@superalbs @muunijs @sxarek I have the opposite preference (feels socially awkward to sleep in a small room with strangers). I’m much happier with the arrangement in an airplane.
English
Dominik Peters
4.9K posts
Dominik Peters
@DominikPeters
CS researcher in Paris (CNRS) on voting theory. I travel by train a lot. From 🇩🇪, studied in 🇬🇧, worked in 🇺🇸 and 🇨🇦, now live in 🇫🇷.
Paris, France เข้าร่วม Haziran 2009
354 กำลังติดตาม493 ผู้ติดตาม
Dalej nie pojmuję dlaczego na kolei tak się wzbraniają przed dwufunkcyjną klasą business, która za dnia jest produktem premium a w nocy tańszym rozwiązaniem od wagonu sypialnego.
Jedynie w Norwegii coś takiego spróbowali ale kompletnie zepsuli puszczając tylko w nocy
Vojtěch Očadlý @vojtaocadly
Proc neni ve vlacich prvni trida jak v letadle? WDYM ze si ve 21:00 nemuzu oblict pyzamo a vycistit zuby, zatimco mi stewardka JLV ustele mou plne polohovatelnou sedacku
Polski
@ryanflorence Absolutely. It’s an overcorrection from the sycophant phase. Now it tries REALLY HARD not to straight up agree with you.
English
"Hostile reframing"
Recasting a neutral idea into something unreasonable before responding
GPT-5.5 does it to me all the time
Ryan Florence@ryanflorence
GPT has a new phenomenon that's driving me nuts and I don't quite know how to describe it. - Ask it if I can do something - It says "no you can't [incredibly twisted restatement of what I asked but also not at all what I asked]" - It then tells me how to do the thing wonderfully - And finishes with an insulting "But you can't just [stupid thing I never actually said]" It goes something like this: "Can I form and coach a youth soccer team for my kid and play in P/D level leagues? Or do I have to be part of a full club?" Then it says: "For official competitive teams in Utah, you cannot just form a random team and enter a league. "There is a lesser-known option: UYSA allows independent teams to enter leagues if they meet requirements. [...lists some simple requirements...] "But it’s not “show up with a group of kids on game day”—it’s more like running a small club team administratively. I never said just "show up with a group of kids on game day"! It does this to me with code too. It's so weird.
English
@thsottiaux @jxnlco I like using 5.4-mini at times but with the new model picker this takes many clicks. Can the model picker remember which models the user has used in the last few days and expose them before going into “more” menus?
English
@thsottiaux @jxnlco When a request fails (eg network error or out of usage limit), there is no button for resending the same message later like “Try Again”. I usually “edit” the message and send it unchanged but that is dirty.
English
@DominikPeters @lefineder Majority switches from all-blue to mostly red with epsilon ^ (N/2) probability. So switching from blue to red yourself is very very slightly better.
English
Pressing red is the Nash equilibrium, meaning that it's the best choice regardless of the choices of others (which you don't control).
So if a group of rational agents (in the game theoretical sense) were presented with this problem, they would all press red and all survive.
Although it doesn't look like it, it’s basically the opposite of the Prisoner’s Dilemma, in which rational agents following self-interest leads to a worse outcome for everyone.
Tim Urban@waitbutwhy
Everyone in the world has to take a private vote by pressing a red or blue button. If more than 50% of people press the blue button, everyone survives. If less than 50% of people press the blue button, only people who pressed the red button survive. Which button would you press?
English
@MannOhneListe @RichardMCNgo @lefineder In practice, yes, but Nash equilibrium is a precisely defined mathematical concept and it deliberately does not contain uncertainty about the actions of others. If you want to reason about that you’ll want other concepts like Bayes-Nash equilibrium
English
@DominikPeters @RichardMCNgo @lefineder The incentive lies in not actually knowing you're in the majority.
English
@RichardMCNgo @lefineder Hmm which Nash equilibria fail to be trembling hand? A clear blue majority remains blue even with epsilon noise so blue is a best response, and same for all-red
English
@lefineder Any blue majority is also a nash equilibrium.
The criterion that distinguishes them is trembling-hand equilibrium
English
@MannOhneListe @RichardMCNgo @lefineder Nash equilibrium means that if we are in a particular situation (eg a 57% blue majority), then each person has no incentive to change their action *assuming that no one else changes their action*
English
@RichardMCNgo @lefineder Uh, no, because a *majority* inherently depends on the decisions of others.
English
First impressions of GPT-5.5-Thinking for mathematics are scary good. Many of my old conversations where 5.4 couldn't solve the problem now seem to get solved (but still need to check correctness).
English
very basic question that's important and relevant to OpenAI and Anthropic: is session limit consumption for subscribers proportional to model API pricing?
Lisan al Gaib@scaling01
GPT-5.5 Pricing & GPT-5.5 Pro Pricing GPT-5.5: $5/$30 GPT-5.5-Pro: $30/$180 (Input/Output per million tokens)
English
@AaronBergman18 This already exists in many implementation and is in use in the countries that have recently implemented age gating for various websites. Works exactly correctly for me. examples: veriff.com/demo/age-estim… innovatrics.com/age-estimation/
English
You could imagine biometric ~proof of age, at the very least
I’d imagine it is not a hard ML task to do “look left, look right, left again, blink, chin up” -> verify that a live person is there + then run an age prediction model and if it says p(>=18) >= 99% or something then throw out the data and let them register
English
Pretty obvious point in hindsight I failed to see: you literally don’t get access by proving your age is above some threshold (not that that would necessarily be good or fine, but still)
Taylor Lorenz@TaylorLorenz
Mind you it’s not age gating, it’s IDENTITY GATING. There is no such thing as “age verification” it’s ID verification. I think we need to stop using the term age verification bc it’s an industry term that obfuscates the reality of what’s happening.
English
@btibor91 Sounds like fast answers are cached responses from a database, maybe through some sort of embedding RAG
English
OpenAI rolled out new ChatGPT updates including Fast answers, ChatGPT for Clinicians with HealthBench Professional, and ChatGPT for Google Sheets
- Fast answers gives quicker replies to simple info-seeking questions when ChatGPT is confident in the answer, skips past chats and memory, works globally on web, iOS, and Android for logged in and logged out users across all plans, and can be turned off in Personalization settings
- ChatGPT for Clinicians is a free version for verified US physicians, NPs, PAs, and pharmacists covering evidence review, documentation, and medical research, with clinical search and citations, reusable skills, deep research across medical literature, CME credits on eligible questions, and optional HIPAA support via a BAA, shown as a separate workspace under the same ChatGPT login, with plans to expand to more countries via a Better Evidence Network pilot
- OpenAI also released HealthBench Professional, an open benchmark built on HealthBench that uses physician-authored conversations and rubrics to test models on real clinician chat tasks across care consult, writing and documentation, and medical research, where GPT-5.4 inside ChatGPT for Clinicians beat base GPT-5.4, other OpenAI and external models, and human physicians
- ChatGPT for Google Sheets lets you create new sheets, ask questions across tabs and formulas, and make updates directly in your sheets
English
@DLPTownSquare When is the last time that a train stopped at every station?
English
I’m sorry, but this is not acceptable in a Disney Park. That train should run all day with all stations open. The state of the Railroad operations in Paris is simply poor.
DLP Report@DLPReport
🚂 Note the summer hours for the Disneyland Railroad, now in effect with the first train not until 12:30pm, but running until 7pm.
English
@GarettJones I wonder if Parisians could correctly rank-order the arrondissements by rent. I think people think the 16th is more expensive than it really is.
English
Average Paris rents differ by about 35% across neighborhoods.
Average DC rents differ by about 150% across neighborhoods.
English
@GarettJones Potentially the DC numbers are skewed a bit if 2BR apartments are bigger in higher-priced areas, since it is not normalized per m^2.
English
@WhatIsPrivate1 @emollick yes, it's interesting that AI doesn't seem to understand in what ways earlier image models are bad. The issue was not blurriness
English
I have been using GPT ImageGen-2 for the past weeks
I didn't think that better image-generators would be a big deal but it turns out that there is a quality threshold I didn't expect, where you can now get text, slides, academic papers
Look at what it does with my "otter test"!
English
@amaldorai @matt_beard_ Oysters and crabs are quite different on the relevant dimensions though
English
@matt_beard_ I know an otherwise-vegan who eats oysters because she says they’re incapable of experiencing suffering
English
only eating large crustaceans because you believe they have the best ratio of sentience : calories is the most EA thing ever
Elena@VirtualElena
dario ordering crab chowder during his "lunch with the ft" interview....not very EA of him...
English
@jxnlco @ajambrosino I suppose it could auto answer in case of no user input for 60s and tell codex that the user was unavailable and to use its best judgment
English
@jxnlco @ajambrosino to be fair, it is nice to know that after I click “implement plan” I know that I can go away from the computer and won’t accidentally stall progress by not timely answering a question. I don’t know the right UI for indicating “I’m open to questions” or not
English
5 steps to making codex your chief of staff
1. download the desktop app
2. install the plugins you need for work
3. paste this into a thread and pin it
4. ???
5. monitor the situation
gist.github.com/jxnl/e96b08aae…
English