People obeying the speed limit encounter fewer people obeying the speed limit than people speeding and vice versa.

My theory: I think this explains why so many drivers hate other drivers and think they are bad drivers.

People that love to speed think they are good because they can drive faster and react quickly (presumably). They inadvertently see more people that don’t drive in this style.

People who drive carefully encounter more reckless drivers.

Driving carefully (resp. recklessly) isn't the same as driving under (resp. over) the limits.

I've been pulled by gendarmes who told me "yes, we know this limit should be 20 km/h higher" but still fined me. Absurd limits targeting the lowest common denominator in vehicle/driver reliability or simply because your local mayor wants to turn his city into a pedestrian/cyclist paradise by making it car hell really aren't rare here.

Nobody cares or even notices all the cars driving normally, it's the sub par people everyone notices. I think there's some minority of drivers who stopped getting any better once they got their license who soak up the hate from everybody. If that minority is say 5-10% it's basically guaranteed that literally everyone else is inconvenienced by one of them on every trip.

> a refreshment, please

> food or drink?

> why do I need to specify. I've been out all day in the heat. Bring me a few refreshments.

That seems an uncharitable view of the reply.

The search space is huge, we sometimes find needles in haystacks by accident, isn’t it exciting that we have tools now that can systematically check every piece of hay?

ML search is more about ‘averages’ based on samples.

Innovations like these are more about ‘shocks’ that surface fitting cannot capture.

Note universal approximation theorem applies only to smooth surfaces.

Not always. Quantile regression exists. And you can develop "no match" categories.

Quantile regression is also about averages.

Averages are formulated as measures of centrality in the L2 norm ("straight line" distance), sum(values) / count(values). Quantile regression uses modifications the L1 norm ("city block" distance); if median (50%) then it is a measure of centrality. Not everything is an average. If you're interested, this is a good (but math heavy) treatment: https://en.wikipedia.org/wiki/Quantile_regression#Computatio...

But the better the mean surface is fitted (in a generizable way), the easier it is to spot outliers.

Well said.

The punchline is inside the article:

> We at Johnny’s Software Lab LLC are experts in performance.

Comment not intended to come across as snidey, just tickled me a bit after reading yours and clicking the archive link.

This is true of ChatGPT 4 with the default prompt maybe but that’s just the way it responds after being given its specific corporate friendly disclaimer heavy instructions. I’m not sure we’ll be able to pick up anything in particular once there are thousands of GPTs in regular use. Which could be already.

But I agree we will probably very often recognise 2023 GPT4 defaults.

There's not really a problematic upper bound for global transaction volume on transactions per second with the lightning network. There are other problems in terms of adoption and infrastructure etc, but the actual potential for humankind to have a single currency global low-barrier payment system is there. But this is more like the early days of the internet. The internet seemed pointless to most people even as late as the 90's.

A future global system could have settlement on the blockchain. Like the current system does at far less regular intervals than the processing power of non-settled transactions.

> There's not really a problematic upper bound for global transaction volume on transactions per second with the lightning network

There unfortunately is, to open a channel you have to make a Bitcoin transaction, and you can't use lightning without opening a channel. Bitcoin processes a max of ~220M transactions per year so to onboard the world onto lightning with only one channel each would take a few decades.

A real layer 2 could solve it, or having some trustless way to use BTC on other chains

It's possible I'm too optimistic on this point. I feel I have a good grasp of how bitcoin works, but when it gets to layer 2 it gets fuzzier. The initial discussions sounded really promising, as they seemed to mirror HTTP on top of TCP/IP. And we all know how much of a slow burning revolution that turned out to be.

But http seems simpler than what I've read about lightning more recently.

Total wild speculation follows:

As for the channels problem, it feels to me like you have something like ISPs or AOL being the thing that's needed to kickstart it. And perhaps the ISPs or AOL in this world are the employers and lightning banks. So yes you need bitcoin to open it, but with an employer you have the potential to have an entity with bitcoin resources and with a reason to send bitcoin to an individual.

Perhaps HD wallets with lightning kickstarter funds apportioned to new employees. This feels similar to the needing a bank account or national insurance number in the UK (social security number? in the US) to start getting paid, or similar to needing an internet connection. Yes you could choose to get paid in bitcoin and have the fees come out of your wage, and this would be similar to getting paid in physical cash today. Or you could get over the slight roadbump to get you onto the lightning network.

What would a real layer 2 look like in your opinion?

A real layer 2 would look more like something built on Ethereum (can see all its L2s at https://l2beat.com).

Essentially it's a separate network that every few minutes takes every transaction and compresses it into a data blob that it saves on Ethereum along with a proof that the computation was done correctly. The Ethereum L1 nodes then only need to verify the proof instead of re-executing all transactions that happened on the L2.

With this design users can go straight from an exchange like Coinbase onto the L2 and never need to use Ethereum, and fees are 10x cheaper because of the data compression. Fees will soon be 100x cheaper as Ethereum is adding extra space just for these L2 data blobs that is much cheaper than normal Ethereum data space.

Unfortunately it can't be done on Bitcoin right now because Bitcoin nodes don't have Turing complete scripting and so can't verify the proof that an L2 posts to Bitcoin.

YouTube is one of those things that consumes more energy than bitcoin.

That is bad, but it doesn't seem part of the public narrative that YouTube is bad because of global warming. Also the general narrative posits that electric cars are good. They also use a lot of energy. There's nothing inherent about EVs versus bitcoin that suggests one uses clean energy versus another. They both use a lot of electricity.

But so does the banking sector. And entertainment, farming, travel and a lot of the other things 8 billion people choose to do or depend on.

Any usage of fossil fuels should rightfully be a target and held to the same standards.

There is another narrative that we need continual growth and that inflation aids this. And deflationary currencies are bad because they don't help growth. This seems so ingrained in modern politics and economics that I slightly hesitate to question it in public.

Well maybe unbounded growth and exponential usage of resources is the actual problem.

I hope we reach the social limit on unbounded growth before it destroys everything we have. It's literally in physical terms unsustainable. My only hope here is that nuclear fusion gives us a respite for a hundred years or so. But we'd still have the same problem at a different scale if we used that to grow to a trillion people.

I can see that a deflationary global currency could be part of the solution to the problem. And I can see why suggesting deflation is good can be seen as a threatening thing to those that have resources.

So I feel in that light, the bitcoin = global warming narrative is really convenient, even if a little ironic.

The current government is very unpopular, but they won a small by-election by opposing London's ultra low emissions zone. So they're now shifting to an anti-environmental stance in the hope that will save them in the next election (they were hardly pro-environment previously).

This is probably down more to the lack of overlap between people who care about carbon accounting and those who celebrate petty jingoistic "victories" like being the first nation in the G20 to achieve collectively agreed targets.

Often there's an archive.whateverTLD link in comments, there's one here:

https://archive.is/AHFlh

This government is leaning heavily into culture wars, they are unlikely to attach themselves to something they are painting as “woke” (cleaner air in school playgrounds, lack of smog, etc)

Needs to support DMCA takedown notices

I think they are saying they manually tested what they thought the automated process was doing, then found out their assumption about what it was doing was wrong. So the manual tests were moot.