16,226 karma · joined April 9, 2019
But being difficult to put into action doesn’t mean the advice is wrong. Sleep deprivation measurably increases cortisol and inflammatory markers. Exercise measurably reduces them. These actions have quantifiable sometimes immediate effects regardless of how we define stress.
They trained on synthetic extractions like "extract equations from arXiv papers" and "extract regulatory information from FDA documents," then tested on more synthetic extractions from the same sources. Essentially, "model trained on synthetic arXiv/PubMed/FDA extractions performs better on more synthetic arXiv/PubMed/FDA extractions than a model that never saw this distribution."
I'd like to see how it handles extractions from a real contract, or a low quality scan of a financial document, or processes a format it didn't see in training. o3 very likely handles these variations better, but we don't have that data to compare.
We need the model weights or tests on standard benchmarks to verify if this generalizes beyond documents that look like the training distribution.
I’m guessing it might be explained by the hygiene hypothesis.
The oral microbiome needs balance, not sterilization.
[0] Mouthwash ≥2x/day → 55% increased diabetes risk Nitric Oxide journal, 2017
[1] Mouthwash ≥2x/day → 117% increased hypertension risk Blood Pressure Journal, 2020
[2] 7 days chlorhexidine mouthwash → 90% reduction in oral nitrite, BP increased 2-3.5 mmHg Free Radical Biology & Medicine, 2013
[3] Daily mouthwash → OR 1.31 for head/neck cancer European Journal of Cancer Prevention, 2016
[4] Listerine Cool Mint increased Fusobacterium nucleatum and Streptococcus anginosus (bacteria linked to colorectal cancer) Journal of Medical Microbiology, 2024
Perhaps minimal, but this does in fact prevent the specific attack vector they demonstrated. The criticism seems unnecessarily harsh given that Google addressed the vulnerability immediately.
But MCP servers are more analogous to a PyPI packages you pip install, npm modules you add to your project or a VSCode extension.
Nobody would argue that pip is fundamentally broken because running pip install malicious-package can compromise your system. That's expected behavior when you execute untrusted code.
edit: after parent clarification
Zuck: Just ask.
Zuck: I have over 4,000 emails, pictures, addresses, SNS
[Redacted Friend's Name]: What? How'd you manage that one?
Zuck: People just submitted it.
Zuck: I don't know why.
Zuck: They "trust me"
Zuck: Dumb fucks.
Mark Zuckerberg, 2004
[0] https://www.esquire.com/uk/latest-news/a19490586/mark-zucker...
It's a bit disingenuous to pretend that the WEF and EU have not been suggesting a global switch to insect protein.
[0] 5 reasons why eating insects could reduce climate change
https://www.weforum.org/stories/2022/02/how-insects-positive...
[1] Why we need to give insects the role they deserve in our food systems
https://www.weforum.org/stories/2021/07/why-we-need-to-give-...
[2] Saving Lives and the Planet: Balancing Diet with Alternative Proteins Can Cut Deaths by 5% and Emissions by 25%
https://www.weforum.org/stories/2021/05/europe-insect-based-...
[3] Europe backs insect-based food in a bid to promote alternative protein
https://www.weforum.org/press/2019/01/saving-lives-and-the-p...
[4] Will you be eating insects soon? Can insects feed the world?
https://intelligence.weforum.org/monitor/latest-knowledge/e9...
[5] Meat: The Future series - Alternative Proteins
https://www3.weforum.org/docs/WEF_White_Paper_Alternative_Pr...
[6] EU Regulation authorising the placing on the market of UV-treated powder of whole Tenebrio molitor larvae (yellow mealworm) as a novel food and amending Implementing Regulation (EU) 2017/2470
The people writing all of these "AI is unprofitable" pieces are doing financial journalism similar to analyzing the dot-com bubble by looking at pets.com's burn rate. The infra overspend was real as well as the bankruptcies, but it existentially foolish for a business to ignore the behavioral shift that was taking place.
I have to constant remind myself to stop arguing and evangelizing about AI. There is a growing crowd who insists that AI is a waste of money and that AI cannot do things I'm already doing on a daily basis.
Every minute spent explaining to AI skeptics is a minute not spent actually capitalizing on the asymmetry. They don't want to hear it anyway and I have little incentive to convince anyone otherwise.
The companies bleeding money to serve AI below cost prices won't last, but thats all more the reason use them now while they're cheap.
We are creating a massive competency gap by treating AI exposure as somehow more dangerous than social media, which we've already allowed to reshape adolescent development with inarguably negative educational value.
AI is already redefining job requirements and academic expectations. Students who first encounter these tools in college will be competing against peers who've had years to develop working usage patterns and build domain specific applications.
The AI literacy measures are fundamentally flawed. The 25-item scale confuses unrelated concepts, mixing questions about HIPAA regs, PCI DSS standards, and password storage with actual AI knowledge. Question 10 asks about HIPAA sharing rules, question 5 about payment card standards. These measure regulatory knowledge, not AI literacy.
Each study uses correlational designs while making causal claims. The title itself "Lower AI Literacy Predicts Greater AI Receptivity" implies directionality that cross-sectional data cannot establish.
Study 1 uses 27 countries with massive unmeasured confounds l;ike cultural attitudes toward technology, digital infrastructure, education systems to claim AI literacy drives receptivity Study 2 has a 41.9% attention check failure rate that correlates with AI literacy scores. Authors even acknowledge this creates bias but proceed anyway.
Studies 5 and 6 test mediation through "magical thinking" using cross-sectional data, violating the temporal precedence requirement for mediation. You cannot establish that low literacy → magical thinking → high receptivity without measuring this over time. The Hayes PROCESS macro they use explicitly warns against this.
The "magical thinking" explanation appears nowhere in the preregistrations but is presented as the main theoretical contribution. The explanation is unfalsifiable and contradicts their own data. Study 6 shows people with low AI literacy rate AI as less capable and more fearful yet supposedly use it more because it seems "magical." This is incoherent. why would perceiving something as less capable but magical increase usage?
The biggest flaw is that without establishing what "AI literacy" actually measures, every subsequent analysis is meaningless.
The study shows different brain patterns during AI-assisted writing, not permanent damage. Lower EEG activity when using a tool is expected just as showing less mental math activity when using a calculator.
The study translates temporary, task-specific neural patterns into "cognitive decline" and "severe cognitive harm." The actual study measured brain activity during essay writing, not lasting changes.
Plus, surface electrical measurements can't diagnose "cognitive debt" or deep brain changes. The authors even acknowledge this. Also, "83.3% couldn't quote their essay" equates to 15 out of 18 people?
The trolley problem presents an either/or scenario. Pull the lever or don't. This is a false dilemma because there arew other options like negotiation, compromise, peace processes, international mediation, addressing grievances.
Scott assumes the "murderer" is irredeemably committed to killing your child, that the threat is existential and immediate, and that defensive actions are the only possible response, then uses these assumptions to conclude you must take defensive action. The conclusion is baked into the circular setup of the thought experiment itself.
A straw man is constructed by characterizing critics only as people who cheer for the murder of Jews, want Israel "decolonized" in a way that implies genocide, or are equivalent to Nazi collaborators. This misrepresents the majority of the criticism of Israeli policies, which usually focuses on specific military actions or settlement policies over questioning Israel's right to exist.
False equivalence applying "Deep Zionism" moral framework across wildly differing scenarios like defending against existential threats, leaving school early, pursuing quantum computing research, and asking women out.
ad hominem attacks, calling critics "snarling nobodies," "self-hating Jews," and "obsessive haters"
Appeal to emotion, false attribution of collective guilt (all Palestinians = murderers, the crowd = the entire international community as if there werent differing opinions.)
By closing comments and declaring there's nothing here to debate, Scott is trying to tie his readers to those same tracks, forced to accept his moral framework or be branded as part of the hateful mob.