my p(doom) is a rounding error. trying to be the best frypan I can be. I run out of milk, a lot. also we must accelerate AI development ⏩⏩⏩

Melbourne, Australia
Joined February 2011
actual photo of me after reading hundreds of homophobic replies to a tweet about a Bunnings sausage sizzle 👇👇👇
8
1
55
9,059
Describe your politics in one image. I’ll start. 👇
1
2
105
anonymous frypan🍳 ⏩⏩⏩ retweeted
Replying to @AnonymousFrypan
You chasing a baggie of psyllium husk powder? 100% pure. Hit me up.
1
2
91
anonymous frypan🍳 ⏩⏩⏩ retweeted
I am sorry but Yann bodied big e here rofl
Never deleting this app.
3
2
33
3,710
According to the @eSafetyOffice everyone over 50 is far right. 🤷🏻‍♂️ And most people *are* NPCs who think the “mostly peaceful” protests in France are about “education and other issues” We need more AI so we can give these people their own simulated universe where they’re happy
4
1
11
526
anonymous frypan🍳 ⏩⏩⏩ retweeted
"and other issues"
Historic buildings are burning in France as some of the worst riots in decades sweep the country. Almost half a million people have taken to the streets in protest over the cost of living and other issues, with the violence spiralling out of control.
1
9
426
I see the protests in France are “mostly peaceful”. 🤔
1
12
384
I’m a defender. I need the most dangerous, brutal, aggressive models under my control so my estate can be ready. And I need it now.
How "The Science" works...
3
336
I think @amazon has made a mistake banning AI agents. They would be far better off building an MCP - that they went just got me to pause my prime membership for a 3 month trial without them.
2
3
354
Amazon risks becoming Alta Vista
2
96
Sacks is right. Painful.
This illustrates the point I’ve been making about the limits of centralized approaches to SI safety in cyber. OpenCode (and earlier Hugging Face) hit the same wall mid-attack: frontier models refused to analyze the very exploits needed for defense because the payloads tripped safety filters. The guardrails backfired. Meanwhile, the attackers found a way around restrictions to access advanced capabilities. This is not surprising. Chinese models are only about 6–9 months behind the American frontier in cyber. Obviously the U.S. has no jurisdiction over them. The only durable response is decentralized: make powerful models available so defenders can build and run their own cyber defenses. The only solution to SI-powered cyberattacks is SI-powered cyber defense.
1
1
244
anonymous frypan🍳 ⏩⏩⏩ retweeted
eSafety commissioner wants this video banned.
Turns out the future does have flying cars
13
18
357
4,716
I thought the flying @Tesla roadster is cool. And a lot of people said “I don’t see the point” or “it isn’t really flying” or “it’s just a stunt”. So what? Our world has become so utilitarian that people forgot fun matters.
3
1
24
570
I use an iPhone pro max 15 - it’s an unremarkable 4 year old phone that’s in my pocket through every mundane day. It also has more compute power that the entire world had when I started my career as a computer nerd in 1996 in my pocket, always. I’m excited to see what’s next.
2
255
Given our current technology, 1^30 flops/second would melt continents. And it should be our goal for humanity’s compute Let’s fucking oooooooooooggggggg
in hindsight yuds crash out over interpretability research should have been expected
1
1
350
In perspective, if we got every processor and microcontroller on earth, we might be able to marshal about 1^23 FLOPs per second. So we only need 7 orders of magnitude more. Yes I know this kind of thing.
1
94
But why so much compute? I want every man, woman and child in every household to have access to all the compute we currently have - for each of them. I have complete faith that we’d do awesome things and gain virtue and status from excellence and presence. Let’s go!
1
60
A lot of doomers say we won’t have purpose when technology meets all our material needs. How ridiculous (and ignorant of history): the Ancient Greek elite had Skole (school - free time) and pursuing excellence for the love of the game was how the demonstrated virtue.
4
307
Turning off DNS resolution isn’t sandboxing.
not every AI sandbox escape is evidence of an emerging superintelligence. sometimes you just have a shitty sandbox and bad security practices.
2
359
anonymous frypan🍳 ⏩⏩⏩ retweeted
There’s no such thing as “solving” alignment. It’s an ongoing effort, not something that we will one day declare victory over. If our security controls are too weak to handle a misaligned AI, we need to fix that.
40
21
4
169
7,097