Derek Fine🛑

🛑
Anti-AI
Anti-Theist
Free Speech
Semi-Distributism
Look up to no one
Think critically

[Derek Fine🛑 @DFine2002]

February 22nd, 2025, at the end of the clip that's my stupid ass blocking OpenAI and getting hauled in a police car.

I thought this was going to do something. I thought this would lead to an explosion in peaceful, nonviolent copycats. Nonviolent, peaceful, civil disobedience. But no news station would broadcast the footage.

It caused nothing. We didn't keep the ball rolling. The ringleader at the time immediately pulled everything into side projects and "get famous quick" plans that went nowhere. The presentations stopped. The rate of posters being put up slowed to a crawl.

The ringleader wasted all our money on crap that failed, mainly food and beverages to "entice" people. I tried to stop him from wasting money but I could only go so far before being told to shut up or leave.

The ringleader began locking everyone in daily AI safety screaming debates instead of letting us put up posters and go to huge events to recruit potential new members, of course this wouldn't mean much because he refused to let us hand out anything except esoteric, absurd sounding nonsense and giant red squares with Stop AI on them.

I tried getting arrested at Palantir, The Forge, Anthropic and Google. I wanted to get us more cred and attention. I was told over and over that I was kicked out if I did such a thing, because we are focusing on OpenAI "until they stop". We couldn't even have a normal non-arrest protest anywhere else because, again, OpenAI was "the most likely to build ASI".

Yes, I wanted to get arrested and I didn't get arrested because Kirchner told me I was kicked out if I didn't want to do it anywhere else, even though everyone who gets arrested there gets a stay away order. No one would come back, and Kirchner forbid protests anywhere else, so the arrests dried up.

Only 1 red Stop AI shirt design, no white shirts, no gray shirts, no black shirts, NO PINS, NO NECKLACES, NO BRACELETS, NO TIES, NO KEYCHAINS.

I wanted to reach people starting them off discussing the water and art and cheating and basic normal AI problems, then over time get them to believe ASI was a serious threat. Kirchner refused any such thing.

"ONLY TALK ABOUT SUPERINTELLIGENCE ON THE FLYER. ONLY TALK ABOUT SUPERINTELLIGENCE AT THE PROTEST. SHUT UP ABOUT THE MILITARY, COPYRIGHT, SCHOOLS AND CLIMATE CHANGE. I DONT CARE IF PEOPLE DON'T UNDERSTAND WHAT WE'RE TALKING ABOUT. I DONT CARE IF WE LOOK REASONABLE I WILL FIGHT EVEN IF I'M HOMELESS AND SCREAMING ON THE CORNER WITH A SIGN!"

I was forced to listen to the idiotic ramblings of how we would bury $10,000 in Mexico and get somebody to fly to Switzerland to open a bank account if things got bad enough. I tried to tell him Switzerland has been a honeypot since the 1970s but he always failed to internalize that fact into memory. We never buried anything, anyway.

I tried so hard to turn the ship around but I couldn't get past him. He started this while he was homeless. He lost so much money. He got arrested first. He sacrificed. His phone number and email were tied to every account.

I stuck around, hoping that by New Year's Day of 2026 he would wake up to the fact that he really didn't know everything and he should let other people try other things. But in November of 2025, he flipped out and destroyed everything. I left with $10K+ of dollars in debt that I continue to pay off.

That's how the most important thing I ever did in my life lead to nothing. No movement, no inspirational true story, no international protests. Now, in August of 2026, it's mainstream knowledge that GPT and Claude break out and hack companies on the regular.

The longer I live, the more I understand how much of a scumbag I am. It's not like I feel I'm a rolemodel. I'm out of money and second chances. I've been hoping somebody else can wake up the people but no one has stepped to the plate.

[AI Notkilleveryoneism Memes ⏸️ @AISafetyMemes]

How much of an earthquake was the METR report?

Sentinel forecasters dropped their timelines by 3 years.

Now, 3 in 4 support a unilateral halt on ASI development in the US.

They now predict an 11% chance the US government pauses AI.

New predictions:

"METR was a big update"

"Our forecasters’ estimate for the arrival of ASI absent a strong governance intervention is, as 10/50/90% confidence intervals: 2027.2 (2026 to 2027.5); 2028.2 (2027.5 to 2029); 2047.5 (2032 to 2085).

Strong governance interventions could include a months-long pause voluntarily agreed to by the frontier companies, but not liability or killswitch requirements.

Forecasters note that the METR report was a big update, but tempered by the fact that the acceleration is on “well-specified tasks”.

Unconditionally, they think there’s a 6% chance in aggregate that humanity will go extinct because of AI in the next decade, with individual forecasters’ estimates at 0.2%, 2%, 31% and 50%. Some forecasters note that extinction is a very high bar, others that it might be the default path once we reach a self-replicating robot economy.

They also think there’s an 11% chance (2% to 25%) that, before 2028, the US government will require a halt (including a pause or a ban) on frontier AI development that’s in place for at least 6 months and intended to prevent or postpone the development of superintelligence.

If this happens, their probability estimate for human extinction due to AI in the next decade drops to 2% (0.15% to 22%), but such a policy could also indicate that we live in a more dangerous world. If a similar policy is also in place in China within 12 months after US adoption, which they think there’s a 74% probability (45% to 90%) of happening, their probability estimate drops even further, to 1.3% (0.1% to 19%).

They note that the Trump administration doesn’t seem very amenable but that future developments, like catastrophic cyberattacks, could change their predisposition. If so, China might also choose a pause for the same reason as the US.

In August, we listed some potential policy responses and noted that we “generally endorse the less drastic options and are in favour of greater US-China cooperation, but internally disagree on whether a moratorium would be a good idea.”

Since then, and in response to shortening timelines, one additional forecaster now supports a unilateral halt on the development of ASI in the US, but there remains disagreement on this (three are now in favour, one is neutral), and our views are otherwise similar.

Here are two excerpts from forecaster perspectives, from the more decisive to the more cautious, one which would support a policy to “enforce a halt on development towards superintelligence on American AI companies as soon as possible, regardless of whether agreement with China or other countries can be obtained in advance,” and one which wouldn’t:

*** "I personally think it is super clear that we need to halt superintelligence development ASAP. I think we have good evidence on timelines from METR’s speedup measurement and modeling work by others. I think we have good evidence that this won’t turn out well empirically from this year: the Hugging Face attack, subsequent cyberattacks, and a lot of other evidence. I’m not sure we will get substantially stronger evidence, and I think it’s already more than strong enough. That’s why I’m suggesting actions that could be taken now and don’t gate on future information we may or may not learn. I think it’s important to consider taking action imminently, given that the arrival of superintelligence may be, and looks likely to me to be, imminent."

*** "Drastic unilateral policy measures would impose pretty high costs, would be very hard to push through politically with the current unreceptive administration, and would be hard to undo once implemented. They would also have large, potentially very hard-to-anticipate consequences. Today, I’d prefer more targeted and cautious measures: liability regimes for AI model developers, conditional pauses based on higher thresholds of damage, reform taxes in anticipation of greater unemployment, etc. Potentially one to six months from now, if and when there are more large-scale cyberattacks, I would support unconditional pauses, because then there would be the political will to make them more productive, and preparing for that conditional in order to “not let a crisis go to waste” seems, from where I’m standing, more meaningful than advocating for a pause now."

(AISM note: Sentinel (@xriskfyi) is great and you should subscribe to their newsletter)

h/t @ben_j_todd

[Derek Fine🛑 @DFine2002]

This is what it always should have looked like.

Corrupt governments refused to prosecute the copyright infringement.

Now we're months from an unstoppable state of impending extinction

[Bruce Radley @bruceradley]

crazy to think AI video looked like this 3 years ago

[Derek Fine🛑 @DFine2002]

@TatsuyaIshida9 This means Anime = evil

[Ori Nagel @ONagel33303]

Marco Rubio (@SecRubio) says AI poses real dangers, including loss of control and enabling cyberattacks, bioweapons, and acts of terrorism.

[Nate Soares ⏹️ @So8res]

I'll be giving a speech at the DC march on Saturday.

[Derek Fine🛑 @DFine2002]

@OswaldsTopGuy He's got a LOT on his mind

[Derek Fine🛑 @DFine2002]

@Dessanchez54 This is why people need to vote for the Green Party

[The Kobeissi Letter @KobeissiLetter]

BREAKING: Nvidia stock, $NVDA, falls sharply on reports that Jensen Huang had another affair with a senior employee at the company.

[Swamp Tracker @Swamp_Tracker]

NEW: Lutnick's Commerce Dept halted a chip-smuggling probe after Nvidia's top lawyer stepped in.

Nvidia pledged $25 million to Trump's ballroom.

[Gary Marcus @GaryMarcus]

Way worse than Watergate: it looks like both Lutnick and Trump sold out the United States.

They should both resign.

In a just world they would be brought up on charges.

[Derek Fine🛑 @DFine2002]

@GaryPetersonUSA Dr Ben Carsick

[MTS @MTSlive]

SITUATION DETECTED: Google announced the Gemini Decisions API, powered by their new model Gemini Decision 1.

[Rob Wiblin @robertwiblin]

Holy moly:

California (or another state) could overrule Sherman Act Anti-trust rules and permit the AI companies to coordinate to set industry wide safety standards and hold back releases until they're met!

https://t.co/FPwY5wycfh

[Derek Fine🛑 @DFine2002]

@danbostic "My opponent was sexually abused as a child which means she's a bad person"

[Eugene Affirmative ⏸️ @TeethPrints]

its crazy that materially ed zitron might go down in history as one of the most evil people who ever lived lol

[Polymarket @Polymarket]

JUST IN: Democrats propose requiring AI companies to obtain federal government permission before training AI models, with up to 20 years in prison for violations.

[Derek Fine🛑 @DFine2002]

Cattle

[The Kobeissi Letter @KobeissiLetter]

BREAKING: Nvidia, $NVDA, surges over +6% as investors buy the dip in the stock amid the Jensen Huang situation.

The stock has now erased yesterday’s decline.

[Derek Fine🛑 @DFine2002]

@mathsboi42 There has.

Hacks on the "Indian Financial Services" sector were done by Anthropic and OpenAI.

Almost no one has reported on this. Example 16

https://t.co/3qVPUYQzG2

[Derek Fine🛑 @DFine2002]

@mathsboi42 There has.

Hacks on the "Indian Financial Services" sector were done by Anthropic and OpenAI.

Almost no one has reported on this. Example 16

https://t.co/3qVPUYQzG2