layman51

Same here. When I use the Gemini app for iOS, it says that I’m on “Ultra” in the left sidebar near the bottom. It doesn’t say that when I’m using Gemini on the web.

This isn’t directly related to the article but if Jared Cooney Horvath argues that learning “depends on sustained attention and interaction with other people”, where would that place some of the newer schools that rely a lot on AI to teach and hire human guides or coaches instead of teachers?

I was running into something similar where I had it working on a single Python file. Maybe part of the issue is that I had memory turned off, and the conversation might have spilled into two conversations. I wasn't using version control either, but since it was a single file with about 1000 lines of code, it didn't seem too hard to eventually get all of my features back.

I have heard of Step 3 too, but shouldn't we bit at least a little bit concerned about the fluoride calcifying our pineal gland? I know some people might take my last sentence and think it's woo-ey but if you search it up, basically it seems like there's nothing conclusive one way or another.

If Google’s top choice is Alice, and Alice’s top choice is Google, then what will happen in the algorithm is that Alice will see an offer from Google and lock it down. Alice could get multiple offers but she will take the best one which is Google.

When Alice and Google are not paired up, then they will always form a “blocking pair” in any matching where they aren’t together so that means any matching that doesn’t have Alice and Google paired is not stable.

I think the confusion is that Alice only has one possible stable employer and thus her worst one happens to be her best one.

If Alice’s top choice wasn’t Google, but some other company, then there might be multiple employers she could be matched with to form a stable pairing.

This algorithm is definitely weird too though because it’s possible for some candidates to be paired with a company they don’t really like, but it’s considered a stable situation because the other companies don’t want them more than their current employees.

I guess in real life, you could try improving yourself and then the matching becomes unstable again if preferences change.

These notes are pretty cool and interesting. I would like to try out for myself since I find it difficult to get into the flow sometimes.

I think it can be a bit easier to think of the algorithm in terms of job applicants (who receive job offers) and employers (who propose job offers). And then due to the algorithm, it must be the case that the number of applicants is equal to the number of employers, and each employer is only trying to fill one position.

An applicant (the receiver) will end up with a "worst" match in the sense that if you look at all "stable" pairings (there can be multiple configurations), the applicant is going to have the least favorable one among the different pairings.

There's an important concept of a "blocking pair", and the Gale-Shapley algorithm is trying to eliminate all blocking pairs, and you could say that pairings are stable when no blocking pairs exist. A blocking pair is a job applicant and employer who aren't matched together, but both prefer each other over their current match.

Suppose that after the algorithm is done with its work, Alice doesn't end up with a job offer from her top choice, Google. This must mean that Google never offered her a job because Alice was just too far down on their list, and Google made a deal with someone else they like better than Alice.

This line of reasoning makes sense, but also I would say that a lot of technologists who are interested in education realize that they have to design something that wouldn't just give the answers away because that's not the point of learning. That seems like it can be pretty difficult to accomplish with LLMs because I bet if a student keeps pestering them, they will give up the answer or give a clue that makes the problem way too obvious.

So usually the claims about replacing tutors like in this article come from a place where they usually give a caveat that maybe for highly motivated and ambitious students, they can figure out how to accelerate their learning with AI. But then that undermines the whole argument of transforming education if it wouldn't work with students who are average or below-average in their academic achievement.

I’ve pondered this too. I was in a group meeting with a consultant who is upfront with clients that he has started to use Claude heavily to speed up his output. There are probably a lot of clients who would still pay for his services because maybe they still appreciate his input or the relationship and don’t want to go through the pain of figuring out how to replace him with an LLM. It has been my experience that some teams would rather just subscribe to some software or platform even if there’s a free alternative.

I’m not allowed to enter legit chess tournaments where if I’m going be using my phone’s chess engine app to tell me the best move in every single position. Because that’s considered cheating by the chess players and tournament organizers and rightly so. But, there doesn’t seem to be an incentive to not do something similar in the world of work unless your LLM output is so bad or obviously inaccurate.

> A library with 100,000 visitors in a town of only 9,000 people

That reminds me that there is a library that I believe became a bit viral on social media and now it seems like it receives so many visitors to the surrounding area and it is still quiet, but it is funny because it seems way more crowded and there's way more movement that I remember in the past.

That's why with this post, I sort of wonder whether this is something that these towns really want to have. I am under the impression that in Japan, they generally want visitors, but then they probably end up with issues from people not knowing what to do and even behaving badly.

Well just as a regular user, I think it is pretty annoying because if I look up anything on Google while in Incognito Mode and hover over a search result, I can see that maybe the top result is maybe Wikipedia, or Instagram, or some other less-known website depending on what I'm searching for. Now, that's all very obfuscated because I don't actually know where I'm going to land for sure.

This reminds me of the time an AI was taught how to play a racing sim game (Gran Turismo if I remember correctly). The AI was able to race its car very well, but it took a lot of risks that a human player probably would not. A human player might be able to copy the approach the AI took, but they would probably crash.

Going back to chess, I think the situation is similar where you can’t expect an amateur player to get better by trying to play like a strong engine. I think even professional chess players mainly use engines to prepare or memorize variations that are counterintuitive for their opponent. In other words, getting into situations that look wild, but that part of one player’s preparation.

I’m not sure how it is in math, but in chess, it seems like top players can play just like engines when they are in “normal” positions, so that is where I get a bit confused as to where the direction of insight is coming from because it’s been my view that AI is able to make leaps that we would never think of taking and I’m not sure that anyone could actually learn how to do that on their own unless they were willing to keep failing over and over.

I consider myself a slow reader so I am interested in this pacing rule. Do you remember how the pacing was enforced? Were there like minute sandtimers or was it just kind of like a 30 minute timer and you either met the pace or didn't?