying differently 🤷

Nicholas Jarvis

Me: Dazzl, experience this dramatic irony for me. Dazzl: "Nooooooooooooo!" Me: *cracks open a beer* Ahhhhhh...

Josephine

Wait was she the whole "female presenting nipples" person?

etienne

naah the timeline doesn't match, she left Yahoo a year before the pornocalypse and the site was offloaded the year after that

Z

Riley, that type of bathing suit is called a sunga. I got them for all my groomsmen at my wedding in Brasil.

Vanessa P

it seems very easy for people who are trying to be reasonable to try to draw this separation where they're like oh well there's a big difference between slapping together a pile of shit that you don't understand and can't change or diagnose by asking claude to "generate app make no mistakes" versus like feeding in some code you wrote yourself to ask it to check it over for vulnerabilities, and the information environment around LLMs is so murky that even skeptics can trip up and concede too much to the AI people. listening to this though I was like "wait HOW is the chatbot finding complex vulnerabilities. WHAT IS THE TECHNICAL MECHANISM HERE." they don't think! they process text! I think there is reason to be extremely fucking skeptical that a statistical text generator can FIND INCREDIBLY OBSCURE, PREVIOUSLY UNFORSEEN SECURITY FLAWS??? I grant there definitely is a risk that LLMs let script kiddies brute-force shit easier by just churning out stuff people have done before but I would like someone to explain to me precisely by what means the statistical text generator can process code in such a complex way as to produce an entirely novel security analysis???

Marley Jay

Great analysis by Hussein here; I’ll add that in the family context, AI is just the latest gadget that’s being offered as a quick fix for structural failures in our society that make parenting much harder than it needs to. ‘You say you need better pay and childcare and flexible work arrangements, so what about an AI that interacts with your kids for you?’

RizzInspector

It seems that parents may have lost the plot of Adam Sandler’s 2006 film Click

Cormac

Just to say, Nova pronounces Dien Bien Phu rather beautifully.

etienne

oh you never want to fire a Sumter or at a Fort Sumter when starting a war, either civil or culture

Rach

Well here's the thing, mathematicians + scientists have not yet figured out why models are able to generalise like this.. You're correct that they don't think, and yes, at their core, LLMs are statistical text generators, but there's also a lot of advancements over a simple statistical model like T9 predictive text. Mainly the transformer architecture that lets them process all the tokens simultaniously instead of one after the other, by giving each token a relationship with every other token. We know this is what makes it appear to capture context and semantic meaning even though it's still just predicting a next word, BUT, that's just *how* they work, researchers still don't have a full theory or proof for *why* they work. (Work as in function, not work as in be useful or reliable ofc) But once the transformer architecture was implemented the models went from generating garbage to comprehensable sentences, so it's obviously the key. As for finding vulnerabilities, there's nothing special there, it's just imitating chain of thought 'reasoning' exactly like it does for prose. If you go to a decent sized LLM, and write your own prose about an argument or opinion but hide a logical error in it, it's likely (although never guarenteed) to find it if you instruct it to. A code vulnerability is the basically mathematically the same as that once it's converted into tokens. (Also, this isn't me gassing them up, I'm a massive AI skeptic. But it's worth understanding that the AI industry is very skewed towards on 'we tried a bunch of stuff to see if it worked' rather than 'we knew this would work because of theory', and this is why they're so dangerous and volatile and harmful as companies)

mexicanhalloween .

you don't know anything about your daughter? that's not something you allow a newspaper article to be published about, that's something you conceal until you take it to your grave. terminal humiliation kink level of posting your Ls

tabula_rosa69

"Dazzle suggested a Venetian mask-making workshop that the family had liked" wait so it was just recommending them an activity that they had already done? is that -- is that something people need to outsource ????

George Toner

But doctor, my wife *is* A.I.-Pagliacci

tabula_rosa69

they had this shit in the 90's, we've known how it works for the better part of a century, stop scaring yourself over magic tricks that were old when you were a kid

Chiaretta ⚜

That is, verbatim, a quote from Dr. Schulz in Django

Em Gravesong

I believe she is saying the family enjoyed it after the AI suggested it, trying to imply this is a good application for the bot.

Em Gravesong

Water utilities are expected to be responsible for the safety of the water they provide. It is a perfectly reasonable expectation to hold; the fact corporations won't do it because of the cost is a problem, not some fundamental physical law we have no choice but to endure.

Keith

Water utilities are a public utility accountable to residents customers, not shareholders. Private corporations are not accountable to their customers. They could care less whether they live or die.

Em Gravesong

Yes. And thus we must capitulate to these artificial realities? You are demonstrating why cynicism is a false-virtue; sure, you can never be disappointed if you capitulate to all that which is rotten in society, and I understand the draw of avoiding this pain. But you also foreclose the possibility of positive change. Allow your expectations to be disappointed, and when the worst does not come to pass it is a pleasant surprise. And when your high hopes let down, it is not a crisis, but an unfortunate eventuality. Life will inevitably demand compromise, so demand the most and best, not just what is already on offer.

Keith

No, I am saying that private corporations are incapable of functioning as responsible members of society because their interests are at odds with it. I am saying the model has failed and the sooner we recognize this reality rather than placing naive faith in the delusion private corporations will 'come around,' the less damage will be done.

Em Gravesong

I have no idea where you got the idea anyone was advocating putting faith in corporations. The rhetoric in the interview seemed pretty well oriented around "something must be done, but the people in charge are refusing to, and that's bad." Must she start quoting Lenin? The EFF is notionally non-partisan. As much as I would like that everyone could extrapolate out the consequences of The Contradictions and behave accordingly, some people aren't there yet. It is not "centrist," "selling-out," or "delusional" to tailor one's rhetoric to not immediately send the average Lib running. The impulse of certain Leftists to beat people over the head with a manifesto is a great way to find Fellow Travellers, and by that exact same token a great way to make it impossible to organize with anyone trained to be reflexively skeptical of anyone left of Teddy Roosevelt. Discretion, valor, and all that.

Tsu

tl:dr because code is formulaic and predictable in a way human conversation isn't. Code is closer to math than speech and computers are good at math.

Tsu

Replying to let you know at least one person understood that reference.

Geoff Blanchard

I love getting more Nate! Nate’s great 👍

Maciej S

I groaned audibly when they said “ask AI to patch this vulnerability”, an absolute disgrace. The guest they invited believes in magic and is not a good person to explain how any of this works. Yes, I am aware of who they are. I am pointing out that "is highly visible cybersecurity professional" and "can communicate clearly without falling into cliches and oversimplifications" are not the same thing.

Maciej S

This interview is baffling, Gasperin opens by an absolutely fantastical (as in "this is a computer genie that grants wishes and does not need verification") description of LLM usefulness in IT security. "You should be able to use AI to write better code. You send your code back to AI and you say 'find the vulnerabilities in this code' and then you tell it 'patch the code' and it writes you code that doesn't have these vulnerabilities and the vision is that one day in the future we will be all writing code with no vulnerabilities that AI can find." And then later says "the reason why frontier labs are not doing much about the vulnerabilities because they believe that one day AI will be able to fix all vulnerabilities but for now they hand-wave the concerns away". You quite literally painted the same vision!