violet13

I don't follow. You can source the chips, in large quantities, right now, and have them delivered to any non-embargoed country in a matter of days.

That they're in limited stock at a particular outlet in China means very little, no? And conversely, that STM32 chips happened to be in stock in 2018 meant very little in the following four years.

If you want the chips right now, you can have them. If you want long-term availability, either buy ahead of the time or sign a contract with the manufacturer (and hope it doesn't fall apart due to a geopolitical crisis or whatnot).

With a side serving of fake parts, multi-week transit times, and the added benefit of propping up the Chinese regime.

I don't understand what you mean. These chips are available in bulk from the manufacturer and from all major distributors worldwide. Why would you want to go through China?

If you mean a PCBA shop, Mouser or DigiKey can ship your order directly to your favorite PCBA place.

UPDI has robust on-chip debugging capabilities.

Very few hobby designs that use more capable chips actually need more capable chips, and the added complexity - hardware and software - often bites hard.

That said, wifi is a major selling point, and you need a beefy 32-bit computer to run the protocol. This is the brilliance of ESP32: you can actually use it just as a wifi dongle for another microcontroller, but since it needs so much computing power, might as well give you some RAM and CPU to do your thing directly on the wifi chipset...

What's interesting to me is why this story leaked. I bet it's just Morgan Stanley (the owner of E-Trade) trying to control they narrative. If they just cut him off, he gets to tell his story, MS can't officially discuss the particulars, and the company ends up getting a lot of hate mail.

But if they leak the official justification beforehand, then Roaring Kitty will have a harder time spinning it.

Not entirely, but either way, land value is tied to all of that. If the city can't grow due to zoning and conservation rules, and if you can't subdivide existing parcels, you basically have more and more money chasing the same lots.

By "not entirely", I mean that while land value often dominates, lowering it doesn't solve the problem, because even if you somehow acquired a city lot for $50k in the Bay Area, you wouldn't be able to build a house on that lot for $100k. So, solving one problem doesn't make it affordable.

New construction is rapidly getting more expensive in many of the markets where it's needed the most. While it'd be cool to point to a single factor, it's a combination of several things. People are expecting more (more square footage, higher-end finish), the government is demanding more (environmental and energy regulations), labor is getting costlier, and yes - there's NIMBYism manifesting through zoning laws and other "conservation" rules.

It is a lossy compressed index. It has an approximate knowledge of law, and that approximation can be pretty good - but it doesn't know when it's outputting plausible but made-up claims. As with GitHub Copilot, it's probably going to be a mixed bag until we can overcome that, because spotting subtle but grave errors can be harder than writing something from scratch.

There's already a fair number of stories of LLMs used by an attorney messing up court filings - e.g., inventing fake case law.

This, along with several other "meta" objections, is a significant portion of the discussion in the paper.

They basically say two things. First, although the measurement is repeatable at face value, there are several factors that make it less impressive than assumed, and the model performs fairly poorly compared to likely prospective lawyers. Second, there is a number of reasons why the percentile on the test doesn't measure lawyering skills.

One of the other interesting points they bring up is that there is no incentive for humans to seek scores much above passing on the test, because your career outlook doesn't depend on it in any way. This is different from many other placement exams.

YouTube runs ads on all my videos, including the ones with fewer than 200 views. I'm not sure what you're getting at otherwise. The point I was making is that paying for YouTube does absolutely nothing to support the vast majority of content creators on the platform. All you're doing is paying Google, and a tiny slice of that supports a handful of celebrity content creators like Mr. Beast. Maybe it's a bargain that makes sense to some, but the parent was deriding the grandparent for not willing to pay.

They by and large don't. Most creators on YouTube aren't even eligible to monetize, especially if they're making niche educational content.

Google just pockets all the revenue if you have fewer than 4,000 "public watch hours". If you put together an a concise 4-minute DIY video, an average view time will probably hover around 1 minute, and you will need 240k views to qualify. Most videos on YouTube get under 1k views.

There is a negligibly small percentage of content creators / content farmers who live in a symbiotic relationship with the platform and actually make good money, but let's not pretend you're patronizing the creators when you're paying for YouTube Premium.

How many times in history have we regretted over-building this kind of infrastructure? We have no shortage of good ways to use electric energy. If AI turns out to be a fad, others will benefit.

Pollution is a concern, but I think don't think that anti-growth or anti-consumption sentiments are useful. All this money pouring in is an opportunity to invest in clean energy.

The US was historically built on the idea that free enterprise, for all its warts, is less bad than governments meddling in your life. Much of modern Europe is built on the opposite idea: that governments know better and that private enterprise is inherently corrupt.

I don't want to pick sides in that debate, but it amazes me how wonderfully ahistorical it is. The US doesn't have a history of oppressive governments; Europe, on the other hand...

Anyway, the disconnect you allude to isn't a disconnect if you consider it from this angle. GDPR is meant to protect you from private businesses. Chat control is supposed to protect you (and the state) from online predators. It was at no point about protecting you from the government.

Let me say that I really like that this is a free textbook, and I realize it's a textbook for EE students, not the general public.

That said, it bums me out a bit when textbooks are written in this style. It's essentially a big data dump of formulas that doesn't explain the "why".

I know that classical magnetism doesn't have a good explanation beyond "it is like that because that's how it is." But at the very least, the formulas are derived from and relate to empirical observations. You don't teach chemistry by treating it as a set of math axioms, right?

"The Art of Electronics" doesn't exactly handhold you through the basics, but it takes a far more accessible approach to EE once you grasp the basics. And their secret is real-world examples, anecdotes, and so on.

Nuclear war is not an extinction event, though. Most people would survive, and while they would face hardships, the "radioactive wasteland" thing is a fairly sloppy sci-fi trope.

Before you say "another reason to stop using Paypal", keep in mind that credit card companies routinely share purchase data with online advertising companies. This is so that Google and others can tie online impressions to brick-and-mortar purchases, among other things.

Paypal is playing catch-up.

Homelessness, especially in places such as the SF Bay Area, really doesn't boil down to just affordability. Yes, there are some folks who just faced economic headwinds and are living in a car while trying to find a way out. But there is also a huge population of people who couldn't function if given keys to a free apartment.

For one, drugs won the war on drugs, addiction is a big part of the problem, and we don't really know how to fix it; harsh punishments don't work, quasi-decriminalization isn't a success, and treatment for people who don't want to be helped is hard. We also don't like to institutionalize people anymore, so folks with severe mental illness often end up on the streets too.

> Anecdotally, in conversations with other drivers, it seems that far too many people overestimate their driving habits.

Or conversely, folks who are making this argument underestimate the flexibility people want out of a car.

If all you factor in is the average daily work commute, then most car owners don't need a car to begin with - even in the US. There's usually some public transport, or work-provided transportation, or opportunities to carpool with a neighbor. Not glamorous, but enough to get you through your average day.

The car culture has very little to do with averages. And for what it's worth, it's the same for most other goods, from computers to kitchen appliances. I mean, how many techies need a kitchen in their home to begin with?

My point is that when you have it configured the usual way, you can type "2 + 2 =" in the address bar and get the result. Which, you know, is sort of convenient. Or "weather mountain view ca", or whatever.

If you change your default search to udm=14, you lose that. Sure, you can click around on the now-useless results page to get back to standard Google and get the answer, but that's friction that wasn't there before. The parameter doesn't just kill the AI stuff.

Unfortunately, this throws the baby out with the bathwater - you also lose all the useful non-AI widgets, like calculator, unit conversions, instant weather, maps, etc.

There are other tools for that, but I confess that I've gotten used to defaulting to the address bar for random queries like that - the browser is always open.

It sounds like an impressive list, but many of these are ghost towns and are of little value to machine learning. Wikiversity is a mess... Wikipedia is the crown jewel and probably the only thing of unique commercial value for ML.