Tumblewood

The US's economy is definitely not "utterly dependent on tourism".

Yes, frontier models can reason outside their chain of thought and manipulate their chain of thought to some extent. The system card for Astra writes:

> We have found that GPT-6 Astra is more capable of controlling its own CoT than GPT 5.6-Sol, and less likely to include incriminating information in its CoT. In adversarial settings (where we push the model to evade our monitors) we find that the model is able to remain undetected when strategically underperforming in evaluations (sandbagging) and can sometimes evade our internal monitors when asked to perform certain sabotage tasks

Your examples are things that most humans cannot do, or things that AI can already do. For example most humans, even most intelligent humans, could not write a well-received book, run a successful company, or make a popular game. On the other hand, AI can absolutely sort through research, draw conclusions on complex topics, and argue them persuasively. Likewise, I don't know what you mean by a "decent" conversation, but millions of people converse with chatbots daily, so I don't know why you say AI fails to meet that bar.

They researched on a 6.9B parameter LLM. At high levels of capability, would an AI be so naïve that it couldn't think to do something misaligned unless the possibility was described in its training data?

Not the person you're replying to, but I see the same thing happen in the Capitol Hill neighborhood of Seattle. Dense neighborhood with a lot of nightlife, but many of its residents exclusively use free street parking to park overnight. There is a lot of contention for spots after about 7pm.

This sounds a lot like one of my favorite games, TagPro, aside from the difference in scale. It has a very tight-knit community, brought together especially by its several community-run competitive leagues. There was even an IRL meetup recently. Sadly it doesn't have anywhere near the marketing budget to become as big as CS.

they do allow you to switch - maybe it refers to google paying to be the default search engine?

that's more about categorization than perception - some cultures don't have a word just for blue, but they still very much perceive it. it's just lumped into the same category as green ("grue").

This line of reasoning refutes pie-in-the-sky doomsday narratives that are extremely unlikely, but the case for AI extinction risk justifies a relatively high likelihood of extinction. Maybe a 0.0000000001% chance is worth ignoring but that's not what we're dealing with. See this survey for the probabilities cutting-edge AI researchers actually put on existential risk: https://aiimpacts.org/2022-expert-survey-on-progress-in-ai/#...

Some of my favorite pairs from the big list:

misrelation / orientalism; superintended / unpredestined; incorporate / procreation (don't mind if i do!); predators / teardrops (a cause and effect); counteridea / reeducation (a bit synonymous); streamlined / derailments (quite opposite!); truculent / unclutter; colonialist / oscillation; renavigate / vegetarian; persistent / prettiness; paternoster / penetrators (hmm); obscurantist / subtractions; nectarines / transience (a story of ripeness); definability / identifiably; indiscreet / iridescent; excitation / intoxicate; discounter / reductions (how logical!)

One small suggestion I have: add a point for pairs with different starting letters, and another point for pairs with different ending letters.

You can filter a discord search by what channel the message is in.

The main point against it for me is not that it is a new drug; it is that, if I reduce my caloric intake, I won't have enough calories to maintain my normal daily activity levels.

Most of their cost comes from API accesses to other services (e.g., Bing), so unless they can replace those API calls the price cannot go down very much.

The two passages are not, in my opinion, similar enough to insinuate copyright infringement. The only similarities are the general melodic contour and some (not all) of the chords. Even though the overall shape of the melodies is similar, they do have many differences, to the point where I would not say it is 'copied'. The chord progression under them is different, but most clearly the groove and delivery are not at all alike between the two pieces. Even though both Shape of You and No Scrubs are hit songs, I had never noticed the similarity until this. If this standard was enough to find copyright infringement, you could nail every third song on the charts somehow.

Feather mining (the style that uses the 1x1 branches) is more efficient per block mined, but it requires a lot more attention compared to a straight 2x1 hole because of all the turning you need to do. So 2x1 is what most people do. People who have enough resources to make beacons usually speed-mine instead, though. That's because with a beacon giving you Haste II and an Efficiency V pickaxe, you can mine stone blocks instantly, allowing you to clear out 2-tall spaces rapidly. This destroys more blocks per ore but is much faster than anything else in terms of time. That said, in the 1.18 update, diamond ore can only be found at levels where deepslate generates instead of stone. Deepslate cannot be instant-mined, so caving is once again the most efficient way to find ores. I consider this more fun than the old style of mining, so I'm glad they made this change.

Advertisement is mostly a zero-sum game, competing for a limited amount of spending. If one player in a market advertises, they gain while the other players lose roughly the same amount. If the advertising world kept operating as normal, but the ads never displayed anywhere—blank sidebars, blank sponsored content, blank billboards—it would be more pleasant for consumers, and the market as a whole would not be any worse off for it.

Then why stop there? No one needs to make the ads. Everyone employed to design, shoot, and display the ad can serve others instead of working to manipulate them. As long as the transactions keep going as normal, it's a huge win for everybody.

Of course, this is business. Corporations wouldn't spend that money just to subsidize newspapers, television, and page views. In a world without advertisement, tons of content would suddenly lose its revenue stream, without an easy way to monetize attention. Attention would not be profitable on its own.

That's a negative thing on the face, but I believe content which cannot survive by donation or by subscription deserves to die. Netflix survives already without (third-party) advertisement; Facebook, on the other hand, would collapse. But a social network does not need as many engineers as Facebook does. Without competing against ad-based sites, a less bloated, less attention-sucking social network could replace Facebook. And so on for other sites. SEO blogspam would die, whereas useful and interesting newsletters would survive (as demonstrated by Substack). High-quality TV could survive, but game shows and reality TV would be forced to greatly pare down their offerings. Valuable, accurate subscription news would survive, whereas clickbait news would have no more reason to bait clicks.

The alternative to the status quo is that many forms of content would become unprofitable, and I don't think that's a bad thing.

This is a poor analogy. Besides the distinction that arson is intentional rather than negligent, arsonists burn things that would not have burned otherwise. While PG&E's negligence sparked the fire, the brush in the area would have burned at some point in the near future regardless of their actions. The Twitter thread mentions the fire that started in the area around the same time because a tree fell, or some other inciting event could have happened. While PG&E should be held accountable for failing to uphold legal standards of inspection, the Camp Fire or a similar one was likely even without negligence, so to compare PG&E to arsonists is misleading.