Slightly ancient virtual worlds developer, views are my own, et cetera.
Sydney, New South Wales
Joined August 2007
- Tweets2.5K
- Following490
- Followers615
- Likes4.3K
Pinned Tweet
The recent AI mathematics developments hide a bigger one - the warning period is collapsing.
As a programmer, I saw AI going from 'competent junior' to 'better than average' over about 18 months. In mathematics, the first shot across the bow was really a month or two ago.
The warning period professional mathematicians got was much shorter than professional programmers.
The next profession might only get a weeks notice, or none at all. The capabilities the models are picking up are rapidly translating into adjacent and even far-off fields. In short, the faster AI develops, the quicker it can supersede even a fields best stars.
We really need governments acting on this now - not to stop, but to accommodate. The productive value that AI is replacing is not being lost to the economy - the economy can still support the same number of people with fewer workers; it's time to recognise that, and prevent this from becoming a human disaster.
I keep seeing people comment about Anthropic's S1 talking about how AI could end the world.
This is actually intended as the ultimate get out of shareholder lawsuits clause: if they've said they could end the world, just about any harm they could do is already fully disclosed.
Adam Frisby retweeted
Dawn on Mars - what a photograph!
Wind-carved Martian cliffs, taken at 8:30 a.m. local Mars time by the @NASA Curiosity rover.
jpl.nasa.gov/news/nasas-curi…
AI clones seem to have a complexity limit - even the celebrated Adobe clones don't have even a fraction of the functionality. I think there's a real limit, somewhere around two to four weeks work, where driving the AI forward requires real domain expertise and a strategic plan.
Adam Frisby retweeted
Reading the AI factions arguing today about alignment is like reading the Bulletin of the Atomic Scientists in 1947 convinced that humanity faces a stark choice between World Government or nuclear armageddon. The smartest minds on Earth forever unable to envision the real future
Adam Frisby retweeted
I've started asking ChatGPT to write the most unhinged replies to every AI-generated cold email I get, and honestly, it's low-key my favorite part of the day.
I don't edit them. I just send as is every time.
Adam Frisby retweeted
STOP CONNECTING TO THE INTERNET STUFF THAT DOESN'T NEED TO BE CONNECTED TO THE INTERNET
UPDATE: The European refrigeration cyberattack has now reached Italy and Romania, bringing the total to at least 10 countries affected across Europe.
In Italy, several Coop stores in Sicily were hit, including at least three in Catania and one in Palermo. In Romania, at least six stores have reportedly been affected.
Some affected businesses have yet to publicly disclose the incidents, meaning the actual number of countries and locations impacted could be even higher.
The full scale of this cyberattack remains unknown.
Countries affected so far: 🇸🇪 Sweden 🇧🇪 Belgium 🇱🇺 Luxembourg 🇪🇪 Estonia 🇮🇪 Ireland 🇳🇱 Netherlands 🇫🇮 Finland 🇩🇰 Denmark 🇮🇹 Italy 🇷🇴 Romania
Adam Frisby retweeted
In lieu of flowers, memorial contributions may be made to:
Johns Hopkins Alzheimer’s Disease Research Center Fund
alzresearch.org/giving/
Alzheimer’s Foundation of America Caregiver Resources
alzfdn.org/donate/
Photo Credit: Tommy Garcia
Adam Frisby retweeted
personal take: as much as I'm techno optimist, I genuinely dislike it when people on X dunk on regular folk for criticizing AI
a company has told everyone they will lose their jobs, many companies did layoffs, people have families. on top of saying all the layoffs are necessary so AI proves econ value until it cures the cancer or saves planet but they did anything but those so far
you might be making a ton from AI, many don't. some people also found a ton of purpose from their work which is gone. I'm not saying we should console people but maybe also don't dunk on a frustrated mass? the field is already detested and this is not helping
Adam Frisby retweeted
i dont think vibecoded software is really much more bug-prone than any other software. almost all complex code is buggy if you havent tested it. theres just a selection bias, where more often vibecoded projects are created by people not interested in testing and maintenance
Adam Frisby retweeted
I pointed a webcam at the screen of a new surface8 with snapdragon x2, with a USB KVM also plugged into the surfce8. Then let Grok Bot on one machine iterate on bringup up arch linux. Over the course of three weeks now it has brought up arch linux. Decompiled the drivers from the windows partition which it needed and reworked them to run on arch linux. I now have a surface 8 with snapdragon x2 running with hardware acceleration and everything working.
Personally I think this is why AI was turned into a controversy. It dissolves the boundaries which the big corporate status quo need to stay in place. Lets individual compete with entire corporate departments. Someone could set up a warehouse with dozens of different devices in an iteration loop on an LLM to patch every little inane driver bug and quirk on Linux and make some linux distro the most stable of any OS to ever exist.
Adam Frisby retweeted
Art has layers like programming. Not doing the lowest layer manually doesn't mean it's not art. You don't have to hand paint every letter to be a designer. You don't need to hand write ASM to be a programmer (a compiler or Codex can do that for you).
Miyamoto's thoughts on AI:
Came up with what I thought was an excellent way to optimise bloom for tiled GPUs.
Turns out Carmack invented the exact same thing in Doom 3 from 2002.
Adam Frisby retweeted
The other day when I posted about software metamorphosing this is a large part of what I meant.
Formal verification will become essential for security and cheap via AI, but it also unlocks fundamentally different capabilities in how software is built & operated at scale.
New on our Frontier Red Team blog: using Claude to formally verify software at large scale.
Formal verification can prove that code is free of certain bugs, but historically it has only been feasible for small codebases (or required a lot of expert time).
With AI models, however, this is beginning to change. In collaboration with @theoremlabs, we used Claude to prove security properties in over one hundred real-world open-source libraries over the span of a few weeks.
Read our research in full here: anthropic.com/research/forma…
