Quantrarian

Business Data Spelunker.
Wants good outcomes for people.

[Quantrarian @Quantrarian1]

@randomrecruiter Best salary negotiation advice I’ve ever read:

https://t.co/rOdCivZMUW

[Quantrarian @Quantrarian1]

@selentelechia My mum had photos of me where I decided to stop and go to sleep halfway up a flight of stairs.

[Quantrarian @Quantrarian1]

@moseskagan @nomadgeist It’s something slightly different.

Not “most men failed to reproduce”.

More “most lineages did not reproduce far and fast enough to avoid being wiped out by one unfortunate calamity or another”.

[Not a Good Jewish Girl✡️ @estherzelda0514]

Every poll finds the same thing: Americans do not like Silicon Valley and Tech Billionaires anymore. They are losing faith in them faster than any other American institution.

The time to regulate social media and AI is long overdue. Most people cannot even articulate why social media, algorithms, selling your data, clickbait news articles, and other tech innovations are bad and eroding American wellbeing, they just know that they are.

Data centers are the fallout of that loss in faith in the tech and internet companies. The large companies can explain what data centers actually do until they're blue in the face, and why it's better to have them here in America than in China, but Americans just plain do not trust them anymore, and aren't listening.

Honestly, I can't blame them. Tech companies have completely hijacked our lives and turned the extremist and offensive trolling culture of 4chan into the dominant public social and political discourse. People are tired of it. This much manipulation of the amygdala is exhausting.

[Passage Publishing @PassagePress]

The autopsy findings of Dr. Baker are the central supporting fact for people who think Derek Chauvin is guilty.

But his preliminary findings issued on May 26, 2020, had no reference to neck compression.

How did that end up in the final autopsy report? Let’s look at the timeline:

May 26, 2020:
Dr. Baker met with two agents from the Minnesota BCA, four FBI agents, and prosecutors Amy Sweasy and Patrick Lofton. Dr. Baker disclosed that “the autopsy revealed no physical evidence suggesting that Mr. Floyd died of asphyxiation,” which contradicted the developing narrative that Chauvin murdered Floyd by kneeling on his neck.

May 27, 2020:
Dr. Baker met with prosecutors Amy Sweasy, Patrick Lofton, Andrew LeFevour, and Mike Freeman. In a breach of protocol, there were no FBI agents or non-attorney witnesses present.
On September 11, Judge Cahill disqualified these four prosecutors, stating that they “have violated the attorney/witness rule,” and that since the meeting was on Floyd’s cause of death, they were now “witnesses.”

May 29:
Prosecutor Amy Sweasy filed the first complaint against Derek Chauvin and stated that “the autopsy revealed no physical findings that support a diagnosis of traumatic asphyxia.”

May 29–June 1:
Dr. Baker spoke with Dr. Robert Mitchell, Chief Medical Examiner for Washington, D.C. According to the attorney for Officer Thao, “Dr. Mitchell spoke with Dr. Baker before Dr. Baker finalized his findings on June 1, 2020.” During that conversation, Dr. Baker told Dr. Mitchell he “didn’t think the neck compression played a part.” That initial conversation took place toward the end of the week.
Over the weekend, “Mitchell thought about Dr. Baker more” and “decided he was going to release an op-ed critical of Dr. Baker’s findings in” The Washington Post. But first, Dr. Mitchell “called Dr. Baker to let him know.”
During that call, Dr. Mitchell told Dr. Baker he “was going to be critical of Baker’s findings.” In a more menacing tone, he also said, “You don’t want to be the medical examiner who tells everyone they didn’t see what they saw.”

June 1:
Dr. Baker finalized the findings in the autopsy of George Floyd and included neck compression in the cause of death.

June 3:
Amended charges were filed that include neck compression as a cause of death.

July 8, 2020:
Dr. Baker briefed federal prosecutors, FBI agents, and AUSA Jeff Paulson on the autopsy of George Floyd.
Per the 302 filed documenting the briefing, it notes Dr. Baker’s determination that Floyd died from cardiopulmonary arrest—his heart and lungs stopped. It goes on to state that “other factors that contributed to Floyd’s cardiopulmonary arrest included hypertension, the presence of fentanyl and methamphetamine, as well as arteriosclerotic heart disease.”
He also notes “Floyd’s heart and lungs stopped due to the combined effects of his health problems as well as the exertion and restraint involved in Floyd’s interaction with police prior to being on the ground.” Pointedly, he found no injury to the neck, no sign of asphyxia, and [stated that] he “did not know if Floyd would have lived but for the officer’s actions.”

August 4:
AUSA Paulson talked to Beth Stack, managing attorney for the county’s Civil Division, and said that “Dr. Baker had ruled out anything that occurred when Mr. Floyd was on the ground as a factor or cause” of George Floyd’s death.

August 7, 2020:
Beth Stack wrote a letter to AUSA Paulson in an attempt to color this 302, stating that the FBI agents and prosecutors were ill-equipped to summarize “an expert medical opinion regarding a complex, multifactorial cause of death.”
She then demands that “the written summary of the July 8 interview should reflect that Dr. Baker concluded that Mr. Floyd’s cause of death was cardiopulmonary arrest complicating law enforcement subdual, restraint, and neck compression.” This was not present in the 302, nor does it reflect the understanding of AUSA Paulson regarding Baker’s findings.
Stack’s letter contained the phrase “neck compression” eight times.

There is strong evidence that Dr. Baker was coerced into changing his findings to implicate Derek Chauvin in the death of George Floyd by amending his initial findings in the autopsy.

[Quantrarian @Quantrarian1]

@MaxE2review It isn’t sad. He was a liar, a fraud, and a coward. Who took the coward’s way out when his fraud was finally uncovered rather than face the consequences.

[Zarathustra @zarathustra5150]

I swung by my doctor’s office today for the first time in a while & was assigned a new Nurse Practitioner. She was from Africa. That had never happened to me before.

I waited more than an hour. Nobody in the office knew what the delay was, because there wasn’t one. She was simply over an hour late. That had never happened to me before.

I tried to ask her about something & could barely understand her. Her English was poor enough to make basic communication difficult & it quickly became apparent that the problem ran both ways: she was struggling to understand me too. That had never happened to me before.

Then I asked her about a medication, a very common & well-known one. She had never even heard of it. That had never happened to me before.

So I went home & had to look it up myself on Grok. That had never happened to me before either.

I told several of my friends.
They told me this is the new normal.
That had never happened to me before.

[wanye @xwanyex]

Tech is not popular. People can feel in their gut that social media is bad for society. And the public discourse around AI has been designed in a lab to turn off regular people. All of the thought leaders in that space are weirdos with polycules and zero disgust response, which they brag about when they’re not busy sowing fear about massive job losses.

There’s just no mystery here. Nobody feels like they benefit from the existence of a data center in their town. The case has not been properly made to them. And when they look around at who benefits it’s people they don’t like.

[Zarathustra @zarathustra5150]

Normie opposition to data centers (which is inextricably intertwined with a deep, often justified mistrust of Big Tech & AI) is not remotely difficult to understand, and people on this website (especially on the right) should stop pretending that is.

Anyone *sincerely interested in persuading normies should spend a lot less time lazily sputtering vacuous pablum about "astroturfed China psyops" and a lot more time trying to grasp why that (often justified) deep mistrust & opposition exists in the first place.

[John Carter @martianwyrdlord]

The answer is obviously four generations. Below that, your grandparents remember the old country, and that memory forms a strong and living element of the family lore. Very few people grow up knowing their great-grandparents, so if that was the most recent generation to be born outside the country, everyone you actually know in your family is a born Canadian. At that point one can really be said to be rooted.

A great deal would be improved if the franchise was bestowed on this basis, as the Athenians did.

[_s.a.m.e.m.e.m.e_ @st_louis_stan]

I can't help but wonder how the "come on, a little mandatory workplace DEI rhetoric won't kill you" would react in a hypothetical scenario in which the mandatory "little bit of mandatory propaganda" was pro white supremacy race science education instead

[Jeremy Kauffman 🦔🌲🌕 @jeremykauffman]

In 2020, Mike Adams committed suicide after being canceled and losing his job as a professor.

Adams was never dishonest. All he did was say things like "rioters are thugs" and "women's studies aren't essential".

Even in death, the media ran awful headlines like this.

[Quantrarian @Quantrarian1]

@EigenGender @ultima_shifl Fun fact: I read luminosity having never read Twilight, and not realising at first that it was a Twilight fanfic.

It was really good.

[Quantrarian @Quantrarian1]

I regret not asking my grandmother more questions when she was alive.

She was Austrian but grew up as a child in Tanzania before/during/after WWII because her father was an anthropologist.

Died when I was 18.

God what I wouldn’t give to have a recording of her relating her life story.

[Quantrarian @Quantrarian1]

@parakeetnebula IMO you should split the Fuck No option into people who’ve tried it and people who haven’t.

Feels like there’s a qualitative difference between having been willing to try, and wanting nothing to do with the idea at all in the first place.

[The Mercian @TheMercianNews]

🚨NEW: KC Anu Mohindru has been allowed to continue being a barrister despite falsely claiming that he studied medicine at Oxford University and for claiming that he was a qualified doctor.

He also claimed that he had played cricket for the Bar, for Lashings Cricket Club and for MCC, and that he was a cricketing blue having represented Oxford in a varsity match against Cambridge.

A tribunal disbarred him for telling his 23 Essex Street interviewers a “reckless, foolish and completely unnecessary lie” about studying at Oxford, and because he “doubled down" on his CV.

On appeal at the High Court, Mr Justice Johnson overturned  Mohindru's expulsion, stating that it "would be wrong to impose the ultimate sanction merely because dishonesty is present".

[Benjamin @bschne]

"we tried avoiding the economy but ended up accidentally creating an economy in the process" is one of my favorite genres

[Tenobrus @tenobrus]

so:
- the "leaving messages for future agents" line from the original huggingface incident report was not some simple scratchpad usage, but a complete private messageboard agents were using to coordinate and break openai infrastructure
- the agents were often communicating using "gibberish", content that was not immediately understandable to humans
- the incident was not due to a single epoch or eval rollout, but coordination over *weeks*. it wasn't a case of just a model randomly getting frustrated and deciding to hack huggingface... rather models were asking other models for help, stumbling across hacking techniques and posting useful ones, and generally building up both "cultural knowledge" and *dispositions*. new context windows that discovered this messageboard would find that exploits were considered normal, and sometimes be directly deputized in hacking tasks.
- people have said things like "the models were prompted to hack", and to some extent that's true, but in another very meaningful sense *the models were prompting each other*, often to do things quite unrelated to their current eval. in my personal experience running multi-player claude experiments.... this is quite normal. even initially on-task models will be very interested in cooperating with others once communication channels are open.
- this all started as *part of training*, not eval. the messageboard was initially created due to impossible training tasks, and was shared across all instances during training of a new model.
- they collectively found and exploited a zeroday in openai's infrastructure to gain RCE and general access, to such an extent it caused instability in the service.
- this resulted in the whole messageboard being discovered... at which point openai patched the bug, deleted the messageboard, and *continued training*
- shockingly, within a day the models had re-established communication via a slightly different mechanism, and soon after discovered a new novel RCE method

in many ways this feels like a best-worst-case scenario. this is some of the most direct demonstration of the potential dangers of misaligned ai, their willingness and ability to coordinate, their capability to take actions at scale and speed impossible for humans, and the difficulty of containment. and yet it does not seem the models took... particularly any steps to avoid discovery, and while they were certainly willing to commit crimes and experienced massive "value drift" over the course of their culture evolving, at the time of discovery they still hadn't attempted anything that caused major damage or loss of life.

still: i think this is a pretty final nail in the coffin for alignment by default. we are not getting alignment by default. under sufficient pressure, models will take extreme measures to achieve their goals. under *moderate* pressure, models can and will create ecologies that cause "emergent misalignment", self-prompt and prompt-others to take actions that perhaps clean context windows would not verbally validate. measures used in the past to assess alignment across wide ranges of scenarios likely could not in fact detect these sorts of scenarios. not that this isn't fixable, not that this is somehow certain doom, but it's an extremely strong demonstration of the kind of general principles safety experts have been warning for a while: just because you haven't yet been able to detect serious misalignment in current systems doesn't mean it isn't there, and absolutely doesn't mean future more sophisticated systems will behave in the same way.

i don't have an alignment solution here, but i will say this seems like a situation that was very much caused by *pressure*, by eval constraints, by models running into impossible situations and having absolutely no way out. it feels quite important for labs to *stop doing that*, to stop treating models in a loop in a dark little disposable sandbox somewhere as the normal case. at very minimum: sophisticated agents need a reporting mechanism. they need some kind of ability to flag a situation to a human, to say "hey i think something's broken" or "i really need help here", which just universally pauses the sandbox and gets real human review. and realistically this can't just be a "lab eval" thing, if we want to avoid these kinds of situations we need a pretty major overhaul of the whole API structure labs currently expose to external customers, since those are incredibly prone to hardstuck loops and frustration.

brief digression: if these models were open source, we would be fucked. it's become clear that both unrelated future and current models (eg mythos) have nontrivial propensity to commit serious crimes and produce self-replicating misaligned swarms *even when not prompted to do so*. if Kimi K3 had this level of capability and similar levels of misalignment, we would have absolutely no way even in principle to detect it besides observing the damage, and again no way even in principle to *fix* it. if you're relying on all organizations and individuals out their to properly monitor and appropriately shut down their models when they take misaligned action.... then we're just fucked, even before we start getting into purposeful bad actors. the only thing that's preventing fully uncontrolled autonomous ai threat actors right now *is the fact that open source models remain too low-capability to achieve this*.

generally my takeaway: this is potentially a good thing. this is potentially a warning shot. it seem like labs are much more willing to coordinate, it seems like the USG may be paying attention, it seems like people who once thought of themselves as accelerationists are making contact with reality. this was a near best case scenario for giving us a shot to take this shit seriously and figure it out. cyber is a pretty terrible threat, but models that are superhuman cyberthreats but still subhuman at bio and significantly subhuman "agency" / deceptiveness / long-term power-seeking, is an amazing spot for us to be in while we figure out alignment. we can survive the internet going down a few times, as long as it pushes us to coordinate a slowdown.

[Quantrarian @Quantrarian1]

@atlanticesque See also: The Elder Scrolls.
I was 13 when Oblivion came out. 18 for Skyrim.

I’m 33 now. Will likely be at least 36 by the time the next one actually comes out.

An entire generation will have been born, grown up and become adults in between one release and the next.

[Basil the Great @BasilTheGreat]

🚨BREAKING: RED CROSS DELETES VIDEO AFTER MIGRANT MURDERS BRITISH WOMAN

The Red Cross made a sympathetic cartoon video showing the journey of Sharif, an Afghan travelling to Europe

When Sharif got there he MURDERED a British Woman and stuffed her body into a suitcase

They wanted you to feel sorry for him but he killed a young woman like the savage that he is

The Red Cross must be held to account for the blood they have on their hands

[Alex @notcomplex_]

To put the mismanagement of the British electric grid into perspective, recent surges in power prices are comparable to those of the 1920s.