aprentic

I'm not sure we actually want to bring back Jon Draper.

https://en.wikipedia.org/wiki/John_Draper#Allegations_of_sex...

Presumably both of these AIs have additional information beyond what was in their initial training.

The hotel agent probably has a RAG that points at their various customer and inventory databases. The user agent has individualized information about the customer. Both of them have also likely had SFT steps that further differentiate them.

The interesting question is if Gibber-Link lets the AIs do something they couldn't otherwise do with natural languages. Does it lower some error rate? Does it reduce the time it takes to send messages? Does it effectively give the AIs additional vocabulary?

If I had to guess at the internals, they probably took the token encodings and mapped them on to tones. Then it just throws text or audio through the decoding filter and passes it back.

If that's the case, the benefits are probably limited to slightly faster communications (It's essentially a simple, lossless compression) and a slightly lower error rate (beeps are easier to correct in noisy environments).

It's clearly not just about the chips. China didn't attack Taiwan before 1987, when TSMC was founded.

There are many disincentives. The biggest one is that China is confident that they'll get full control of Taiwan without violence.

They believe that the US is in a state of terminal decline and Europe will never overcome centuries of infighting. A significant portion of the population of Taiwan wants to reunify. They fully expect that they can just continue to nurture supporters within Taiwan and wait out the West and Taiwan will just fall into their lap.

Why would they go to war when they think they can get everything they want without war?

Exactly.

The US has unrivaled force projection capabilities; half the worlds carrier fleets, the largest navy by tonnage, and the technology to execute unprecedented combined arms maneuvers.

China has insane defense-in-depth; more missiles than you can shake a stick at, the largest navy by number of ships, a vast arsenal of countermeasures, and a 4:1 population advantage.

China can't stand toe-to-toe with the US anywhere except the immediate vicinity of China. In that vicinity, nobody can get close if China doesn't want them to.

They're extremely careful not to state that as their plan.

In general, they make a huge effort not to talk about their plans vis-a-vis Taiwan at all. They just keep repeating that Taiwan is part of China.

The closest they come to stating that they plan to use force is that they'll sometimes say that they won't reject the use of force.

Given that nobody has proposed a scenario where China actually could do something like "take the IP by force" (since the IP would be gone if they ever tried to invade) and we can generally see that the Chinese leaders aren't complete idiots, it seems highly unlikely that they're planning an invasion any time soon.

This feels like somebody telling Bob Vila, "This house needs a kitchen renovation."

It might actually need a kitchen renovation but you can't actually do it until you repair the foundation under the kitchen. The bathrooms might be a higher priority, you don't have any tools, you just kick out all the contractors, and the next meeting with your parts suppliers is in the courthouse.

China has repeatedly demonstrated 2 things WRT Taiwan. They're patient and they're serious about their red lines.

I can't recall a single instance where China announced anything about Taiwan that wasn't reactive. They just keep repeating the "one China" policy.

Their official stance is that there's no need to invade Taiwan because Taiwan is already part of China and they reserve the right to use force to enforce their territorial integrity.

The practical manifestation of that policy has been that China and the US both get to pretend that their view on Taiwan is the reality and nobody will do anything if the other side doesn't rock the boat.

Their red line is a formal declaration of independence by Taiwan. As near as I can tell, all but one of their "military exercises" has been in response to actions that get close to that line in diplomatic terms https://globaltaiwan.org/2024/10/chinas-military-exercises-a...

During many of those exercises effectively blockaded Taiwan. They did that for a week after Pelosi's visit and they experienced no pain in response.

I draw 3 conclusions from these observations:

  1) China will not invade Taiwan without some external stimulus  
  2) China is prepared to blockade Taiwan in the event of any attempts at secession  
  3) China has established that secessionist behavior is casus belli for a blockade in the eyes of the international community

Theoretically, yes. It's pretty unlikely.

The US alone currently has more than 2x the military expenditure of the entire EU. The US also has a larger GDP than the entire EU.

The US supplies the main bulk of Ukrainian military aid. https://www.ifw-kiel.de/topics/war-against-ukraine/ukraine-s...

The EU would have to make some very serious budgeting changes if they wanted to fill that void. A bunch of EU politicians would need to make the case for deep budget cuts, tax increases, war bonds, or some combination of those.

"Whataboutism" was originally a valid criticism of "tu-quoque" attacks.

Now it's frequently used as a defense for hypocrisy.

There are problems with PPP but it does try to address some real (no pun intended) issues with nominal GDP.

A US resident might be able to afford a lot of Chinese goods and services but can't get many of them. For a Chinese consumer, the average price of goods or services is less relevant than the price of the subset of goods and services that are actually accessible to them.

It attempts to address the very real problem that a Chinese resident can buy a Chang Li truck for around $2,000 but an American consumer spends over $9000 once the realities of international shipping are factored in.

A theoretical economist might abstract these differences away and just claim that arbitrage will eventually balance out these price differences but empirically we can see that's not what actually happens.

That’s old data. China’s GDP is larger than that of both the EU and India.

The college numbers have been changing too https://www.axios.com/2023/05/08/chinese-students-us-educati...

I'd say the area where cooking videos particularly shine is technique and visual assessment.

It's much harder to describe the proper motion in text than to just show a video. A still image of dough doesn't show you the elasticity as well as a video does.

Areas they're particularly bad for is anything involving data; recipes, ratios, times. Those are all much easier to see in text.

They're both pretty bad at teaching "feel". There's no way around tasting and poking at any recipe you're not familiar with. Neither videos nor text will teach you how to do that. They'll tell you it's important but you really just need to do it a bunch.

That's kind of like asking any variety of, "Can X videos make us better at X?"

Absolutely, but the more relevant question is if a particular cooking video or series, out of the vast set of "cooking videos" will make a particular viewer a better cook.

Probably not. I'd suggest a few questions to determine if that's true.

What is the specificity of the video? Some "cooking" videos only pretend to be about teaching cooking but they're actually about something entirely different. I love watching old episodes of Iron Chef but I don't expect to learn anything practical. Some cooking shows are about a particular aspect of cooking; a video on cooking with Spam won't make me a better cook because I don't cook with Spam. A video on how to make cakes look attractive with fondant won't make me a better cook because I like food made of food ingredients rather than construction materials.

Does the author know their subject matter? Many videos tell you stuff that's just plain wrong, or even dangerous (don't make popcorn on a soda-can stove with methanol or it could blow up in your face).

Are they level-appropriate? When I left for college, my mom gave me a copy of "How to boil an egg". I got a lot of benefit out of that book but something similar would be useless today. By the same token, years later, we got a lot of great information of of Rose Levy Beranbaum's "The Bread Bible", but it wouldn't have helped me at all as a freshman with a communal kitchen and minimal cooking ability.

There's also the question of what the viewer is going to do with it. You can watch videos all day long without getting any better at cooking. If you want to get better at cooking you primarily need to cook. Some videos may aid that effort but even the best ones will only help to the extent that you're willing to implement what they teach and practice it.

This is a well known effect and there have been a number of attempts to address it.

Slashdot went through a number of iterations. They initially added a points system and ended up adding a meta-moderation layer to try to address biases in the previous layers. https://slashdot.org/faq/mod-metamod.shtml

Reddit lets viewers choose from a number of sort algorithms that allow them to emphasize posts other than those most heavily upvoted.

Lemmy uses a log scale to try to bring more voices to a conversation after the initial "New" votes come in. https://join-lemmy.org/docs/contributors/07-ranking-algo.htm...

They're all trying to address the issue by offering sort orders which should show the "best" posts at the top. The fundamental problem with all of these is that they assume that everyone agrees on what is "best" and that small set of sort options will allow the majority of people to choose one which matches their ranking of "best".

We've also seen several large companies provide "algorithms" which provide customized ranking lists for individuals. The glaring problem with these approaches is that they rank by what is "best" for the company, not for the individual.

We've been missing is a way for people to choose sort lists that reflect their own views of what is "best". Otherwise the best we'll be able to get is some sort order that some plurality agrees on, even if that plurality isn't even a majority.

This was actually the very first statistics project I ever did. We had the basic high school level "Probability and Statistics class". We learned some simple combinatorics, some formulas for the normal distribution and the basics of hypothesis testing. The we were told to do some project (it was junior or senior year so they gave use fairly wide latitude).

I wrote up a survey of people and distributed to classmates who had either taken the SAT more than once or had taken the PSAT and then the SAT. I asked them if they took a prep class and what kind.

The results strongly supported the meme.

There was a big gap between people who did and did not take prep classes. There was a small but statistically significant difference between the Princeton Review and Stanley Kaplan (I don't remember which one won) and there was a large gap between people who took group classes and those who had private tutors.

I didn't collect any demographic information but it was at a fairly expensive private school in Manhattan. Mostly white kids. The minority kids were mostly from prep-for-prep.

There’s Samuel Slater (the traitor). https://en.m.wikipedia.org/wiki/Samuel_Slater

He was the most spectacular but hardly an isolated case. https://hbswk.hbs.edu/archive/trade-secrets-intellectual-pir...

Do you know if any of the "delete my reddit posts" apps support this?

Maybe even an option to keep it running and randomly update posts with GPT nonsense over time?

That would be a reasonable hypothesis if our information was limited to a few isolated inconsistency issues.

The more posts we see deleted the less likely it is that it's an accident. I just checked and found out that posts from an account that I scrubbed and deleted 5 years ago are popping back up. That stretches "eventual" pretty thin.

We also have a large amount of external information that suggests that reddit is not acting in good faith, much of it from Spez himself.

Between the two of those, malice is more likely than stupidity to be the motivating factor.

Apple is an interesting company.

They didn't invent computers, mobile phones, mp3 players or tablets. Instead they tend to take products where the whole industry is essentially still in alpha, polish it to a mirror finish and then show the world how those products will make their personal lives better.

Computers don't just have to be for big companies running accounting software; they can unlock the creative potential of writers and artists. Mobile phone aren't just for drug dealers and business executives; they let people share their lives with their friends.

There were a few misses too. The Magic Mouse looked sleek but was mostly annoying and didn't really do anything that other mice didn't already do better.

I'm afraid this headset falls in the same category. This seems to have marginal improvements over the HTC Vive I bought 6 years ago. The resolution is better, it looks lighter and less bulky. It also gives you pass through mode for an AR experience vs just a VR experience. It looks like a good headset.

That's just not enough. Not only does it fail to meet the high innovation standards that we've come to expect from Apple, it fails to meet the standards that we already expect from AR.

Unlike with all the previous hardware innovations, we already know what we want from AR. Everyone who has played modern video games is already used to various forms of HUDs and we've seen many iterations of what works well and what doesn't.

When I put on an AR headset I want all the things I've seen in video games. I want name indicators. I want automatic highlighting of relevant objects. I want to replace my map with floating directional indicators. I want to see my health stats floating off to the side. I want to be able to make all my friends wear silly hats. I want a virtual cat following me around.