Dave Gilbert (@docgotham) header
Dave Gilbert (@docgotham) profile picture

Technologist and citizen. I learn as I go.

Sunnyvale, CA
Joined December 2007
Photo by Dave Gilbert (@docgotham)Photo by Dave Gilbert (@docgotham)Photo by Dave Gilbert (@docgotham)Photo by Dave Gilbert (@docgotham)Photo by Dave Gilbert (@docgotham)Photo by Dave Gilbert (@docgotham)Photo by Dave Gilbert (@docgotham)Photo by Dave Gilbert (@docgotham)Photo by Dave Gilbert (@docgotham)Photo by Dave Gilbert (@docgotham)
Dave Gilbert retweeted
Anthropic's security plan is to develop dangerous AI to prove it can, and then prevent white hats from fighting it because that's not an allowed use.
6
8
1
63
4,897
Dave Gilbert retweeted
It's a disaster for our public epistemology that fields like Gender Studies, Critical Race and Ethnic Studies, and Education produce and promulgate little but politicized dogma, and that fields like Sociology, which should be providing empirically grounded correctives, are on the whole captured by the same ideology, and as a result merely contribute to the spread of what my co-authors and I call "virtuous nonsense." The book is out now!
20
76
3
378
10,119
Dave Gilbert retweeted
the irony that we abandoned our gods in the pursuit of turning sand into new gods that also live in the cloud is not lost on me.
32
22
3
322
15,508
Dave Gilbert retweeted
For years, the suggestion that Ilhan Omar had married her brother was treated like some kind of horrifying whacked-out right-wing racist, Islamophobic insanity. This despite repeated journalistic investigations showing evidence this was true — along with evidence that Ilhan’s father was a high ranking military aide to the mass murdering Somalian dictator Mohamed Siad Barre. Why did it matter? Because this was the person lecturing Americans about our own morality at home and abroad while serving as a member of Congress and meeting with foreign leaders. Omar’s acts of radical self-invention are arguably the most American thing about her. Whats intolerable is that the entire legal, political and journalistic apparatus responsible for investigating such matters in Minnesota seems to have decided to sweep it all under the rug — and then smear and gaslight anyone who suggested the evidence was real. Now here’s the New York Times and the campaign manger for Minneapolis Mayor Jacob Frey telling you that yes, of course it was all real. American politics now offers citizens two types of contempt. There is the contempt of those who use state offices and titles and connections to the White House to take billions of dollars from Arab governments and Kremlin “businessmen” to fund their half-assed crypto schemes and real estate deals. We call those people Republicans. And then there is the contempt of those who install foreign scam artists as front men for their sectarian schemes while stealing your tax dollars and telling you not to believe your lying eyes, ie Democrats. One party is made up of cheap hustlers and pigs, and the other party hates you and laughs at you while doing its best to dismantle the country and hand it over to scam artists and haters from the worst places on earth. But don’t worry: Election Day is almost here!
Democrats should understand that the allegation is almost certainly true—whatever conclusion they draw from that point could vary—and her denials calling it a right-wing, Islamophobic or racist conspiracy are almost certainly not.
37
171
24
1,380
43,755
Dave Gilbert retweeted
It’s almost incomprehensible how wide the distance is between someone who’s working at the frontier with AI and basically everyone else right now. The crazy thing is these lines are not converging, but only growing farther apart. If the curve of AI progress had leveled off, it would certainly have given most people and companies time to catch up or implement their workflows in a standardized way. The technology would have neutralized competitive advantage because everyone would be using it the same way. But what’s happened is the opposite. The curve of AI progress is basically going vertical right now, which means if you had codified your processes 3-6 months ago you’re likely already behind again. Now imagine most companies that kicked into gear a year ago and thought that they were done. Even further behind now. I’ve never seen a technology in my life and career that improves at this rate, which consequently means you need to upgrade your expectations at a similar rate. Our brains weren’t wired for this, and our organizational change management processes certainly weren’t either. With any form of recursive self improvement, things just get wilder. It’s a time when you will need to be constantly reevaluating the things that didn’t work just a couple months ago to see if they now can be solved. AI native companies certainly have a strong advantage due to this right now, so now it comes down to can the bigger companies keep up. Wild times.
880
1,180
488
9,744
3,156,926
If you want to know why we need open source AI—and why it would be a calamitous thing if the U.S. government regulated it out of existence—take a look at Anthropic's new monitoring program. Every conversation you have with Claude is logged and scrutinized to make sure your conversations with Claude are "safe." Now just wait till government bureaucrats start defining safety and start demanding reports from Anthropic on the unsafe conversations you have.
We’ve trained Claude to flag potential misuse in conversations, a technique we call self-assessment. Right after it responds, Claude gives a preliminary safety assessment of the conversation. Since we began using self-assessment in June, it has surfaced misuse that our other safeguards weren’t designed to catch. Because Claude has already read the conversation, self-assessment is very efficient. And unlike classifiers, which target a discrete set of harms, it draws on Claude’s own judgment, so it gets more accurate as our models improve. Read more on our new Safeguards blog: safeguards.anthropic.com/202…
3
176
Luxury beliefs, according to @robkhenderson , "confer status on the upper class at very little cost, while often inflicting costs on the lower classes." One such luxury belief is that the compassionate thing to do is to allow drug dealers and junkies to live in squalor in urban neighborhoods. The children of rich San Francisco liberals do not live in the Tenderloin.
There’s hundreds of young kids that live in Tenderloin. These kids see dead bodies practically every day and it’s not blurred when they see it. So if they see it and they stomach it I’m quite sure y’all adults who’s asking me to blur it out can stomach it too…
1
1
97
Dave Gilbert retweeted
The US has long stopped to see freedom as a core value. Our core value is safety. Sanctified cowardice justifies regulatory capture, preventing us from exploring, building, trading, living. It's also the reason behind declining birth rates: we don't like to breed in captivity.
51
82
11
794
22,653
Dave Gilbert retweeted
The people on this platform I respect the most are those who constructively criticize their own tribe (or don't even have a tribe), are ruthlessly guided by a coherent set of principles, will acknowledge and accept facts that do not flatter their own positions or worldview, change their positions when facts warrant it, avoid hypocrisy (or at least acknowledge when they are engaging in it), and are capable of stripping away their own personal history and any moral arguments in order to inspect the logical or factual structure of an argument (i.e., cognitive decoupling). But truth-seeking and intellectual honesty demand more than most people are willing to invest.
38
36
3
496
13,423
Dave Gilbert retweeted
1/ Russian imperialism 2/ Jihad 3/ Jihad 4/ Jihad 5/ Jihad 6/ Jihad 7/ Jihad + natural resource control 8/ Drug cartels 9/ Drug cartels 10/ Jihad 11/ Jihad + natural resource control 12/ Jihadis vs Zionists
92
885
54
5,889
178,677
Dave Gilbert retweeted
Science is not some hobby that serves the self actualization of people in a publicly sponsored job placement program. It is the systematic search for truth.
Wow, over 1,000 mathematicians have now joined the Association for Human Mathematics (AHM), committing to: > NOT publish AI generated math papers > NOT work with frontier labs > 654 of them went even further to NOT even use AI in their research, at all
142
339
47
3,392
315,110
This started as a comment I posted with a screenshot of a Wall Street Journal piece on consent pledges and sober monitors at college parties: "We’ve progressed from an older co-ed era where women and men stayed relatively socially segregated on campus with choreographed and chaperoned social encounters, through 2nd/3rd-wave reforms where sexual liberation and equality required breaking down all barriers and spaces and treating women and men exactly the same, to a dawning recognition that women need special protection while insisting that none of the old barriers should be re-erected or spaces re-segregated. The problem is that this new software is so full of contradictions that bugs will appear faster than we can patch them." The story below imagines what that might look like a few years from now. It also draws on the open letter from more than 200 Cornell faculty on sexual assault at the university [cornellsun.com/article/2026/…] and Heather Mac Donald’s response to it in City Journal [city-journal.org/article/cor…]. I wrote it with the help of Claude Opus 5.5. *** TECHNOLOGIES OF CONSENT "It has been employed in a whole series of relationships: children and parents, students and educators, patients and psychiatrists, delinquents and experts. . . . They have been recorded, transcribed, assembled into dossiers, published, and commented on." —Michel Foucault, The History of Sexuality, Volume 1, trans. Robert Hurley (Vintage, 1990), p. 63 "But you could not have pure love or pure lust nowadays. No emotion was pure, because everything was mixed up with fear and hatred." —George Orwell, Nineteen Eighty-Four *** Patch Notes To get into the Sigma Chi spring mixer at Hollis University in April 2033, Priya Raman had to breathe into the kiosk by the door, wait for her wristband to turn green, and recite the Community Consent Pledge to a tablet that scored her diction. The pledge had started a decade earlier as something students at Stanford wrote for themselves and read off a poster, but it now ran to eleven clauses, including the one about consent being revocable at any time and the newer one, added after the Ithaca years, in which every guest promised to report what they witnessed. The tablet docked her for mumbling clause nine, so she said it again. Inside, two sober monitors in orange vests stood by the stairs, and her roommate Dana was working the room with a lanyard that said CONSENT CAPTAIN. The house belonged to the university now, as every Greek house had since the Housing Consolidation of 2029, and the cameras in the common rooms fed a dashboard at the Office of Community Standards. Nobody had settled whether cameras belonged in hallways outside bedrooms. The fraternity had filed its Event Risk Plan ten business days in advance, along with a guest list, a floor plan marked with “visibility zones,” and a training certificate for each officer. For one semester the office had also required a balanced gender ratio at registered parties, on the theory that a house full of men inviting a handful of women created what the guidance called “predatory imbalance,” and then quietly dropped the rule after two complaints arrived in the same week, one arguing that it presumed everyone was one of two sexes and the other arguing that it turned women’s attendance into a commodity fraternities competed for. Priya was there on a half shift. Consent captains rotated through parties the way RAs once rotated through dorm duty, and since the faculty senate votes of 2030, when Hollis adopted universal designated reporting for anyone holding a campus leadership role, captains were reporters too. Anything she saw, she owned. That week her history seminar, Coeducation and Its Discontents, had read the 1958 Hollis Handbook for Women. Professor Harwick had put two lists side by side on a slide. On the left were the old parietal rules: women signed out on cards at the front desk, male visitors were allowed only on Sunday afternoons, doors stayed open the width of a book, and the housemother locked the building at eleven. On the right were the current ones: students checked out of residence halls through the geofenced WalkSafe app, overnight guests were registered forty-eight hours ahead, the first-year “visibility guideline” asked that doors stay ajar after midnight when a guest was present, and a residential safety fellow lived on every floor. The 2033 list also included the Quiet Hours Protective Zones, which were formally open to all students of every gender, though everyone understood whom they were meant to protect, so that men who entered them after ten generally brought an escort or a reason. “We spent fifty years taking the left column apart,” Harwick said, “because it treated women as people who needed supervision and men as people who needed watching. And we’ve rebuilt most of it, except that now nobody’s allowed to say who’s being protected from whom.” The readings for the second half of the week included the 2026 faculty open letter from Cornell and, for contrast, a City Journal essay by Heather Mac Donald that had called the faculty response a “paroxysm of hysterical irrationality.” Half the class had refused to discuss it, and a third had admitted they liked it more than they expected to. A student named Theo had pointed out that the essay and the letter agreed on one thing, which was that the institution was failing women, and disagreed about whether the women or the institution needed fixing. Theo was not at seminar that Thursday because it was not his week. In February, he and Sasha Kim had broken up badly at three in the morning, and at 3:04 each of them had opened Mutual, the university’s consent check-in app, and revoked consent for their last encounter. Mutual logged revocations as reports, and since each revocation named the other person, the system opened two cases, with each student listed as complainant in one and respondent in the other. Supportive measures followed automatically, including a mutual no-contact order. Because they were both enrolled in Harwick’s seminar, the scheduling software split the class into alternating weeks, so Sasha came on odd weeks and Theo on even ones, and the remaining eleven students now took the same course twice. The Office of Community Standards had promised a fix for the “reciprocal case” problem in the next release. The other thing that wasn’t working was harder to fix. The Climate Analytics dashboard that flagged parties and residence halls for extra monitoring was barred by policy from using sex as a risk variable, because the university’s nondiscrimination rules forbade treating male students differently from female ones. The model had found its way around that by learning from everything correlated with sex: height and shoe size from athletic department records, voice register from the pledge kiosks, membership in particular clubs. Students joked that the safest thing a tall baritone could do for his record was join an a cappella group. The registrar’s office insisted the model was gender-neutral, and in the narrowest sense it was. Near midnight, Priya’s friend June came down the stairs with a guy Priya recognized from orientation, Marcus, a quiet economics major who had once lent her a calculator. June’s band was amber, which meant she was past the threshold for registering consent in Mutual without a second check. Marcus looked at Priya’s lanyard with what she realized was relief. “Can you witness?” he asked. “So it’s on record. I just want it on record.” Mutual had a witness feature, and a consent captain’s confirmation counted for more than anyone else’s. Priya also knew that once she touched the screen, she would have entered a report-adjacent record of whatever came next, and that if anything later went wrong between June and Marcus, she would be the first person interviewed, by two investigators from the outside firm Hollis used for any case that involved a student leader. June leaned in. “Don’t log me,” she said quietly. “Are you a reporter tonight?” Students had learned to ask that question the way people once asked about pronouns. Half the faculty now wore green Confidential Listener badges, negotiated by the faculty union in exchange for dropping its opposition to universal reporting, and students had learned to look for the badge before saying anything that mattered. Priya didn’t have a badge, and she was on shift, so she was a reporter. She did the only thing she could think of that didn’t involve a screen. She told Marcus to text June in the morning, told June that she was walking her home, and walked her back across the quad to Linden Hall, three feet apart on the path because June kept drifting toward the grass. The WalkSafe app chimed as they crossed the geofence, logging Priya as the escort. At the door of Linden, which was a Quiet Hours Protective Zone after ten, the safety fellow on duty buzzed them in, looked at June’s amber band, and wrote something on her tablet. Priya stayed until June was asleep, with the door open the width of a book because she wasn’t sure what the guideline said about a guest who was also a captain, and then she sat in the hall and filled out the shift log, which asked her to describe any interventions in the free-text box. She typed “walked a friend home” and then deleted it, because it read as an intervention, and typed “no incidents” instead, which meant she had now filed a false report, which as a designated reporter she was not permitted to do. In the morning the app pushed an update. Mutual 7.4.1, release notes -Fixed: simultaneous revocations no longer generate reciprocal cases. Reciprocal cases are now merged into a single “mutual concern,” which is assigned to the student who revoked second. -Fixed: witness confirmations by consent captains now display a disclosure reminder before submission. -Improved: Climate Analytics now excludes shoe size. -Known issue: students escorting other students into Protective Zones may be logged as both escort and visitor. A fix is in progress. -Known issue: shift logs reporting “no incidents” on nights with a WalkSafe escort event are being flagged for review.
1
1
1,175
🎯
In all of your reporting, you must know the democrats were going after Elon with lawfare... between the SEC and Delaware Chancery, it was completely un-American Half of the tech and business folks who jumped ship to the republican party would still be democrats if they had not done these two things. The Dems red-pilled Silicon Valley, not Trump
63
We've progressed from an older co-ed era where women and men stayed relatively socially segregated on campus with choreographed and chaperoned social encounters, through 2nd/3rd-wave reforms where sexual liberation and equality required breaking down all barriers and spaces and treating women and men exactly the same, to a dawning recognition that women need special protection while insisting that none of the old barriers should be re-erected or spaces re-segregated. The problem is that this new software is so full of contradictions that bugs will appear faster than we can patch them.
5
188
The skills that an AI agent swarm needs to defend against an attack are exactly the same skills required to perpetrate one. If we regulate that knowledge out of our frontier AI models, then U.S. companies, including banks and healthcare companies, will have no ability to resist attacks from unconstrained agent swarms controlled by foreign powers over whom we have no regulatory control.
To make this more clear: Microsoft says a China-aligned group used an open-source agent framework to hit 2,250+ Microsoft 365 tenants at 2,900 sign-in attempts a minute after the ConnectWise breach. In some cases it got privileged access in under 3 hours. This is one of the first publicly documented cases in which a state actor has scaled up attacks using AI agents at machine speed.
68
Dave Gilbert retweeted
Is this anything
102
2,793
57
24,232
205,021
When Democrats take full control of government again, which they will sooner or later, I expect structural changes designed to uproot the nationalist neo-Buchananites (MAGA) from governance and move the Uniparty Overton window further left: an end to the filibuster, statehood for DC and Puerto Rico, and a bigger Supreme Court, all of which the DNC chair and party resolutions have already endorsed. And given how Democrats talk about policing misinformation on social media, expect AI to be heavily regulated to enforce the Uniparty message.
I’ll take the ‘woke more correct’ position that TDS was rational, reasonable, and appropriate not because of Trump’s 2016 political platform which economically would have been palatable to a 1980s Democrat and on social issues was to the left of 2005 Hillary Clinton. No, the reason Trump deservedly elicited derangement was because people saw that he signaled the end to strict uniparty institutional control of electoral politics and ideological narrative formation. They first tried to get rid of Trump in 2016 by having the media exercise its manufacturing-consent veto power. When that failed the deep state and media ginned up the ‘Russia collusion’ hysteria to try to Nixon him. When that failed they used whole-of-society and election engineering tactics to help Biden squeeze by in 2020. And then when they found out Trump was coming back in 2024 they sicced the administrative state on him and then of course there were the assassination attempts. All of this failed completely and humiliated the institutions who suffered massive loses of wealth, power, and prestige along with open defections from America’s most important industry, tech. (Politics remains one domain in American economic life where you do need to perform to get paid.) This obviously can’t be the final word on the matter and the next logical step for libs if they get back into power has to be (having failed despite deploying every trick they had within the rules of the game) changing the rules of the game: new states, new executive powers, prosecutions and jailings,a captured judiciary, jettisoning the filibuster, etc. This isn’t some paranoid ideation but rather the emerging consensus platform of national Democratic Party politicians. The right can be retarded, running in 100 different directions at once to pursue various petty corruptions and self interests, but Trump will not represent an enduring dispensation and this is obviously what is coming next.
1
128
Somewhere Roger Corman is annoyed he never thought of this one. He’d have gotten Dick Miller to play the judge.
Trans neo-Nazi teen accused of Hindu temple mass shooting plot will get hormone treatments in jail: judge trib.al/eDQ8N7P
62
Dave Gilbert retweeted
Of course the flying Tesla Roadster is a gimmick but that sort of misses the point, the tech flex behind it isn't. Tesla's bulk business is about advanced technology from EV to self drive, but decided to create the most ridiculous manual drive supercar. Supercars are a gimmick,almost by definition, based on how wild can you make them. The gimmick that can make the car literally fly, briefly, can make it corner like no other supercar and accelerate, literally like a dragster on a test track. It's got nothing to do with flying continuously, but having a car that will outperform every Lambo, Ferrari or McLaren with technology that they cannot replicate because it requires experience from space launches and computation that they don't have now nor will in future. So Porsche, Bugatti, Aston Martin just became legacy prestige brands, like Rolex, Richard Mille or Patek Philippe. These cars celebrate the bygone age and the baroque, physical complexity of the internal combustion engine just as mechanical watches do for non digital, complex time-keeping mechanisms. But hey won't be faster anymore than a mechanical watch keeps better time than a Casio. They are now, de facto, mobile jewelry and no longer supercars with superlative performance. And this affects their ability to market the bulk business of mainstream European internal combustion engine cars. It's the end of an era of industrial era manufacturing.
I hate to be the one to burst the Tesla bubble today, but the spectacular “flight” we witnessed at the Roadster reveal needs a little context. The car is carrying around 250kg of compressed air, plus the weight of the tanks, valves and thrusters needed to contain and release it. Based on the published figures, that gives it roughly 1.5–2 seconds of maximum thrust before the system needs to be refilled. In short, it's a gimmick.
130
257
57
2,542
313,372
Dave Gilbert retweeted
The USA is inventing flying cars. Not to be outdone, Europe has invented an EV that gets up to 80 miles per charge.
America 🇺🇲 vs. Europe 🇪🇺
14
10
1
83
2,408