paytonjjones

Title is misleading. The vans are manufactured by an Israeli company and bought by US government entities, not vans affiliated with the nation state of Israel.

I always thought that retransmitting the full chat every turn is a foundamental problem

So did many NLP experts who worked on LSTM, GRU, etc. But transformers empirically won out and here we are.

It will be interesting to see if future model architectures need to take this into account again.

If humans are any indication, one thing we're still missing is efficient movement of information from short term to long term memory. Fine tuning in principle does something akin to this, but not efficiently.

from more than 83,000 automated license plate reader

This feels purposefully misleading. Did they manually watch hours and hours of footage, or is this the equivalent of "I searched an online corpus of hundreds of millions of documents" (I Googled it)

I can understand saying "will be" a disaster, but "have been"?

How do you square that with data like this?: https://ourworldindata.org/cdn-cgi/imagedelivery/qLq-8BTgXU8...

What was the first DeepSeek moment? (genuine question, I'm out of the loop on what you mean)

People like being needed and important to other people. That isn't some artifact of capitalism; it exists across all times and cultures.

Is there any entitlements program that's not politically impossible to repeal?

Taking away "free stuff", especially once people have come to rely on it, is a political nightmare no matter the structure.

It's hard to tell whether this is sarcastic, because like the safety of nuclear energy, it's literally true but few people believe it.

A big political problem with YIMBYism is that almost by definition it will take years to see the effects.

If you champion it as a politician, there's a very large chance you get all the flack for new developments, but the next guy gets all the credit for affordable housing.

Texas seems to be the only place that actually does it, and I honestly have no explanation for why.

Easier with a duopoly for sure (if there's something restricting new entrants to the market, that's a big problem in its own right), but even so, it's a prisoners dilemma.

They are better off cooperating, but if one defects first they can screw the other one, so it relies on trust.

That's true, but the thing about price fixing is that it basically guarantees these conditions.

The price is artificially high -> there's a ton of demand waiting to be unlocked by the "potential energy" gated behind the unnatural price

Capacity is easy to plan around; get too much and you can just raise the price again.

informal agreement to keep transaction fees about the same

This would be like, ridiculously illegal. It's also not practical; if you can steal customers, the incentive to undercut is very strong.

Someone shared this with me the other day. Haven't used it much yet but it seems like it will easily beat my previous strategy of plugging things in randomly and getting aggravated: https://www.whatcable.uk

Something I never considered much is what happens when everyone else is using the Whispering Earring.

You may be more free and independent, but you may also be unable to compete as everyone else easily gains wealth and success. Natural selection doesn't particularly care about freedom of consciousness.

Bleak.

Selection effects are ridiculously strong for almost anything related to health.

I strongly believe that's why nutrition science is soooo far behind the rest of medicine. There aren't nearly enough serious RCTs (whereas regulations make them abundant for other medical interventions).

Hmm you are right. But looking closer at the methodology, I find myself even more confused.

It seems the metric is something like "most successful stewards of shareholders' investments" which I guess is interesting. But now I'm tripped up on why the metric would only consider founders rather than CEOs more generally. Imagine Gates didn't start Microsoft, but instead became its CEO a month after some other founder started it and that founder sat on the beach in Hawaii while Gates did well, what he did. The founder would appear on this list but not Gates.

Edit: basically, all my intuitive "this doesn't make any sense" alarm bells are going off, but I think I need someone who really knows what they're talking about to help me understand exactly why, or what would be a more sensical version of this

I'm no economist but I'm having a hard time grokking any meaning out of this metric.

So if Elon decided to sell all his shares today (and likely destroy his companies in the process), he'd shoot to the top of the list? What's the point in that?

My 401k has benefitted from the growth of e.g. Amazon for sure, but the main 'wealth' I get from them is my ability to buy anything and get it delivered in a day. That is, I benefit from their infrastructure existing, regardless of who the shareholders are.

I feel like...that's terrible advice?

Voluntary retrieval (the memory "mode" of flashcards), even for simple pairs, is perceived as highly effortful by almost everyone, on a similar tier to doing mental math.

It's incredibly efficient for learning, and achieving your goals quickly can be very rewarding, but I don't think "learning flashcards should be fun!" is a reasonable expectation for the vast majority of people.

there is "natural" or "intuitive"

Your argument is sound but this overstates your case a bit. There's a reason we don't type with our toes.

I find it more plausible given my priors as a research psychologist.

Correlational overlap between self-report scales is common. It's actually so common that at decent sample sizes, almost any arbitrary self-report will significantly correlate purely on valence. For instance, one should expect a correlation between people self-reporting "I had a tough childhood" with "I frequently suffer from skin rashes". This is because trait neuroticism (the degree to which people are prone to feel negative emotion) is a major driver of how people interpret and respond to scales generally.

In contrast, psychologists mightily struggle to find replicable experimental results. That's what you need here.

A better version of this study would be to run an experiment where you take away (or heavily restrict) parental phone access over a month or two and measure the parent-child relationship vs. a control group.

"Pay more attention to your kids", which is a pretty good universal message to put across.

I wouldn't be too sure of that actually: https://www.archbridgeinstitute.org/the-secret-to-parenting-...

This is a weak study that is exemplary of psychology's weak experimentation culture and correlation/causation laundering, especially with regard to self-report.

The heavily hinted implication is that device use damages relationships. But look at what they actually measured. They ask adolescents to answer questions like:

"My primary caregiver ignores me when they are on a device." (DAIS, their new scale)

And then also ask them to answer questions like:

"I often worry that this person doesn't really care for me." (ECR-RS)

And then act like it's a revelation that these two self-report scales are correlated.

A much more plausible causal explanation is that a single psychological variable (e.g. a bad relationship) causes both self reports, rather than the implied pathway that device use causes A, which then causes B.

Today's SOTA also sounds totally sufficient to me, but I wonder how much our standards will inflate by 2028. Maybe a lot, maybe not at all...very hard to say.

I've treated people with panic disorder who, if there were no cost involved, would probably have called an ambulance monthly for suspected heart attacks.

I wonder how countries with universal healthcare coverage deal with the lack of a (dis)incentive here. Maybe they just eat the cost?

doesn't matter how many more nines you add

I don't get this argument. People say the same about autonomous driving.

But humans also have some number of nines. If you can get it better than humans, that's better!

Yep. It will be interesting to see if things move that way or not. AI scraping might get annoying but a lot of sites (eg e-commerce) have obvious incentives to allow AIs to scrape them for free (or even pay AI companies to prioritize them, like they already do with Google).

These do make up a huge % of tool calls, but I don't think these make up a huge % of tool call failures.

I see models fail on tool calls that involve API requests to a specific API, internal or cloned Makefile calls, npm run commands, etc.

Better business model would be some sort of micropayment setup - allow humans, but make scrapers pay a hundredth of a cent for access.

The cap does not apply to Grok, Mr Musk's own AI system

Super misleading headline. Please add "external AI" or "non-Grok AI"

I wonder if we'll start to see that pattern with every new release. Tool use likely changes rapidly, so the newest, rather than most intelligent, model may always have an edge.