Twitter Wars 1

Daniel Golliher 🗽

@danielgolliher

“I don’t like you, and here’s your permit” is the essence of a pluralistic, small-l liberal society. https://t.co/f9B7Y8dCd7

Daniel Golliher 🗽

@danielgolliher

I made https://t.co/9xjg6jllWK with NYC Open Data. Browse all events in NYC parks, and sign up for emails for events near you. @NYCParks had a lot to offer: exercise, star gazing, movies, and more!

Daniel Golliher 🗽

@danielgolliher

"If somebody sends me a spicy email about data centers, I say, 'Please, please, please hand-walk that to my office.'" https://t.co/IaJiqkr2Dq

Daniel Golliher 🗽

@danielgolliher

Imagine learning about government here https://t.co/EEISvuDyI6 https://t.co/GfRKqan18L

Daniel Golliher 🗽

@danielgolliher

People tend to be scared to talk about public labor/union reform, and it largely kills any hope of the civil service reform many state capacity advocates want. So: let’s talk about it. I declare it not taboo. It’s required to get where we want to go. https://t.co/JNbMjZAVwv

Daniel Golliher 🗽

@danielgolliher

The president is supporting data centers, the mayor is pushing back on the teachers union. They’re both right in my view. https://t.co/hcIgEG761T

Daniel Golliher 🗽

@danielgolliher

Cohort 16, baby! Application are now open. We start September 14. https://t.co/0k4lrctqFf

Daniel Golliher 🗽

@danielgolliher

From a February NYT piece. Let’s hope this is the beginning of a trend 🗽 https://t.co/AkZSN1kBsv https://t.co/ee7HJGBKH8

Daniel Golliher 🗽

@danielgolliher

Sometime this fall Maximum New York is going to have events with invited members of both major parties to discuss, in a sensible fashion that examines priors and validates them live, hot-button issues at various levels of government. These could be labeled “cultural issues,” even as they have deep technical points. The overall goal is not to agree on a conclusion, but to lay out everyone’s priors and examine how they overlap, with a north star of converging on mutually agreed upon priors. Agreement or disagreement on terminal policy will be left for attendees to do as they socialize. These include, but are certainly not limited to: - Israel/Palestine - Trans athletics - Immigration On the internet, these conversations get dominated by the unhelpful and the ill-tempered. It is crucial for people of good faith to come together (in person), and fully pull apart specific claims, validating them live on a projector in front of everyone in a spirit of Maximum New York concreteness. These “cultural issues,” if you want to label them that way, make people walk on eggshells, make them fearful and wrathful, and shut down the thinking mind in the process. But this is no way to get to the bottom of things and move productively forward as Americans. We must not fear discussing these things, discovering what our friends think, and embracing an attitude of curiosity and truth-seeking. People who are more traditional policy wonk types often stay away from working on these issues more often than not, ceding the realm to the overly confident and under-challenged. So, it’s a straightforward fix.

Daniel Golliher 🗽

@danielgolliher

Legally Blonde is properly viewed as a piece of fun entertainment, but, like all quality entertainment, it also possesses the ability to pull on deeper threads for audience members who are so inclined. I saw it first in late elementary school, before I had a full appreciation of what I was seeing. But it would become my most-rewatched movie, then and now. It's a classic American tale. As Elle (@ReeseW) moves through the acts of the movie, we are presented with the following: - No matter who you are, hard word and determination will allow you to improve your lot. - Do not judge people hastily or in a prejudicial fashion. We all contain surprising multitudes and capabilities. - You can change your life by deciding to move into new domains, and you don't need anyone's permission to make that move. - You can overcome setbacks and mistakes. You can summon inner determination and help from your friends. - Others want you to succeed. They are rooting for you. - It is important to make good friends. Your characters reflect back upon each other. - You can extend grace, forgive those who transgressed against you, and go forward as friends. - A cheerful worldview won't get you everywhere, but it will get you far. - There's nothing like stick-to-it-iveness, mental toughness, and a positive attitude! Have some can-do, follow-through, and gumption! Not everyone will be as deeply touched by this movie as I was, and that's to be expected. We all attach to media in different ways, for different reasons. I love the movie because I love the law; because I love a fearless upstart; because I love the American northeast; because there is a 2000's glamour to it that is evergreen for me; and because I love stories of Americans who pick up sticks, move across the country, and--fish out of water fears and realities be damned--establish a great, new life. "What, like it's hard?" compresses a lot. Elle worked hard to earn a near-perfect score on the LSAT, and clearly possessed the capability to achieve at the highest level (even if she needed a kick). When she displays surprise at Warner about her getting into Harvard Law, it reflects a worldview that we've already seen play out by that point in the movie: just do what needs to be done to get where you want to go--what's hard about know that? It's the same lesson that comes across in the famous 1960 interview with Orson Welles about making "Citizen Kane" (https://t.co/UbJH3MkS7O): >> Interviewer: What I’d like to know is where did you get the confidence from to make the film with such — >> Welles: Ignorance. Ignorance. Sheer ignorance. You know, there’s no confidence to equal it. It’s only when you know something about a profession, I think, that you’re timid, or careful or — >> Interviewer: How does this ignorance show itself? >> Welles: I thought you could do anything with a camera that the eye could do or the imagination could do. And if you come up from the bottom in the film business, you’re taught all the things that the cameraman doesn’t want to attempt for fear he will be criticized for having failed. And in this case I had a cameraman who didn’t care if he was criticized if he failed, and I didn’t know that there were things you couldn’t do, so anything I could think up in my dreams, I attempted to photograph. >> Interviewer: You got away with enormous technical advances, didn’t you? >> Welles: Simply by not knowing that they were impossible. Or theoretically impossible. And of course, again, I had a great advantage, not only in the real genius of my cameraman, but in the fact that he, like all great men, I think, who are masters of a craft, told me right at the outset that there was nothing about camerawork that I couldn’t learn in half a day, that any intelligent person couldn’t learn in half a day. And he was right. >> Interviewer: It’s true of an awful lot of things, isn’t it? >> Welles: Of all things. ==== "You changed your life for the better in a big way?" "What, like it's hard?"

Daniel Golliher 🗽

@danielgolliher

If there's one conviction a fine-grained look at New York City history has given me, it is this: We have always been swimming through a river of graft, corruption, and people using all that is valuable to fuel anything but the public good. Building things here has never been easy, free, or clear. And yet, here we are, with everything we have. There will always be massive incentives for powerful interests here to grab the public purse, and to use the law as their covert ally. But New York always raises an equal, and occasionally far more potent, force. That force is how Central Park survived Tammany patronage, how the boroughs were consolidated, how the subways got built, how the city rebounded from '75, how the MTA was saved and the subways rebuilt, and so much more. Our history is sometimes a stolid reminder that we can achieve big things and recover from almost anything. But sometimes it shakes you by the lapels in reminder: we have a duty to achieve for the future what the past achieved for us. A better city and a larger legacy.

Daniel Golliher 🗽

@danielgolliher

One of my go-to coffee shops in Manhattan has “Good luck out there” beautifully set at the exit. Reminds me of the Hang In There cat in the best way https://t.co/B21tGzxLYX

Daniel Golliher 🗽

@danielgolliher

Take a listen, it was a delightful conversation! https://t.co/NmKsRChjfW https://t.co/3q7LuhKoKE

Daniel Golliher 🗽

@danielgolliher

Despite increasing pro-supply sentiment in the past few years, the City Council as an institution would still rather slow or halt housing production below the level that NYC needs (>=50k units a year?). It remains the squeaky wheel of city government in that respect. This became quite apparent in its opposition to last November’s CRC ballot questions, and that same institutional impulse is on display here. @NYCMayor’s appointment is well qualified for the role.

Daniel Golliher 🗽

@danielgolliher

I have long enjoyed playing hymns to myself in the evening, and ChatGPT has helped me enjoy it even more. I’m also just playing and improvising more now. I can ask it to transpose a hymn to any key I want, or ask it to help me with beautiful embellishments. I take a picture of the hymn page, and just ask for what I want. I do this when I want a different sound, or, frankly, when anything has more than three sharps or flats and I’m not so inclined (I am not a gifted piano player; I am a chop-wood-carry-water player).

Daniel Golliher 🗽

@danielgolliher

“We will keep doing more government like this, and we will get Congress to plug in too,” is a compelling pitch. City Hall has a few things to learn here. https://t.co/nUQQ8W26CZ

Daniel Golliher 🗽

@danielgolliher

NYC in process CivLab 🤝 Maximum New York https://t.co/Lf1M5TejfN

Daniel Golliher 🗽

@danielgolliher

It is important for my wellbeing that I live within a 5-minute walk of a 24-hour diner.

Daniel Golliher 🗽

@danielgolliher

Would you guess that the report this chart is from would contain this phrase "Nevertheless, when comparing across broad occupational groups, there appear trends toward more wage convergence, i.e. less inequality"? The image the mayor posted below is chart 2 from @NYCComptroller's "Recent Trends in Personal Income & Wage Inequality," published September 2 (https://t.co/CaiV9Ox2IG). The Comptroller's report is great, and everyone should read it. However: the mayor is presenting this chart *without attribution* (how are people supposed to read the whole thing), and without any context. This manner of presentation is leading people further away from prudent policy and truth, not closer to it, and is unhelpfully stirring up resentment in a way I consider to be unwarranted and unhealthy to our politics. First: this chart (and many others like it) is often posted with the intent to provoke the following thought in viewers: "the rich are gaining most, and everyone else is left with little/their lot isn't getting better/our society doesn't help them." Even if it does not have that intent, it is clearly being taken that way by a significant part of the mayor's audience. However: is that true? The poor and not-so-poor in America and NYC, in fact, benefit from an extraordinarily generous tax-and-transfer/welfare system. About half of NYC is in our state's extremely generous medicaid system (https://t.co/sztBSeuznR), for example. The chart that the mayor posted is about income--what people earn--and says nothing about the other side of the ledger: what people pay, and what people gain in welfare and other transfers. Second: the wealthiest New Yorkers (city and state) pay most of our income taxes, in addition to earning the most income! (There are other taxes, like real estate, and so forth, but this post is about income.) In NYS, the top 50% pay almost 100% of the income taxes (see chart below from the state). The richer you get, the more disproportionately you are represented in total income taxes receipts. NYC/S already have a very progressive income tax system. Third: without commenting broadly on what our tax structure should be, it's important to note again that NYC/S already has a very progressive system, where the richest already sustain the whole system, and the poor already receive massive transfers from the city/state/feds and pay little, no, or negative taxes. Posts like the mayor's make people pessimistic and stir up class resentment, because people think our society does nothing to help help the poor, or that it isn't generous. But it does a tremendous amount of generous helping! This should be good news for people who wish it were true, because it is! Finally: the Comptroller's report itself is very honest about what it shows and what it doesn't, and it ends on a measured note that I wish the mayor would have posted--but the conclusion does leave people with a more even view of the prospects of wage earners throughout and at the bottom end of the spectrum. I've screenshotted the report's final paragraph below, but here it is again (also: please read the whole thing): "The evidence also points to what is—and is not—driving these trends. The divergence at the top is predominantly a story of non-wage income: capital gains, dividends, interest, business income, and rents, which make up most of the income for the top 1-percent and grew faster than wages did. Wage and salary income, viewed on its own, shows a much milder widening, and occupational wage data that exclude bonuses show base pay growing faster in lower-wage occupations than in higher-wage ones, and within many occupational groups wages are converging rather than growing more unequal. The labor market, in other words, has been modestly compressing pay differentials even as the overall income distribution has pulled apart. Rising concentration in New York City reflects the returns to wealth, changes in who lives here, and the mix of jobs in the city—rather than a deterioration in the relative pay of its lower-wage workers." ===== To create a "city that works for everyone," as the mayor says, means (to me) creating an economically dynamic place where people have tons of opportunity to improve their lot. Secondarily, it means helping people who fall and need a hand up. If you over-index on the second thing, you get neither the first or second thing, per: https://t.co/luwagxPBg5 And if we want to make sure we are economically dynamic, we need to make sure we increase the local rate of wealth formation: https://t.co/PLBAdoY244 Finally (actually): when I think about the things that hold back people in NYC from getting ahead (moving up, getting a nicer place to live, having cheaper utility bills, etc), I do not think of "the rich" as blocking these things at all. It is bad laws like our overly burdensome "environmental reviews," our inflexible zoning, a utility system that can't authorize and connect new energy to our grid quickly, and so much more. The mayor addresses these kinds of things, and is on board with, for example, SEQRA reform (https://t.co/w57bIck0Pe)! But this post is not healthy for our politics, and leads people away from understanding what our political economic system actually is.

Daniel Golliher 🗽

@danielgolliher

I have done the thing where I booked a 6am flight, and it seems like an excellent idea right now, due to scheduling reasons. And yet, I know that it will not be an excellent idea when the time comes, due to 6am reasons.

Daniel Golliher 🗽

@danielgolliher

Dolly is one of the few who deserves the recognition and honor of a whole civilization, and I believe she will receive it. https://t.co/mlg1SQ7yhW

Daniel Golliher 🗽

@danielgolliher

There’s a version of “if he wanted to, he would” that applies to the New York State Legislature (and to a lesser extent the NYC city council). It has massive constitutional authority to cut through our law and command change. Relative to what it could do, it is very meek. And so, one must conclude “he doesn’t really want to.” This applies to building housing, transmission lines, new subways, reshaping the government itself, and so much more. Many legislators actually just have priorities that they put above wielding the full might of their branch to change New York in certain ways. An example that shows a legislature in action to the degree I’m talking about was in 1895–it removed the private right of action for people to sue to stop the construction of the original subways. Or when it consolidated the five boroughs into one city in 1898. Big stuff! And they did it often! @nicolegelinas discusses this legislative “if he wanted to, he would” phenomenon (my words, not hers) wrt Robert Moses on here and in her book Movement. The things he did were enabled and backed by the legislature, because it wanted them. See, for example: https://t.co/RWAMy8v1br This isn’t to say that today’s legislature does literally nothing, nor that they don’t have their own institutional constraints. There’s unlimited throat clearing one could do to explain all of that in detail. But it is not a mighty legislature that wields the full measure of its power to get the things many of its members say they want. Congress currently suffers from a similar lack of institutional ferocity, but Albany is the bigger bee in my bonnet.

Daniel Golliher 🗽

@danielgolliher

You too, can experience me talking about city government from the floor. https://t.co/d5L0UvROEe https://t.co/GfRKqan18L

Daniel Golliher 🗽

@danielgolliher

I’m planning a “How to be a good political party member” seminar via Maximum New York for the fall. Participants will be Democrats and Republicans. DM if you are a party official who likes what I do, and who wants to speak/have ideas. This post is a preview of how I approach party issues: https://t.co/yTL4aCcRau

Daniel Golliher 🗽

@danielgolliher

It’s going to be a great autumn https://t.co/qIU5R1B1RN

Daniel Golliher 🗽

@danielgolliher

I'm rereading @patio11's "The story of VaccinateCA" in @WorksInProgMag. It's well worth anyone's time, but this paragraph is, from a civic-spiritual perspective, fundamental to how I see many things: https://t.co/vRDG7UOmcR

Daniel Golliher 🗽

@danielgolliher

I think Nicole has the right view of Moses, and she lays it out well in her book, Movement, which you should buy and read! https://t.co/jFovoOFgV3 https://t.co/grMvtPbSbh

Daniel Golliher 🗽

@danielgolliher

“Choo choo choo” is like the Transatlantic accent of trains. It’s the way they talked on the silver screen, has a certain verve to it, but generally doesn’t reflect modern train speech. https://t.co/q4jaNTI2gY

Daniel Golliher 🗽

@danielgolliher

What this @axios piece doesn’t mention is that in 2016, according to the same survey*, only 1 in 4 could name the three branches. We’re up to 2 in 3. We’re way up! @dsallentess wrote about this in the NYT earlier this year. https://t.co/LwZwwmmstN https://t.co/vjh2mKuJL9

Daniel Golliher 🗽

@danielgolliher

I agree, @jasminewsun's piece (linked in QT'd thread) is great. The policy response across states has not been uniformly responsive to real concerns, but many of them are being reflected in policy across the country. Although: even if you're a policymaker or a data center fan, you still have to communicate well that concerns are truly being reflected in policy! One particular challenge is protecting the public against downside risk, (like the "Foxconn fiasco" described in the piece)--a project running up a tab of investment and expectation, which people and governments plan around, and then pulling out. Good policy can mitigate this by: (1) setting up selection effects that screen out less stable projects, and (2) requiring deposits and other terms that pay out in the event that a project does materialize, or terminates too early. In my recent piece on data centers connecting to the grid, I wrote about how Texas addressed these concerns (see screenshot below for an example), and compared it to New York, which is not as far along in the policy process. @Prigoose -- data center policy event at @fractaltechnyc??

Daniel Golliher 🗽

@danielgolliher

Here are my brief, live notes on the first ~5.5 hours of today's NYC Council AI hearing. I will have to catch the rest of the hearing asynchronously. https://t.co/bvWp92Zbdc

Daniel Golliher 🗽

@danielgolliher

If you are a writer in NYC (policy/substack/state capacity/abundance) who is interested in writing about reforming public sector labor as part of the broader "reform civil service" push, get in my DMs. We have schemes to do.

Daniel Golliher 🗽

@danielgolliher

A lovely blog from a cohort 15 alumnus, on the course requirement of memorizing the poem “The New Colossus” https://t.co/wAykL4hnjt https://t.co/GfRKqan18L

Daniel Golliher 🗽

@danielgolliher

People are saying https://t.co/T0Mx43bDAC

Daniel Golliher 🗽

@danielgolliher

@zdch Has anyone written about this??

Daniel Golliher 🗽

@danielgolliher

This is creativity in government hiring, love to see it. https://t.co/nxIAG3GPAr

Daniel Golliher 🗽

@danielgolliher

Speaker Menin has gaveled in the council's hearing on AI. https://t.co/esLlkuLTlN

Daniel Golliher 🗽

@danielgolliher

NYC before post-WW2 zoning and planning: https://t.co/s2nfYn30y5 https://t.co/RJirUHsooy

Daniel Golliher 🗽

@danielgolliher

What need for a heart, when one has New York City? https://t.co/pvrYSes4s0

Daniel Golliher 🗽

@danielgolliher

One of the best ways to learn about NYC government is to read its municipal bond prospectuses ("official statements" for each bond issuance). Same for the MTA, and other bond-issuing entities. Consider: an official statement is the legal document wherein the city and its engaged counsel make *legal representations* about why it can credibly repay bonds, which involves an overview and evaluation of the political processes, demographics, budgeting, taxing power, and more that impact the city treasury. These kinds of documents mix boilerplate review language with more temporally updated, material review. Understanding which is which is vital, and is why, although the document is excellent for getting a grip on fundamental city government, it is not simple. This is perhaps my most @patio11 coded municipal opinion, although there are other strong contenders.

Daniel Golliher 🗽

@danielgolliher

Beautiful post, and also: "permanent middle class" is a wonderful sentiment https://t.co/PtK3X8dyGu

Daniel Golliher 🗽

@danielgolliher

Endorse https://t.co/NPj0GF9fRQ

Daniel Golliher 🗽

@danielgolliher

Before “you can just do things,” there was “What, like it’s hard?” https://t.co/QCKa7Z9u78

Daniel Golliher 🗽

@danielgolliher

Love these stories https://t.co/XX2TqIquCu

Daniel Golliher 🗽

@danielgolliher

Office hours are tomorrow from 8-10am. You can find the info on the Maximum New York Luma calendar. You can come if you're a current student or not, and you can come to discuss something specific, or just say hello! https://t.co/xabACfQ8Ko

Daniel Golliher 🗽

@danielgolliher

Foundations of the Liberal Arts, Cohort 3, applications will open this week. It will start in about three weeks, and will be Sunday evening to give you a refreshing start to the week. https://t.co/GfRKqan18L

Daniel Golliher 🗽

@danielgolliher

I remember when I really internalized Cartesian coordinate systems in 7th grade, and realized what they represented. I think in terms of them every day, but they've faded so far into the furniture of my thought that I sometimes forget that they were once an innovation, both to the world and to me. Working with them casually has gotten so much easier and prettier with AI tooling.

Daniel Golliher 🗽

@danielgolliher

Yes https://t.co/t8iqGZKaM2

Daniel Golliher 🗽

@danielgolliher

Learn AI, meet pals, grow into the future! https://t.co/AYj7sBZYz9

Daniel Golliher 🗽

@danielgolliher

This cohort is going to be bonkers. https://t.co/GfRKqan18L

Daniel Golliher 🗽

@danielgolliher

She was one of the greatest Americans, and her life will endure as a great hope and inspiration to all of us. I grew up with her music, and have referenced her many times in my own writing. Dolly was a true model of how to live, across any domain you could think of. In business, family, culture, politics, philanthropy, and more, she was the best of us. I will miss knowing that she walks on the same world as me, but she will always be in my heart, and the hearts of millions. I will always love you, Dolly.

Daniel Golliher 🗽

@danielgolliher

For my friends, paying the fare. For my enemies, the saloon-style entry gates. https://t.co/RnREAAsPxA

Daniel Golliher 🗽

@danielgolliher

The @nycibo published an updated "city budget in $100" overview for the budget passed in June. The full report, linked at the bottom of the thread, has answers to questions like "what's in Miscellaneous"? These are "budget categories," rather than agency-by-agency. https://t.co/oZEvs29dGo

Daniel Golliher 🗽

@danielgolliher

This is a great report that highlights the hard realities of our power needs: (1) Both the red and blue teams have goals that require far more generation, (2) we’re about to retire a lot of coal generation, and (3) it will take a diverse energy generation portfolio—including natural gas—to get us where we want to be. “All of the above” energy is the future!

Daniel Golliher 🗽

@danielgolliher

Stripe Hog https://t.co/L03WHyUJQW

Daniel Golliher 🗽

@danielgolliher

“NYC's unserious, lopsided general elections after closed primaries no longer make sense for anyone” Co-sign https://t.co/fHIoou1moU

Daniel Golliher 🗽

@danielgolliher

I would be happy to pay much more in federal taxes if I had good assurance that they would go toward closing the national deficit within a year or two, and then tackling the national debt. Like, if there were a “close the deficit by fully funding the legislatively enacted budget, not weaseling out of trade-offs by debt financing/future taxes” tax and a “reduce the debt” tax, I would be satisfied to pay them. But the government would have to commit to wiser fiscal management to effectively pull that off!

Daniel Golliher 🗽

@danielgolliher

From Asimov’s “I, Robot” (1950). A lot of societal ideas packed in two short paragraphs, more relevant than ever. https://t.co/izw6yfaij3

Daniel Golliher 🗽

@danielgolliher

Excelsior https://t.co/Z7k212Wvph https://t.co/Yyum4CJXWc

Daniel Golliher 🗽

@danielgolliher

Underrated point https://t.co/HgL0pFzPtK

Daniel Golliher 🗽

@danielgolliher

Catch me in DC! https://t.co/VECj62sQmy

Daniel Golliher 🗽

@danielgolliher

One of my niche views on NYC governance: the public advocate would ideally be a Republican if the other two citywide offices (+ the Speaker) were Democratic. The PA is supposed to operate as a third citywide, ombudsman voice. It would be helpful toward that end if it were more formally adversarial.

Daniel Golliher 🗽

@danielgolliher

This is an example of what to do https://t.co/uSHc4NFRmZ

Daniel Golliher 🗽

@danielgolliher

Who is in charge of New York's cybersecurity? https://t.co/GbTyRkCPub

Daniel Golliher 🗽

@danielgolliher

"Who should be the third party validators." - Q from Menin "Whoever is qualified and competent to do the evaluation." - A from Coxon "You can have more than one third party validator..." - A from Kokotajlo "California has SB53...it's not clear whether this would have required any reporting on Hugging Face..." - A from Turner "Should the standards for [SB53 and NY's RAISE Act] be lower before they kick in?" - Q from Menin Yes - A from Turner

Daniel Golliher 🗽

@danielgolliher

Same kind of deal when I got my license updated. Smooth, efficient, excellent customer experience. https://t.co/LEEkSDrVoS

Daniel Golliher 🗽

@danielgolliher

It is finally clear that autumn is upon us, so it’s important to emphasize: Emily is the superior Gilmore Girl.

Daniel Golliher 🗽

@danielgolliher

The use of AI is everywhere in my life, but not in a way that makes it feel smaller. I also recently used it to help me figure out why I suddenly got blisters while running (despite being a seasoned runner with nothing obvious going on). In this particular use case, it has directly supplanted search engines, which would have been my first stop before. But also: I use it in a different way than search engines. AI is both search and a "rubber duck" (https://t.co/4B8NKIgayu) at once. Anyway, I notice easier problem solving in general for everyday things. (And all of this sits side-by-side with things like the HF attack; that's 2026.)

Daniel Golliher 🗽

@danielgolliher

This particular post is at odds with the governor's expressed interest in figuring out a way forward with data centers. The governor went on @NY1 just a few days ago and laid out an approach to hyperscale data centers (which she defines as >50MW in her executive order banning them for the next year) that was thoughtful, even if you don't ultimately agree with it (see: https://t.co/DfYjGVFIJP). Her appearance on the Odd Lots podcast earlier this summer with @TheStalwart and @tracyalloway was also thoughtful, and full of pro-progress points that were widely applauded: https://t.co/ylreTw3mLQ And her use of AI (powered by hyperscale infrastructure!) to improve state law shows her willingness to embrace the new technology and its requisite physical infrastructure to improve government (https://t.co/9cbOKJ7ZsF). The president's post, even if not the way I'd put it, is correct. And the governor's approach to data centers and AI has been thoughtful, even though it isn't exactly what I would have done. This post seems to work against all of that. Data centers are an opportunity for New York and the nation.

Daniel Golliher 🗽

@danielgolliher

I will sharpen the point a little more: “Natural gas is bad” is a new “developers are evil” front in abundance-land. Significant work remains ahead to change minds and memes. https://t.co/WtvBG5Chwy

Daniel Golliher 🗽

@danielgolliher

What a time to be alive https://t.co/BkuvtAVY7c

Daniel Golliher 🗽

@danielgolliher

I am suspicious of a person who smokes in public spaces (especially more acrid marijuana), who also decries air pollution more generally. “Externalities for thee, not for me!”

Daniel Golliher 🗽

@danielgolliher

Broadly speaking: help peer

Daniel Golliher 🗽

@danielgolliher

SF: August 6-8 DC: August 19-20

Daniel Golliher 🗽

@danielgolliher

People are saying: https://t.co/36icrauYNu https://t.co/fvp7qbxljG https://t.co/payxvCd53Y

Daniel Golliher 🗽

@danielgolliher

@CarlMichaelW asks: what guardrails on misleading advertising and truthful disclosures do you wish your competitors had to abide by, and where do you think consumers are likely to be misled about AI tools' usage and risks. @OsseChi asks if any of the companies will contribute free compute to NYC (indefinitely) to harden the city's defenses, and whether they will commit anything today. He aims the question specifically at Anthropic's Graham. Graham says Anthropic is already working with NYC and NYS on this front. OpenAI's Dwyer highlights their Daybreak for Defenders $1B initiative to give subsidized access to models for defense, and highlights that OpenAI is already working with NYS and PANYNJ. He then asks if the companies support the federal government preempting localities and states on AI safety regulation. OpenAI's Dwyer says they support sub-federal *harmonized* legislation to raise the floor on frontier safety. @OsseChi then says that their companies/contributions lobbied against political candidates who support safety legislation (as far as this claim goes, I would point out that it is mixed, and AI company contributions go both ways, as manifest in the recent NY-12 race with @AlexBores). Meta's Cahill asserts Meta's support for a uniform federal standard, as does Google's Friend, who says that it's also perfectly fine for state's to legislate for AI, because many relevant laws are at the state level anyway (one could potentially think of torts, I would note). Anthropic's Graham says they were the only company on the panel to support CA's SB53 prior to it becoming law, which was the basis for NY's RAISE Act, and that they also support a strong federal framework.

Daniel Golliher 🗽

@danielgolliher

There is a consensus that is trying to be born here. “All of the above” energy is nominally endorsed by New York’s Democratic governor and Republicans in the Assembly. It is the way forward for New York. There is a left/right fusion that can be achieved here, and I am happy to talk to whoever is interested in meeting in the omni-partisan middle. DM me.

Daniel Golliher 🗽

@danielgolliher

Applications close this evening! It’s a wonderful time to accelerate into gov/politics 🗽 🇺🇸 https://t.co/GfRKqan18L

Daniel Golliher 🗽

@danielgolliher

Applications close Friday https://t.co/GfRKqan18L

Daniel Golliher 🗽

@danielgolliher

Published in the 1990s, still holds up! https://t.co/nkaX5JDrMq

Daniel Golliher 🗽

@danielgolliher

Important lens to have when observing political actors https://t.co/TEuhBlggqn

Daniel Golliher 🗽

@danielgolliher

🫵 go study https://t.co/KtgYxSBHvK

Daniel Golliher 🗽

@danielgolliher

Well I know how I'll be spending at least a few hours this week https://t.co/2j1MAttP5X https://t.co/wacLo7UpFS

Daniel Golliher 🗽

@danielgolliher

@NYCMayor I don’t think this post leads New Yorkers to a more truthful view of our political economy: https://t.co/piLSp7hWkN

Daniel Golliher 🗽

@danielgolliher

Up my alley https://t.co/abalUQnnQM

Daniel Golliher 🗽

@danielgolliher

I noticed that she quoted the book "Empire of AI" by Karen Hao in her opening remarks. This is the book that seems to have kicked off the "AI uses huge amounts of water" issue, which was based on an order-of-magnitude error she later corrected after @AndyMasley caught it.

Daniel Golliher 🗽

@danielgolliher

https://t.co/K0e3UVsbGH https://t.co/6YSI2M7lbj

Daniel Golliher 🗽

@danielgolliher

Points like this are missing from the Flock discussion. Technology can be built in a wonderfully pro-social way. This doesn’t mean that it’s perfect, but it can be a massive improvement over the status quo. https://t.co/PcWSXmKTx9

Daniel Golliher 🗽

@danielgolliher

“Compromise” vibes: watering down an ideal “Common promise” vibes: building up mutual ties that bind

Daniel Golliher 🗽

@danielgolliher

@davidshimel @fractaltechnyc Spotted: @__drewface, @liamdanielduffy, and These United States https://t.co/lejROMAFKJ

Daniel Golliher 🗽

@danielgolliher

The MNY piece: https://t.co/P3OXg6Q4AU

Daniel Golliher 🗽

@danielgolliher

Good lens. Government ought to be able to execute, or things proceed off the rails. https://t.co/B0AQUxcmqu

Daniel Golliher 🗽

@danielgolliher

I just follow orders. Godspeed, @AlexTISYoung. https://t.co/bBtmQdtTia https://t.co/xcedUpB4fj

Daniel Golliher 🗽

@danielgolliher

https://t.co/zSjmK5nNQR https://t.co/yXdtTHylTM

Daniel Golliher 🗽

@danielgolliher

This post was partially motivated by @sebkrier’s recent post (https://t.co/mkLClLrCRB). After Monday’s City Council hearing on AI, it seems clear that many advocates and lawmakers are getting pulled into a narrow, overly abstract way of thinking about AI. But we can be more concrete and productive than that. The goal of my post, which is just the first on New York’s cybersecurity, is to pull people in a more concrete direction.

Daniel Golliher 🗽

@danielgolliher

A better version of the @NYCCouncil hearing on AI yesterday would have incorporated many of these ideas. https://t.co/kt64KHwkDn

Daniel Golliher 🗽

@danielgolliher

@ShahanaFromBK refers to the February story of AI precipitating a missile landing on an Iranian school, and asks about their processes for tracking and stopping that kind of action (https://t.co/dWcylcAdZj). @CMAlexaAviles asks if the companies sell products directly or through intermediaries to ICE or CBP. And she asks for agreements those companies do have with those agencies. She also asks if the companies will commit to not using NYC resident data for law enforcement or immigration enforcement actions. She asks if any company audits "downstream law enforcement use" of their products. She got universal pledges to follow up with details.

Daniel Golliher 🗽

@danielgolliher

This comes out when reviewing park events on https://t.co/9xjg6jkO7c. https://t.co/7pnuS3LWg5 https://t.co/sfrE4WAvH2

Daniel Golliher 🗽

@danielgolliher

“Why the need for DEFRs? Because of the escalating demands of electrification of buildings and vehicles in order to meet CLCPA targets on top of the 2040 100% emissions-free power mandate. It was not about data centers or any other large loads, and they're still needed for those same reasons.” ☢️

Daniel Golliher 🗽

@danielgolliher

"Terraform is hiring!" is an opportunity klaxon https://t.co/rgIjOVwYpx

Daniel Golliher 🗽

@danielgolliher

“All of the above” is abundance, both in housing and in energy. I think NYC’s governing establishment has come a long way on this for housing, but still needs to learn the same lessons for energy. https://t.co/FFp5E9KJNc

Daniel Golliher 🗽

@danielgolliher

I agree completely. To save and productively change America, one must love it. I think the same is also true of political parties. Like the nation, it is possible to change a party while actively hating it. But that’s a bad path. Be a “party man” or “party woman”! Love your party. If you cannot find that within yourself, I would say that is more probably an issue on your part, as with those who cannot find a path to loving America. Both major parties have a lineage worth evolving well into future, even if one ultimately chooses one over the other.

Daniel Golliher 🗽

@danielgolliher

Run, don’t walk https://t.co/LKkRVYjzGo

Daniel Golliher 🗽

@danielgolliher

Agree https://t.co/rGmTG0db8d

Daniel Golliher 🗽

@danielgolliher

FYI https://t.co/LnnGrF5JqC

Daniel Golliher 🗽

@danielgolliher

One day they will get their subway link to the other boroughs! https://t.co/VJpMgkGyJe

Daniel Golliher 🗽

@danielgolliher

@CMLindaLee highlights the potential benefit of using AI to coordinate agency actions together to reduce the work burden to them. But, she says, how do we ensure unnecessary information isn't collected, and highlights the Trump administration's "public charge" rule change as a example of what could happen with increased data collation (that is bad, in her view). The company reps all assert their user privacy frameworks are robust, and that they care about privacy/data privacy/user security. Anthropic's Graham goes a bit further to say that their user policy has restrictions on how law enforcement and others can use that data. @vmmaloney asks how the company's quantify improvement/decline in model safety, and what counts as an incident. My summarized response: the industry and others are still debating what counts as an incident and when to report it--one doesn't want overreporting or underreporting.

Daniel Golliher 🗽

@danielgolliher

Next the Speaker asks how many times any company's agent has escaped their sandboxes, or if there any incidents that haven't yet been reported to the public. None of the company representatives gave a number here, or a firm yes/no on the non-public question. Anthropic's Graham says that Anthropic releases these things when it makes sense on the basis of security and technical understanding of what happened. Meta's Cahill says he's not personally aware of any incidents, but the Speaker asks if the *company* is aware, rather than just him personally, and he says he will follow up. Google's Friend answered in the same way as Cahill, with the same response from the Speaker. The Speaker then asks the company's reps to commit, live, to not releasing any model that fails an audit by a third-party evaluator the company didn't select. She says she's looking for a simple yes/no. She does not get any simple yes/no answers; all company reps assert that they don't release models they believe are unsafe according to their own internal processes. Menin calls this "equivocation."

Daniel Golliher 🗽

@danielgolliher

Next the Speaker asks the company panelists if they believe their companies should be held legally liable for any rogue AI actions. The speaker is looking for a yes/no answer here. No company rep gave the yes/no answer; various of them reiterated their company's dedication to safety, and that this question is too complex or nuanced for yes/no. "Existing legal frameworks do apply to AI...we've long said if it's illegal without AI, it's still illegal with AI." - Google's Friend gets the closest to the kind of answer Menin was looking for.

Daniel Golliher 🗽

@danielgolliher

"Can we teach [AI agents] a moral code?" Q from @CnDelarosa. "We don't have the science for this yet." - A from Coxon "In principle, it probably is possible...but we are very far from achieving this in practice." - A from Kokotajlo "I think it is quite possible, even as someone who is relatively optimistic, we can look out and see the recklessness of these companies...[but] we have a really good chance of not solving it." - A from Turner

Daniel Golliher 🗽

@danielgolliher

Autumn is the most millennial season. Ideal for: - Gilmore Girls rewatch - Stomp clap music

Daniel Golliher 🗽

@danielgolliher

1975: - Fiscal crisis - Birth of ULURP - Passage of SEQRA - Banned pay toilets 1975 is the "what happened in 1971" of New York City/State https://t.co/p6tp2U6UpF

Daniel Golliher 🗽

@danielgolliher

Excellent https://t.co/75NFvzVp46

Daniel Golliher 🗽

@danielgolliher

@GarettJones and @akoustov are excellent follows generally, but if you want honest truth-seekers when it comes to the perennially thorny topic of immigration, they’re your guys. https://t.co/GYTmE50XfE

Daniel Golliher 🗽

@danielgolliher

If you want to read the City Council's committee report for today's hearing on the Pied-a-Terre rollout, I will post the 12 pages as images in this thread, followed by a link to the report. https://t.co/oxUQCllg4h https://t.co/sT5SXZmeXU

Daniel Golliher 🗽

@danielgolliher

Excellent https://t.co/Ur0dRK8lfb

Daniel Golliher 🗽

@danielgolliher

It is wonderful to have vision and pragmatism at the head of NASA. https://t.co/EzdagJmFrp

Daniel Golliher 🗽

@danielgolliher

I’ll allow it https://t.co/0JcGdn0yXJ https://t.co/bOPWdB68cL

Daniel Golliher 🗽

@danielgolliher

It is midterm season, but you gotta have the @NYCCouncil AI hearing up on the big screen during (currently @AlexBores speaking) https://t.co/edefLdQKA1

Daniel Golliher 🗽

@danielgolliher

@nycdistrict1 asks what legislation the companies are against on a state or federal level? None of the company reps answer the question, they just say what they have supported. No surprise there. @kaymsanto asks Anthropic's Graham how many people are on his red team (he says like 25 on his team, but several hundred people overall are working on different pieces of the catastrophic risk problem). She asks if he feels this is sufficient. He says "yeah...and we need to continue doing more." She then asks the staffing question to OpenAI, the rep says she will have to get back with the council. I will note: the amount of council members in this hearing continues to drastically dwindle, and we are still on panel 2!

Daniel Golliher 🗽

@danielgolliher

Personally, I could sit through a 40-hour @NYCCouncil hearing and find it fascinating the whole time, both because I do like many of the members, and because legislatures are one of my special interests. But legislative hearings, as an instrument, have specific advantages and disadvantages. People take time out of their day, and time off work, to attend and pay attention to these things. Council members and staff should get as much answered on the record or via research as they can *prior to* hearings, and use the hearings to pursue anything lacking after that work. This allows the hearings to be shorter, makes their content punchier and higher value per capita, and prevents repeat questions that should be aggregated beforehand. I believe that Congress has been doing a good job with their recent AI hearings, and provides an example that the Council could learn from.

Daniel Golliher 🗽

@danielgolliher

Now the opening statement from former OpenAI researcher Daniel Kokotajlo (@DKokotajlo), who just gave testimony at a Senate subcommittee hearing on AI last week (Sept 30). Last week's hearing: Subcommittee on Disaster Management, District of Columbia, and Census "Rogue AI: Securing the Homeland Against AI Agent Attacks" https://t.co/yJQsIudm5b

Daniel Golliher 🗽

@danielgolliher

Today in @politico, the president’s admin via @jacobhelberg shows alignment with the top cyber official in New York State, Colin Ahern, on the nature of foreign threats. https://t.co/UV47vJAqAW https://t.co/2GkvKPrTQO

Daniel Golliher 🗽

@danielgolliher

The status quo hasn't been the status quo for very long! We can get back to functional rent collection. https://t.co/jIOo7cYNWw

Daniel Golliher 🗽

@danielgolliher

Superb https://t.co/H6KyKNF3xK

Daniel Golliher 🗽

@danielgolliher

One of the marvelous things about America is that it’s enormous, and many kinds of people want to live in different kinds of ways. Rural, suburban, urban. It can all be productively unlocked with prudent policy to allow us all to pursue our version of the American Dream. https://t.co/61SbyKywM9

Daniel Golliher 🗽

@danielgolliher

Yes! And also: it was delivered in 1893. Essentially every online source, and every published physical book I could find, gets the date wrong and pins it at 1883. Even the @PBS "original source archive"! I wrote a post about how I tracked down and verified the correct date: https://t.co/UvGiLBCUSJ

Daniel Golliher 🗽

@danielgolliher

Putting it out there if you're doing fall programming: I would be interested in doing talks, debates, etc on New York's CLCPA (Climate Leadership and Community Protection Act, 2019) implementation. I'll definitely be doing my own talks about this, but the more the merrier!

Daniel Golliher 🗽

@danielgolliher

There are more locations for nuclear plants than we think. @GovKathyHochul ☢️🗽 https://t.co/8YeCsfrPLZ https://t.co/sz1fIAegGv

Daniel Golliher 🗽

@danielgolliher

And now the opening statement from Alex Turner (@Turn_Trout), formerly of Google Deepmind. Speaker Menin confirms "the Council has issued a subpoena for Mr. Turner's appearance." https://t.co/QxTk8pAvQx https://t.co/37YPAoPWNS

Daniel Golliher 🗽

@danielgolliher

Go get these jobs! https://t.co/3UqvZs54tW

Daniel Golliher 🗽

@danielgolliher

https://t.co/6uAxmWYheV https://t.co/wvcc9GfjAy

Daniel Golliher 🗽

@danielgolliher

Embrace your power, citizen! https://t.co/TgAQty7LwO

Daniel Golliher 🗽

@danielgolliher

In today's episode of the All In podcast, @NASAAdmin said: "We are living through an extraordinary moment in history, aren't we? It took just 65 years from Orville and Wilbur's first flight to Neil and Buzz walking on the surface of the moon. Now, it's been 57 years since Apollo 11, and for a while there the pace of progress hasn't been all that inspiring--but look at what is on the horizon right now. We've got artificial intelligence, quantum technologies, robotics, additive manufacturing, the promise of fusion energy, biotech to heal the disabled, autonomous transportation, this whole abundance thing everyone is talking about--all converging in the years ahead, is beyond what most people can comprehend." (0:49-1:33) Note: Abundance mentioned! cc @DKThomp, @ezraklein Can't wait for Space Abundance

Daniel Golliher 🗽

@danielgolliher

Re: Citizens United, people will talk about big/dark/corporate money in politics. All fair game for discussion. But one point about the case often gets lost in the wash--there was another kind of entity that had its expenditures limited by law.

Daniel Golliher 🗽

@danielgolliher

Go write! https://t.co/2plwpwdhcM

Daniel Golliher 🗽

@danielgolliher

The Chinatown birds will absolutely steal your dumplings if you’re not careful https://t.co/6xGHwfa36V

Daniel Golliher 🗽

@danielgolliher

https://t.co/ECDjxjq4PV https://t.co/Djvmw8J5uN

Daniel Golliher 🗽

@danielgolliher

You can learn many things here: https://t.co/lp4elFI3zu

Daniel Golliher 🗽

@danielgolliher

What interesting times we live in https://t.co/IIlDf00fd8

Daniel Golliher 🗽

@danielgolliher

Also: https://t.co/09vIMjT9AU

Daniel Golliher 🗽

@danielgolliher

@frankmorano is up next, now with a "one-minute" limitation, due to the hearing dragging on (my note, as above: that is because it was poorly planned. My second note, the one-minute limit will not be enforced, because that is just simply not enough time to ask questions here). "Does your company believe that there is any non-zero chance that advanced AI could escape human control and cause catastrophic harm?" OpenAI's Dwyer -- "I don't know..." (got cut off) Google's Friend -- "Do we entertain the possibility...yes." Meta's Cahill also says they model that possibility. CM Morano asks whether economic pressure might override safety concerns on pressing forward, and whether chatbot logs can be used to target advertisements without affirmative user consent. @galeabrewer has three questions: (1) on third party evaluations, if a model doesn't pass, how does that impact whether/how it is released? (2) on China--what information are they getting from US companies? (3) Will the companies commit to not helping ICE at all "We should not be shipping immigrants out of this country at all." @CMMercedesCD46 asks if any of the companies sell products to NYC hospitals, and says "people need to be diagnosed by a human." She wonders if AI is being used properly for diagnosis and adds a remark about AI changing interactions with insurance. OpenAI's Dwyer says that doctors should remain in charge, that AI shouldn't hold itself out as a licensed professional, but that it can be very helpful, especially with doctor administrative tasks. Anthropic's Graham shares a personal story about using Claude at 1AM to help triage an eye problem that led him to (properly) go get medical care. This is an increasingly common story! I have had this kind of moment myself, as have my friends.

Daniel Golliher 🗽

@danielgolliher

The hearing is now taking a 5-minute break. One wonders exactly how long this is supposed to take. I will note: while the hearing started off brisk when only the Speaker was addressing witness panels, it is slowing as the questioning is fragmented among members, and the company reps often answer with pledges to get back with the council members with more details. This hearing is starting to resemble the e-bike hearing as described by @nicolegelinas: https://t.co/SPH71U8VTy

Daniel Golliher 🗽

@danielgolliher

@CnDelarosa asks about the meeting that the AI companies had with the president last week, and the agreement they signed to voluntarily bind them to certain standards. She asks what exactly they agreed to. OpenAI's Dwyer highlights legislation that OpenAI supports, like mandatory federal pre-deployment review, and independent technical assessments of their models. Meta's Cahill says they also agreed to independent review of their models in the federal accords. Google's Friend answers similarly, regarding independent reviews. And she highlights that the end of the accords say that the voluntary commitments may need to become law, which Google supports.

Daniel Golliher 🗽

@danielgolliher

@AndyMasley Jacob Coxon is the first to give testimony on the "whistleblower panel." https://t.co/29Fyzx0MUm https://t.co/CLSgePubnL

Daniel Golliher 🗽

@danielgolliher

I’m watching this live view from orbit, on the phone in my hand, on the subway going into Manhattan. What a time to be alive! https://t.co/hUTNLYOXQT

Daniel Golliher 🗽

@danielgolliher

I do think cybersecurity (which is the principal point of intersection with AI that many people are concerned about) is still a fundamentally bipartisan arena. https://t.co/3jMQkeWg1D

Daniel Golliher 🗽

@danielgolliher

Starbase municipal government seems like an adventure for the right kind of person https://t.co/qW6J1kbBSR

Daniel Golliher 🗽

@danielgolliher

@RichardMorrison I asked dear sweet ChatGPT https://t.co/N0Wy2EiPUH

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/PjIqHChFXM https://t.co/mCPyUYDEaN

Daniel Golliher 🗽

@danielgolliher

@__drewface Related: https://t.co/UFXBg3wDdY

Daniel Golliher 🗽

@danielgolliher

@aarmlovi cc @__drewface this is a post we could collaborate on

Daniel Golliher 🗽

@danielgolliher

It is indeed a political nerd Christmas. In my gov class, we cover c3/4s (and the whole 501c chunk of the Internal Revenue Code) towards the end, and the questions and discussion are always great. https://t.co/2KAEi5G2qz

Daniel Golliher 🗽

@danielgolliher

Cohort 1, the Ur-cohort https://t.co/Fo3aIcd51q

Daniel Golliher 🗽

@danielgolliher

From @ahall_research and @danmthomp: https://t.co/asOE0mo5Im https://t.co/MdlnSBTr7x

Daniel Golliher 🗽

@danielgolliher

@CMPiSanchez is next. She asks: What updates have the products made to protect children, and to limit their practice of medicine (she references this story: https://t.co/z8nlhkCjTq)? And also: how do the AI products currently use info shared with them by users (like for advertising), how do they use any kind of proprietary data put into them, and who are they paying for the data they receive? Again, I'll point out that this ground has mostly already been covered. OpenAI's Dwyer opened by *once again* discussing ChatGPT for Teens.

Daniel Golliher 🗽

@danielgolliher

@CnDelarosa I will note that the chamber does not seem friendly to the AI companies. Their audible reactions are often angry/incredulous laughing in response to AI company reps, and laughing *at* them in response to council member responses.

Daniel Golliher 🗽

@danielgolliher

"What percentage of New York's workforce do you estimate will be replaced by AI over the next five years?" Q from the Speaker. OpenAI's Dwyer says her company not only wants productivity improvements, but to help workers. Menin cites a BCG report that ~20% of workers could be displaced by 2030 (I think I caught that it was 2030), which would be 600k New Yorkers. Anthropic's Graham cites their economics team's work (https://t.co/E1z1DrWiBM). He doesn't give an exact answer to the Speaker's question, but Anthropic has definitely given robust answers in their economics analyses (in my view). They don't confidently forecast an exact job loss figure for NYC, but their analysis also shows why that is difficult.

Daniel Golliher 🗽

@danielgolliher

"As you are all under oath, we want to hear how you quantify the [risk from AI] in the worst case possible scenario." Morgan Dwyer - "I don't know, and it* doesn't matter...we should not train models [that have a chance of catastrophic events]." *she clarified that this means putting an exact quantity on it, because any number is too much. Speaker Menin: "Honestly, I'm incredulous at that answer." Speaker Menin then asks about some OpenAI employees who were terminated, and whether that had anything to do with their views on AI safety. Dwyer says "I don't know." She then turns to the other panelists for an answer to her quantitative question. None of them provide the kind of answer she's clearly looking for, which seems to be "You say you want to be more transparent, but you're not really answering the question." --Menin says in response to no company representative giving a quantitative answer to her question. "There is no reference class on which to base quantitative probabilities...there isn't really a rigorous scientific way to do that yet." - Google's Alice Friend. I think she had the best short response to the difficulty of giving an exact probability, but clearly the speaker and others still want a more exact probability, and think this is necessary. @logangraham also didn't give one exact quantified answer, but he explains that Anthropic does release their own detailed analysis of risks regularly. Which is true! This risk report from August 2026 (https://t.co/Ml1SNGjqMN) lays out how Anthropic thinks about the risk from its tools, and also reveals how hard it is to give one-risk-number-to-rule-them-all.

Daniel Golliher 🗽

@danielgolliher

Panel 1 of 14 is concluded https://t.co/MhcLWw0cNR

Daniel Golliher 🗽

@danielgolliher

"move on from past partisan hobbyhorses" is a good rule of thumb for what the moment requires. https://t.co/UGtzXRuO8H

Daniel Golliher 🗽

@danielgolliher

v2 iteration https://t.co/ltAq27XIg1

Daniel Golliher 🗽

@danielgolliher

@trillionjohns Good example of this kind of thinking: https://t.co/BxnPJs8A9h

Daniel Golliher 🗽

@danielgolliher

Bless the chaos monkey! https://t.co/vBfDKvWWk0

Daniel Golliher 🗽

@danielgolliher

Get writing 🤖 https://t.co/ZFw9tRh9n2

Daniel Golliher 🗽

@danielgolliher

Get ready https://t.co/IoZFvw35mo

Daniel Golliher 🗽

@danielgolliher

@__drewface @lorenz_system I even QT’d it

Daniel Golliher 🗽

@danielgolliher

@emhuntersmith 🫡

Daniel Golliher 🗽

@danielgolliher

Neat! https://t.co/ITBJ5d4P0P

Daniel Golliher 🗽

@danielgolliher

Speaker Menin says that the Council didn't get all the answers it wanted from company representatives, and that they will send questions to them that they want answered in writing. We now move on to panel 3 of 14, the city administration. https://t.co/xZjfWyWQjp

Daniel Golliher 🗽

@danielgolliher

Question from @KamillahMHanks: her bill will require 24-hour reporting to the city's cyber command (NYC3). She asks a good question--is this too soon? Can they realistically report if the investigation is still ongoing? This was also the question that came up in the 2018 council tech committee consideration of intro 1153 (that bill did not pass); the then-head of NYC Cyber Command, Geoff Brown, said "As currently written, the bill would require us to divert resources from responding to an attack to brief the City Council, in a manner outside of the traditional hearing and oversight processes outlined in the Charter, and it may force the public exposure of information that would make the City an easier target of cyber attack." The AI company reps all reiterate their commitment to reports, and stress that reporting needs to happen to the right people at the right time, which is variable. If I were to read in between the lines: the 24-hour requirement is not a good idea, and disclosure needs to follow the specifics of incident context. This is aligned with the 2018 NYC3 response. At the end of CM Hanks' questioning, two hours in, the council "committee of the whole" has noticeably thinned out.

Daniel Golliher 🗽

@danielgolliher

Next the Speaker asked if any company carries insurance against catastrophic events. None said yes, but Google's Friend said she'd consult the company's lawyers and get back with the council. Next the Speaker asks if the company reps can confirm: "AI agents will always comply with the safety guardrails imposed upon them to secure against catastrophic harm." Again, she's looking for a yes/no. Again, she did not get any yes/no answer. "To promise perfection would not be possible for any product on the market, but our goal is to get as close to that standard as possible." -Google's Friend. Friend seems to be giving answers that get closest to what the speaker wants, but still clearly aren't what speaker wants exactly.

Daniel Golliher 🗽

@danielgolliher

Good thread on a vital issue https://t.co/Y4mCCtQGd2

Daniel Golliher 🗽

@danielgolliher

Regarding the United States: “Our democracy” but never “the democracy.” On the other hand, “our republic,” but often “the republic.”

Daniel Golliher 🗽

@danielgolliher

Great read. https://t.co/xpKRcQKsC4 https://t.co/5YDWH5kBQX

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/CRAR5q2DrV

Daniel Golliher 🗽

@danielgolliher

https://t.co/ARcm5RvwNT https://t.co/jvKH7SiUCw

Daniel Golliher 🗽

@danielgolliher

@trillionjohns @tasshinfogleman 👀👀👀👀

Daniel Golliher 🗽

@danielgolliher

@tasshinfogleman @trillionjohns 👀 👀

Daniel Golliher 🗽

@danielgolliher

@RichardMorrison Perfect meat-to-meat-to-meat ratio

Daniel Golliher 🗽

@danielgolliher

Good science fiction uses the context of different technology to examine society. I wonder if this passage from “I, Robot” is relevant—advanced machines banned from Earth, so they were put in space https://t.co/LrshopiB85 https://t.co/ypKQD6HxUK

Daniel Golliher 🗽

@danielgolliher

Many good points to contemplate here. (See next post in thread for my related thoughts.) https://t.co/8kEE2t15H3

Daniel Golliher 🗽

@danielgolliher

Neat! https://t.co/NNmMPX9GLq

Daniel Golliher 🗽

@danielgolliher

@s_sullivan @EphieBernstein His story pops up in Stripe Press's "Maintenance of Everything: Part One" https://t.co/c81emCPK5p

Daniel Golliher 🗽

@danielgolliher

Also: https://t.co/HgVYf3EDOv

Daniel Golliher 🗽

@danielgolliher

OK I will take a break here. I will have to catch the rest of the hearing later, and those notes will not be live. I have other obligations to tend to!

Daniel Golliher 🗽

@danielgolliher

Panel 3 https://t.co/0mbFbUogso https://t.co/yPK279TBGN

Daniel Golliher 🗽

@danielgolliher

@CMJenGutierrez says everyone has emphasized safety, but she doesn't feel as thought that's true, and that answers are "too nuanced" (I take her to mean, not specific enough to be useful). She asks what specific safety actions they're taking, and "make me believe that it is of the utmost importance to you all." Rhetorically, this landed well as far as I can tell. In this latter stage of the afternoon, she also kept well under her time allotment, which many other council members have struggled with. No new answers we haven't heard before, but the phrase "reiterate" came up. This hearing has heard a lot of company answers many times at this point!

Daniel Golliher 🗽

@danielgolliher

@juliej_won asks what is the exact failure rate/benchmark threshold that would get them to halt a model release, and if a model successfully generates a bioweapon design 5% of the time in internal testing, do they still release it? She also asks if they will commit to releasing redacted logs of the prompts that jailbreak their models for biorisk activities during internal testing. These are decent, targeted questions! I think @juliej_won is one of a small number of council members who have sufficiently specific questions to be useful, or at least easily answerable in a way that reveals internal processes the public might be interested in. But the council should have supported her answers by structuring the hearing in a way that would have permitted better answer acquisition, as per ongoing comments in this thread. Anthropic's Graham notes that sometimes it doesn't make sense to publicly release jailbreaking prompts, because people might try them. He also says that Anthropic releases lengthy reports that explain the nature of benchmark failure rates in bio (they do, see up thread), and that it doesn't boil down neatly as a single percent.

Daniel Golliher 🗽

@danielgolliher

@PhilWongNYC asks about RSI and AI training. He wants to know how much company code is generated by AI itself, and how much a human is still in the loop currently (do humans check all the code, for example). @CMSusanZhuang asks about the environmental impacts of datacenters, and about the impact of potentially incorrect AI answers on "security." @Harvey4council asks whether the companies think self-regulation has failed, whether more civil/criminal liabilities should exist for bad actions, what reporting turnaround time should the companies have after an incident, and will the companies collate and disclose all bad incidents to the NYC Council (as they do more robustly to the EU, says DM Epstein). Google's Friend says the existing torts system can do a lot under current law, and while I think that's probably somewhat true, the tort system probably still needs to be improved here. This was discussed in a more explicit way in last week's U.S. Senate hearing on rogue A.I. with Georgetown's Paul Ohm: https://t.co/6QFfzRe3O4

Daniel Golliher 🗽

@danielgolliher

@RitaJosephNYC asks about products sold to students, and asks about that data security. She also asks about what responsibilities the companies think they have for students entering the workforce whose work might be threatened by AI, citing a "new Center for Urban Future Report" (I assume she is talking about the CUF report cited in the committee report, which is actually from 2018, and it says that ~41% of NYC jobs could be automated within 10 years, or by 2028, but I don't believe we've seen that play out.) Note: I'm sort of skipping the AI company answers here at this point, because they are often not that substantive. They are some version of "we will get back to you with more details, and we are about the thing you mentioned, and we are working to address it." This isn't to say the companies aren't sincere or truthful/untruthful, but I don't think most people would take anything substantive from their answers. These things should have been researched and ironed out prior by the Council, although I also understand that some of them want the soundbite of them asking a basic question and not getting a certain kind of answer. I will highlight anything I think is interesting, but will otherwise do a more detailed overview later. In the middle of AI reps answering CM Joseph's question, @CnDelarosa had to warn some people in the chamber against disruptions. (Note: the council sergeant at arms will remove you!)

Daniel Golliher 🗽

@danielgolliher

@Lynn4NYC asks about biorisk. However, she says that "Most viruses in nature can be contained fairly quickly," as opposed to artificial ones that could potentially be made with help from AI. Natural viruses can be plenty troublesome, and I wonder why she said that! The company representatives all assert that they are concerned with biorisk, that they evaluate biorisk in training and release, that they watch for adversarial actors using their products for this reason, and that they are continually working on safeguards. Google's Friend: "It is very important to consider the dual-use nature" of bio capabilities, and she highlights their AlphaFold model from Google Deepmind as a good example of controlled deployment of advanced AI in bio. @CMBrooksPowers asks several questions about using AI for impersonation or furthering things like "racism, anti-semitism, and islamophobia." All the reps say that they have controls for fraud, and that they are concerned with hate and work to make sure their products are responsive to this in training and deployment.

Daniel Golliher 🗽

@danielgolliher

@crystalforbk asks what their approach is to self-improving AI (RSI), and she asks who should bear the costs for increased utility rates from AI datacenters. OpenAI's Dwyer agrees that RSI is important, and says that OpenAI is committed to paying their own way for datacenters, and highlights the tens of millions OpenAI has committed in public benefits agreements. (Note: I think infrastructure costs and public benefits are one of the most defensible aspects of AI development, but council members keep treating this question as if AI companies haven't answered it over and over with both policy, action, and committed resources. Per my previous post, this is something that should have been researched more prior to the hearing, and there are way deeper questions to ask here if the hearing were being run more properly. But the surface level question keeps getting asked and answered.) CH Hudson notes that it seems like RSI is core to everyone's AI strategy, and that she is concerned about that given the testimony at the beginning of the hearing.

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/vdfYVqA19t

Daniel Golliher 🗽

@danielgolliher

Next @EricDinowitzNYC asks about education, and the impact of AI/LLMs on the ability of students to learn to independently read and write, among other things. He asks what they are doing to prevent emotional and technical (reading/writing) reliance. He also asks about preventing AI-assisted hacking to improve student grades. OpenAI's Dwyer highlights the release of "ChatGPT for teens" for those aged 13-17, which voluntary parental controls, and says that ChatGPT is not for those under 13 at all, and she says OpenAI supports legislation to hold them accountable, like SB1119 in CA for safeguards for those 13-17. Meta's Cahill says that Meta's AI products are also only for 13+, and highlights suicide/harm resources, parental controls, and controls against sexual/romantic interactions with minors. CM Dinowitz also asks again about restricting their models for cognitive development protections (like restricting essay writing) rather than just emotional/mental protections. No definite answers. @nycpa then arrives and cuts in line. He says the opening remark from the OpenAI rep was "One of the most god-awful fantastically terrible answer that I've heard at the beginning of a hearing." He also seems to briefly compare the AI companies to Nazi medical experiments wrt how they discuss AI cost/benefit. He then asks what guardrail the AI companies DON'T want to see, and for what reason. Google's Friend makes a point: "guardrail" is a technical term with specific meaning in the AI world. But colloquially it means "policy requirements," and on those terms she shares that, as a cancer survivor, she doesn't want to slow AI development in a way that impairs advancement in fields like medicine.

Daniel Golliher 🗽

@danielgolliher

Now on to other council member questions: Deputy Speaker @CMNantashaW asks how the company's approach deep fakes/misinformation, and how they approach AI replicating artists' work. @voteshekar forcefully asks about the economic dependence that the economy, NYC's pensions, and others have on the valuations on the AI companies. I want to point out that he made several false statements and accusations, including that AI companies "steal public land for data centers." One might not like that land use, but "steal" is facially incorrect. He's the first council member who I'd say has let ideology get the better of him in his questioning. @CMKevinCRiley asks what challenges do they face id'ing malicious uses of their products, asks would stronger protections/reporting mechanisms making it easier for them to report problems, and last Q as land use chair he asks about data centers--what safeguards should companies have when putting them near residences. Reps all highlight their internal processes for reporting and cybersecurity work. Several report that they've pledged to "pay their own way" for datacenter energy, and will follow up with more info.

Daniel Golliher 🗽

@danielgolliher

Next @CnDelarosa asks if any employees have been fired for developing models that go rogue. OpenAI's Dwyer, Meta's Cahill, Google's Friend, and Anthropic's Graham say that they head policy, not HR, and don't have knowledge of that. @CnDelarosa's next question: she asks about AI literacy in communities, and she says that the AI companies have widened the wage gap and income inequality in NYC (this is an empirical claim that I'm suspicious of--I will revisit it in detail, just typing it out for now), and what the companies are doing/thinking about it. OpenAI's Dwyer says that OpenAI has a $10M training program for teachers in NYC to help them learn to use AI. I wonder how this interacts with NYC's recent one-year moratorium on most AI in its public schools! Meta, Google, and Anthropic also explain that they have made investments in AI upskilling. Anthropic's Graham highlights their economics team's work on quantifying wage and employment impacts.

Daniel Golliher 🗽

@danielgolliher

Panel two speakers (company representatives) are now giving their opening statements. OpenAI - Morgan Dwyer Anthropic - Logan Graham (@logangraham best virtual appearance award) Meta - Shane Cahill Google - Alice Friend https://t.co/fiHvDHaWMV

Daniel Golliher 🗽

@danielgolliher

Here's how all of our tax revenue plugs into those budget categories: https://t.co/amhIczvnfH

Daniel Golliher 🗽

@danielgolliher

@RichardMorrison Give him double time!

Daniel Golliher 🗽

@danielgolliher

@NYCCouncil @drewczimmerman Instant follow

Daniel Golliher 🗽

@danielgolliher

Will you listen to what the bonds are saying https://t.co/c4oXJSaO1d https://t.co/mjHiPhrOai

Daniel Golliher 🗽

@danielgolliher

@CharlesFLehman It's a very FOILy day https://t.co/tGKBI7QDKQ https://t.co/ztXplGg3hZ

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/SshZyDqu5L

Daniel Golliher 🗽

@danielgolliher

🗽 https://t.co/lOqCit2T4r

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/WOXFTlWGEp

Daniel Golliher 🗽

@danielgolliher

@Prigoose @christineist @funplings @Liface :0

Daniel Golliher 🗽

@danielgolliher

Needs immediate city-wide deployment https://t.co/V60SeWASqU

Daniel Golliher 🗽

@danielgolliher

Some background: https://t.co/qFbEnLEJuE

Daniel Golliher 🗽

@danielgolliher

"The great privilege of the Americans is to be able to have repairable mistakes." Tocqueville's "Democracy in America," Volume 1, Part 2, Chapter 6. https://t.co/NdXaqahLVv https://t.co/t7ogTkJ8qy

Daniel Golliher 🗽

@danielgolliher

@IsaiahMenning @ACC_National @JoinFAI Congratulations on the move!

Daniel Golliher 🗽

@danielgolliher

Good question for anyone to contemplate: does your institution/social graph have an outgroup advocatus diaboli to consult? https://t.co/Hd3fXDAog0

Daniel Golliher 🗽

@danielgolliher

@Prigoose 🚨Helen Keller and Anne Sullivan mentioned🚨

Daniel Golliher 🗽

@danielgolliher

*applications. Proof of humanity!

Daniel Golliher 🗽

@danielgolliher

@Liface https://t.co/VBuIICLNlw

Daniel Golliher 🗽

@danielgolliher

I’m just scrolling along and get to spectate so many cool developments https://t.co/etHLiO08WA

Daniel Golliher 🗽

@danielgolliher

@Prigoose Sunday night, and also every night

Daniel Golliher 🗽

@danielgolliher

Yes https://t.co/2qM2OGMkvA

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/0I7JBTbDpC

Daniel Golliher 🗽

@danielgolliher

@NYCCouncil @CarlMichaelW @OsseChi Side note here on company support for the RAISE Act—the law’s author says otherwise, as do many people who reported on the bill when it was in process: https://t.co/FvCNoF1ZH5

Daniel Golliher 🗽

@danielgolliher

@tasshinfogleman @zencephalon 👀👀👀

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/bXmqGsyWG9

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/vb21TLP9us

Daniel Golliher 🗽

@danielgolliher

@axios 2 in 3 can according to the survey, and that’s way up over time! https://t.co/9aIms4cmRJ

Daniel Golliher 🗽

@danielgolliher

One of the turns of phrase I've been watching this year is "to know ball" (I like it, but I did notice an uptick). I asked ChatGPT to make a graph of the evolution of "to know ball" from a sports phrase, to a more general application of knowledge in any domain. https://t.co/d5m5zC5WeJ https://t.co/iD65kLp8AI

Daniel Golliher 🗽

@danielgolliher

@m_adams Sea to shining sea, bay-bee

Daniel Golliher 🗽

@danielgolliher

@GarettJones Generous? Yes. Wise? Separate consideration.

Daniel Golliher 🗽

@danielgolliher

@kvallier Very interested in following this work!

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/IJuvcnkVkH

Daniel Golliher 🗽

@danielgolliher

See also: https://t.co/WKbo4wSBny

Daniel Golliher 🗽

@danielgolliher

Again FYI, get @NYCComptroller’s newsletter https://t.co/V1jQdO7IaF https://t.co/TuNKeflbAe

Daniel Golliher 🗽

@danielgolliher

@cojobrien Several people have said this to me! Always down to assist a Foundations of DC endeavor

Daniel Golliher 🗽

@danielgolliher

!! https://t.co/z6LfXUHU8y

Daniel Golliher 🗽

@danielgolliher

Heavens https://t.co/8KKJ8Fnfjr

Daniel Golliher 🗽

@danielgolliher

Seems like a good idea https://t.co/qRvACpOF6v

Daniel Golliher 🗽

@danielgolliher

@IrvingSwisher https://t.co/ZupAD8XnWz

Daniel Golliher 🗽

@danielgolliher

https://t.co/31BNJ02fPv

Daniel Golliher 🗽

@danielgolliher

https://t.co/djRFHMlDgO

Daniel Golliher 🗽

@danielgolliher

Also: https://t.co/NJjM1vOAU2

Daniel Golliher 🗽

@danielgolliher

@GarettJones @ufhamilton https://t.co/ZRlabtgH9u

Daniel Golliher 🗽

@danielgolliher

@zdch 🫡

Daniel Golliher 🗽

@danielgolliher

@nicolegelinas I strive to deliver

Daniel Golliher 🗽

@danielgolliher

@ShahanaFromBK @NYCCouncil 🫡 I look forward to the balance of the hearing, and seeing your further thoughts too once it’s done!

Daniel Golliher 🗽

@danielgolliher

@EricCrampton @CMPiSanchez Me too! I’ve been helped greatly by them, as have close friends. Anthropic’s @logangraham has a story that I’ve heard many and increasing versions of: https://t.co/O5f9pOWutg

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/WtvBG5CPm6

Daniel Golliher 🗽

@danielgolliher

Go grab the full report: https://t.co/4dpd5XnRvV

Daniel Golliher 🗽

@danielgolliher

@JNicholsWorley Stay tuned for a post I’m cooking up on almost exactly this 🫡

Daniel Golliher 🗽

@danielgolliher

🚨New Bagley book!🚨 https://t.co/q5YxQVPRAo

Daniel Golliher 🗽

@danielgolliher

@fstflofscholars Mine is 9/14! Getting excited

Daniel Golliher 🗽

@danielgolliher

This, on the other hand, is more my speed if the goal is a jocular post: https://t.co/YAdtkNwISM

Daniel Golliher 🗽

@danielgolliher

@emmma_camp_ Well earned! Congratulations!

Daniel Golliher 🗽

@danielgolliher

@Trent_STEMpunk 🫡

Daniel Golliher 🗽

@danielgolliher

All-of-the-above is the name of the game with energy, and natural gas is a big part of that. https://t.co/UzRvFbIQTY

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/eQOFb0cEg8

Daniel Golliher 🗽

@danielgolliher

Also: https://t.co/19ZVeHKIlr

Daniel Golliher 🗽

@danielgolliher

@mbateman At the end of my government classes we read Lincoln's Lyceum Address from 1838, and discuss the overall societal context that necessitated its delivery. Grim! And yet here we are today.

Daniel Golliher 🗽

@danielgolliher

@markfishman_xyz They will! True public servants

Daniel Golliher 🗽

@danielgolliher

@jasminewsun's piece: https://t.co/Bjji4Mg9QP

Daniel Golliher 🗽

@danielgolliher

@lorenz_system Need

Daniel Golliher 🗽

@danielgolliher

@neejkapoor @AlexTISYoung 💙

Daniel Golliher 🗽

@danielgolliher

@ahall_research @danmthomp https://t.co/BWeijPBMVn

Daniel Golliher 🗽

@danielgolliher

@ChrisGillett It just wanted to explore!

Daniel Golliher 🗽

@danielgolliher

@lorenz_system @NYCHousing Yesssssss

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/744wSxl4qc

Daniel Golliher 🗽

@danielgolliher

@ajdlinux oh NO

Daniel Golliher 🗽

@danielgolliher

Wickard came up in class. Always a fun one!

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/fp2kqGiT3Y

Daniel Golliher 🗽

@danielgolliher

The piece: https://t.co/xm20FEyQyS

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/dEfdLbrIAP

Daniel Golliher 🗽

@danielgolliher

From "Facts of the case" section: https://t.co/v2cyfbeNib

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/tI7GGaZBxb

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/TOfDwF64l5

Daniel Golliher 🗽

@danielgolliher

Related: https://t.co/FXp5akb4kw

Daniel Golliher 🗽

@danielgolliher

Ahem: https://t.co/ryV9Dvoihv

Daniel Golliher 🗽

@danielgolliher

The legistar link: https://t.co/dwsbqjEMrQ

Daniel Golliher 🗽

@danielgolliher

@Paul_Taske YES

Daniel Golliher 🗽

@danielgolliher

The post: https://t.co/7WNqJIxSnL

Daniel Golliher 🗽

@danielgolliher

@chill__berrt YES