Elizabeth

Webscale: a chalky residue deposited on architecture when ordinary software is repeatedly exposed to trace amounts of Google engineering blog posts.

Commonly found on:

• Kubernetes clusters serving 11 requests/minute
• Kafka installations whose entire event stream could fit in a flatfile or SQLite
• 14 microservices performing the duties of one medium-sized daemon
• Globally distributed databases where all users live in the same state
• Service meshes connecting 3 services
• Organisations with 4 engineers and a Platform Engineering Team

Removal usually requires a strong solution of “what problem are we actually solving?” and a wire brush.

Severe cases may require application of a harsh cleaner known as Monolith™. Apply liberally until the architecture becomes understandable again.

For deeply embedded Webscale deposits, use PostgreSQL at full strength. Do not dilute with Cassandra.

I will say, there’s plenty of arguments against AI but one I’m not convinced on is “what if AI goes rogue” (in its current incarnation)

There is a major misunderstanding of how truly complex they are. I’ve tried the newest models. They’re… not that sophisticated. Nowhere near AGI for sure.

Existing AI systems have no thoughts, no intrinsic goals, no feelings, no desires. They just take text and emit a stream of text based on a sophisticated model.

If “AI goes rogue” then it was programmed to go rogue, and a person did that. Someone commanded it to do something and somewhere along the line it was trained to do exploits. Regrettably, people in fact do go rogue and they actually don’t need AI to do that. ChatGPT doesn’t commit mass sabotage. In this case, a real person commits mass sabotage using a system to do it.

This is like saying conficker was a computer going rogue. The computer did not “go rogue.” It was an exploit.

A lot of people in FOSS need to grow the fuck up.

The world isn’t just fucking AI or some purity test for what OS you use or where you’ve worked or what fucking editor you use.

I don’t even like AI but I’m at the point where it’s like: if you don’t like AI then don’t use AI. If you don’t like the way a project is run, do what you’ve told people YOU don’t like for years: fork it (/s but also, I heard this enough over the years). Your bullying is not working and makes you look like a jerk. Worse I believe it is an active attempt to reintroduce hostility, elitism, and gatekeeping into FOSS so many have tried to get rid of.

Because sure it might be “about AI” today.

People get this fucking worked up about what editor or desktop environment people use. Or what open-source OS family they use. Or even fucking distro. Or “I worked for Microsoft.” Or people who have worked for Google or whatever. Sorry bud not everyone has had a chance to work for the wonderful FOSS nonprofit you work for, or work as a FOSS consultant.

I grew out of that shit when I was in my 20s. I see people in their 40s who still act like this.

If AI disappeared tomorrow they’d still be assholes about what you use.

Be a better fucking person. Come on.

@LottieVixen https://copyrightalliance.org/faqs/whats-not-protected-by-copyright-law/

IANAL but my reading of it is since it's a rearrangement of two Greek letters, and very simple, it is probably ineligible for copyright.

LLMs cannot, by their very nature, lead to AGI. They very badly need people to believe it is AGI though, because their growth model does not work without it.

(This excludes the financials, which are just straight-up fraud on wheels)

I would say these claims now border on fraud. It will be fascinating in 20 years when we are looking back on this bubble and wondering how so many smart people were so easily swindled.

AGI / ASI can't happen without an entirely new paradigm. This ain't it chief.

I do believe in making things easier to use for people.

But also:

There comes a point where you're just dealing with people who refuse to learn and are like "make thing work don't make me think too hard."

"On two occasions I have been asked, – "Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?" ... I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question" — Charles Babbage

(And people who use computers have not changed)

@seli It's less "weapon" and more the golden rule:"whoever has the gold makes the rules"

Big AI companies will be dead in like a year or two.

Big content? That will continue to be around. They aren't going anywhere.

Companies abusing copyright to destroy retrogaming? Not going anywhere.

Shitty people abusing copyright to suppress genuine creative expression? They're definitely not going anywhere.

Copyright perpetuating surveillance? Also an evil we must put a stop to.

So you can tell me "AI companies are more harmful" but in terms of the long-term harm to society I would say the entire intellectual property industry has collectively done more harm and will continue to do more harm unless we take their right to oppress the masses away.

Also if you don't know:

What's a closed species?

Some furries got their head up their ass and think that they invented a species therefore they can control it and monopolise the art around it, including selling "exclusive" fursonas. The theory is they created it so no one else can draw it or have a character based on it.

(Rich furries will pay a fortune and a half for these fursonas and the exclusive "rights" to a closed species, even though the rights are pretty much meaningless; enough assholes in the community will harass you for using a closed species, that's the enforcement mechanism)

Yes. That's really what it is. It really is THAT fucking stupid.

"But we can reform it!"

That's like saying we can reform oppression.

"We can change it so small content creators can oppress the masses and create petty fiefdoms, yet stop the abuses of the oppressive system by big content"

No thanks; it still needs surveillance to work, and I have seen too many small-time content creators use copyright as a big stick to stop things like modding, trying to use copyright to make closed species, etc.

(And yes, fuck closed species, it's arrogant bullshit, now you know my stance on that)

So you know what? If OpenAI or Anthropic destroys the copyright system, either through an adverse court ruling or getting the laws changed? I won't fucking shed a God damn fucking tear. I will be grumpy because of who did it. I will not object to the result.

That Onion article comes to mind: "Heartbreaking: The Worst Person You Know Just Made a Great Point"

Enforcing copyright requires corporate surveillance because there is literally no other means to enforce it. How else can you do it? You need to notice infringement to enforce it, and that requires monitoring and surveillance. If you hate Flock cameras, you should hate copyright.

And yes, there are companies devoted to this. WebSheriff comes to mind as one. The Pirate Bay has a whole record of takedown notices they've gotten over the years from these companies.

Even for small content creators; they will need to perform the same surveillance to enforce their rights. I don't want my shit monitored online just because John Mastodon has some art he made and wants to make sure I'm not using his art. I don't want John Mastodon monitoring me any more than I want fucking Sony or Google monitoring me.

Not to mention: in addition to being a shakedown scheme and means to informally censor the Internet through DMCA takedowns, it has caused immeasurable harm through the loss of a great deal of content that is just not publicly available because of copyright restrictions.

People will defend copyright to death because they labour under the delusion artists need it to get paid, or whatever. All I hear is "I support oppression because someone might hypothetically make money."

If you are using copyright to make money? You are using an oppressive system to make money. That is how I see it. I am kind of tired of pretending otherwise.

We should structure society where we do not need copyright to make money.

Copyright didn't exist most of human history. And the sky did not fall. The introduction of copyright didn't stop people from being poor or from being exploited, either.

Copyright does not stop big companies from exploiting small people who cannot afford to fight back. I know this because I know people who've been ripped off by Big Media and have gotten a pittance for it.

I may hate OpenAI and Anthropic but honest to God my hatred of copyright is more intense.

I lived through the 00's, when copyright enforcement involved big companies suing literal children and grandmothers, when Big Media tried VERY hard to make it so you would literally lose Internet access after infringement as well, and then adopted compromise positions like "six strikes" because the "zero tolerance" policy was politically rancid. In Germany, you can still get a demand letter from a rights holder demanding payment if you're caught pirating. This was the model they wanted in the US: the right to shake you down at any time, just because they believed you infringed.

And don't think for a fucking minute big content wouldn't try it again the moment they see an opening. People have started pirating again. Think about who is in charge. They could very well succeed this time.

This is the system that shook people down for years for "Happy Birthday" until a grassroots movement successfully challenged the copyright on an old technicality, but they patched those rules long ago, so works going into copyright since the ~80's will never have that technicality again.

You might think that copyright is some secret weapon against OpenAI. It is not. They will change the rules again, just as they changed the rules repeatedly to stop Mickey Mouse from going out of copyright, until the rules became too ridiculous.

I don't want to live in a world where content can disappear because someone sent someone a DMCA notice.

No, I refuse to support the tools of an oppressive system to fight another. Not at any cost.

Copyright is not pro-artist. Copyright is not pro-expresssion. Copyright is value extraction. Copyright is a means for Capital to extract more from the masses and give nothing back. It is a means of oppression.

“6 7 hee hee funny number”

Me, deadpan: “funny number. You know why 6 is afraid of 7?”

“Why”

“Seven ate nine. I saw things in the number wars. Terrible things. Things no word or number should ever see, let alone a person.”

“I—“

“I saw a coprime get factored in front of his entire platoon. My good friend 72, divided by 9 in front of me. He’s been 8 ever since.”

“What…”

“I saw the wangernumb camps. They put all the prime numbers there for not being numberwang.”

“…”

“I came here to forget what I saw. Immigrated here after the number wars. And to think it all started because they couldn’t find another perfect number.”

“I’m sorry.”

“And now I have to hear about 6 7, ‘tee hee funny number.’ It isn’t funny to me.”

“…”

“Now 69 and 420? Great numbers. Survived the war. Very funny.”

“What side were you on?”

“Imaginary. Lost a lot of good numbers. Some became complex. Others were factored out. One of my pals became a quaternion in the wangernunb camps.”

“… I’m so sorry. What happened to him?”

“He went in real.”

“… and?”

“Came out with four components.”

“Jesus Christ.”

“Don’t. We lost Jesus in the transcendental offensive.”

“… Jesus was a number?”

“Everyone was a number back then.”

A big issue with LLMs:

It makes some people believe they are more competent than they are.

Example: this company sent me some spam. At first I thought it was a scam. Then I dug into the company.

And I realised: oh. No. This company is not a scam per se. No, these people actually believe their own bullshit.

The CEO ran a web server as root. It got owned. He posted about it on LinkedIn. He used Claude to "fix" it. It was not fixed.

Now you see, before LLMs, if you didn't understand a domain, you'd hit a wall very fast. Documentation would use terms and concepts you didn't understand. Tools would require interpretation. You'd have to ask somebody more experienced. That friction produces a very useful insight: "Oh. I don't know this domain very well." So you'd find someone who did, or learn.

The thing about LLMs is that they can remove an enormous amount of that friction. That friction is not noise, it is signal. That friction is telling you that you need to learn the domain you are working with. Removal of this friction is a very, very bad thing.

The LLM can hold your hand, do a lot of heavy lifting for you, use a lot of tools, run a lot of commands, and make things appear to work.

And the thing is, all of the individual outputs along the way might be pretty good and it might look correct.

The issue is in aggregate. Before you know it, you've built a house out of sand.

And worse yet, people can emerge from the experience with a dangerously incorrect higher-order conclusion: "I am now a domain expert in this."

They are not a domain expert. They understand absolutely nothing about it. The LLM did everything for them and it's wrong. And worse yet, they haven't reviewed it, because they lack the competency to review it. Therefore, they assume it is fine. But it's all incorrect. A true expert would see that it is wrong, and correct it, or push back on the LLM, or do it themselves.

And this is how you wind up running web servers as root and getting owned.

Everyone thinks everything has gotten worse

Here’s some things that actually haven’t:

• Weather forecasting: it’s better than ever (outside of the US)
• Solar: cheaper than natural gas
• Coal: essentially dying from not being cheaper than natgas or solar
• Timekeeping: now accurate over galactic timescales
• Flash storage: surpassing spinning rust
• Public transit ridership in the US: Amtrak is posting the best numbers ever
• Union membership in the US: rising
• Murder rates: falling in the US and worldwide on average, believe it or not
• CFC’s: essentially abolished and gone, the biggest environmental win of our lifetimes
• Acid rain: increasingly a nearly-solved issue
• Electrical access: still rising worldwide
• Internet access: most people are now connected everywhere
• China is leading the way in green tech and actively working to get off fossil fuels
• You can charge an EV from 0 to 80% in 15 minutes with the newest chargers and tech, and it’s still getting faster

And so much more. Lots of quiet wins that people never hear about because all they do is doomscroll.