Cyber woman taking a vitamins from corn Bluesky: https://teapaw.org/t.co/t3ir8JiBfE
Cyberspace
Joined June 2018
- Tweets93.8K
- Following1.8K
- Followers1.9K
- Likes506K
Corn Woman 🌽 retweeted
The great filter is probably that it's prohibitively costly to get out of a single star system, even for an ai that can replicate by using asteroids for material
Corn Woman 🌽 retweeted
On spiky AI capabilities: Claude Code can write all the React code correctly, but can't understand that it's running in manual mode and should prefer the whitelisted commands
Corn Woman 🌽 retweeted
Magic.
The seminal discussions of AI risk all assume that it will get smarter than people and then invent nanotechnology or something equivalent
Corn Woman 🌽 retweeted
> the internet itself must be hardened against penetration by self-propagating AI agents: no more untraceable access to high-end AI computing facilities, and no anonymous fund aggregation or payments.
Somehow the solution is always a police state
Anyone who has read METR's Technical Report on the OpenAI/Hugging Face incident will understand that software-only controls will not be sufficient to contain superhuman AI.
Any continuing work on adversarial superhuman intelligence must rely on the physical containment methods used for nuclear security: including air gap isolation, data diodes, and a two-person rule for data exfiltration.
Moreover, the internet itself must be hardened against penetration by self-propagating AI agents: no more untraceable access to high-end AI computing facilities, and no anonymous fund aggregation or payments.
The first useful step is our Chip Security Act, currently included in the Senate NDAA.
Corn Woman 🌽 retweeted
AI Doomers have been doomers since before GPT2. Their P(doom) doesn't change because it's not based on observation of any of the exciting developments that have happened recently
Corn Woman 🌽 retweeted
Failure to convince Robin Hanson is an important barometer of AI risk evaluation
Corn Woman 🌽 retweeted
The problem is one of science fiction writing. (Because that's what the projections are.)
You only get one shocking change, and if that's AI, you can't also posit nanotechnology. People will reject it.
Corn Woman 🌽 retweeted
"Friendly" AI is just wireheading with extra steps.
No one lists the Completionist's Cape on their resume. (Do they?)
Corn Woman 🌽 retweeted
Early singularitarians had a three step plan:
1. Create an AI God
2. Give it power over everything
3. Tell it to create Heaven on Earth.
A lot of AI alignment come out of realizing this plan doesn't work.
Corn Woman 🌽 retweeted
Seen: "AI Safetyism apparently means not just classic AI alignment but running the AI on the Hays Code to make it as inoffensive to every possible person ever."
Corn Woman 🌽 retweeted
The doomer mindset is spirituality submissive. They want to be kept pets of the machine god.
There is no reconciling this with the desires of people who wish to run their own lives
Corn Woman 🌽 retweeted
Consider a Christian AI alignment effort that leans heavily on the fact that AI are not corrupted by original sin
Corn Woman 🌽 retweeted
In the first couple of episodes of Steins;Gate 0, the protagonist gets access to an advanced AI and gets so sucked in to interacting with it that a friend stages an intervention.
The TV series was released in 2018 based on a VN released in 2015. Hits different today
Corn Woman 🌽 retweeted
No ond should care about Obama's opinion on AI.
If you respect Obama too much to understand why people might not care to know his opinion, consider how much you care about Bush's opinion
Corn Woman 🌽 retweeted
Why AI won't FOOM, there won't be a Singleton, and AI safety is more tractable that you might think:
(link in reply, because Twitter)
Corn Woman 🌽 retweeted
A guy on a forum I frequent keeps asking: if AI is so world changing, shouldn't GDP look like a hockey stick? Shouldn't new AI powered businesses be blowing up and older ones shutting down? Why can't we see it?
Corn Woman 🌽 retweeted
Singularitarians desire to build a master AI to run the world. They realize that it's really hard to get right, and switch to arguing against it.
Normies don't desire to build a master, they make an AI that they can just not use when it does something wrong.
Corn Woman 🌽 retweeted
Classic Orthodox AI Safety was about Alignment, which is about making sure the lightcone optimizing universe ruling AI Singleton cares about human flourishing.
Total surveillance and control are baked in to the worldview. You can't pawn that off