teebs

I've mostly used it for reading textbooks or academic papers (and taking notes on them) and it works fine for me. It can be a little small but you can zoom in if you need

It's likely because GPU calculations are non-deterministic and small differences in floating point numbers could lead to different outcomes (either in the way you described or somewhere deeper in the model)

That's not how the First Amendment works. Cornell University is a private organization and it can choose not to do business with people if it wants to, provided that it's not breaking other laws by doing so. Students get suspended or expelled for speech all the time. (I would guess public universities can too - the First Amendment says "Congress shall make no law..." - not sure where its boundaries are drawn though.)

Do you have specific mathematical topics you're interested in?

I have tried reading textbooks a few times to teach myself but found it hard to stay motivated, so I found a tutor on Upwork to assign and grade homework problems and answer my questions. Along with math textbooks and YouTube videos, this has been super helpful for fleshing out my knowledge of college-level math that I never learned properly. It's also great because they can go at the pace you want and focus on topics you find challenging, interesting, or useful.

I like yoga but I don't feel that it's a super social activity. Maybe I just go to the wrong studios. Do you find that you talk a lot with other attendees? Or do you think the yoga itself is social?

Thanks for posting this! I found myself feeling quite lonely at the end of a 5-year relationship a few years ago and have spent a lot of time in the past few years trying to work through loneliness. I can't compare it to your situation - divorce with children must be really difficult - but I think male loneliness is a super important problem in our society and unfortunately it's something I had to put in a lot of work to address.

> The thing is I don't even like hanging out with people most of the time but I get into this space where it feels like a biological necessity.

Connecting with people is a biological necessity! It takes work but it's totally worth it.

I definitely agree with others that therapy is a good place to start. Also, it's okay to feel lonely. It's also okay to want to connect with people and to put work into it.

A few things that have worked for me:

- Just because older friendships have withered doesn't mean they can't be revived. Reach out to people to catch up - everyone loves feeling wanted every once in a while. Most of these won't turn into close friendships, but some might. And even if they don't, seeing old friends will make you feel more connected to more people and make life more colorful. The world is a lot more interesting when you hear more people's perspectives, I think.

- Join a club. Or join multiple clubs, even. A lot of people are looking for social outlets as pandemic restrictions recede. I like running so I joined a running club, but I'm sure you can find other clubs to join. There will probably be a lot of new members like yourself wanting to socialize - my running club is 20 years old but most of the active members joined since June 2021.

- You might consider finding roommates. Living alone is nice in some ways, but having roommates gives you built-in social interaction every day (as long as they're people you can tolerate). IMO more than one roommate is best.

- Make friends at work. If people don't eat lunch together, try to invite people to lunch. If they don't hang out after work, try proposing going to a bar one day. This might be higher effort than other ideas here, but I usually find that having casual work conversations fills at least a little of my social needs each days. (Obviously this doesn't apply if you work remotely.)

- Be vulnerable. You might be worried about others judging you for feeling lonely, but I don't think they will. You can express it to them in less direct terms, too - maybe "I feel like I've lost touch with my friends and I wanted to see how you're doing" or "I've been trying to rekindle friendships since the pandemic/the divorce." I had a conversation with a high school friend just a few weeks ago just like this, in fact. I think being vulnerable about things in my life (including but not limited to loneliness) helps deepen the relationships I have and makes my social interactions fill more of my emotional needs too.

For me, a mix of all these things have helped. I still feel lonely sometimes, but I feel a lot better than a few years ago. It's a lot of work to build friendships and a social structure, but it's well worth it for your happiness (and it'll make your new friends' lives better too).

You can just do that by adding large hidden layers or more hidden layers, up to a point. But eventually the signal that's coming from the input data is so diluted through all the neurons and layers that your models stop performing better as you add more neurons. Many of the advances in NNs come from structuring the neurons in particular ways - for example, in computer vision, convolutional neural networks. These are kind of a small neural network that looks at each part of the image and gives an output - basically, shrinking and summarizing the image - which is then fed into another small neural network that shrinks & summarizes some more, and so on until you have only one result, like "is this a cat" or something like that. Transformers, which are mostly used for natural language processing, have small neural network layers that let the network figure out which other words in a text are relevant for understanding a given word (or, token, really). These help simplify the problem for the NN so it doesn't have to sort through all the data it has at once, which is a bottleneck on scaling NNs.

Not sure that explains it like you're 5, but hopefully it addresses your question

It's kind of funny watching SGD slowly get closer and closer to you, and then overshoot. I was surprised that it's so much slower than me at learning this task. But then again, my brain is probably devoting a lot more computing power to solving this problem than it is so it's not really a fair fight

> This gets to the other mistake Diehl makes in that article, which, ironically, echoes a similar mistake made by many crypto absolutists: there is no reason why the Metaverse, or any web application for that matter, will be built on the blockchain. Why would you use the world’s slowest database when a centralized one is far more scalable and performant? It is not as if WhatsApp or Signal are built on top of the plain old telephone service; they simply leverage the fact that phone numbers are unique and thus suitable as identifiers. This is the type of role blockchains will fill: provide uniqueness and portability where necessary, in a way that makes it possible to not just live your life entirely online, but as many lives simultaneously as your might wish, locked in nowhere.

Halfway through this paragraph I thought Ben was arguing that phone numbers are a centralized service that shows that blockchain isn't necessary. Instead, he's arguing that blockchain will create "unique portable identities." This is an old vision that many companies have been pushing for a long time - "log in with Facebook" "sign in with Apple" etc. Today, we've pretty much settled on phone numbers and email addresses as a unique, portable identifier (plus accounts from big tech companies). You can even easily create new email addresses if you want multiple identities. Why do we need the blockchain to solve this problem either?

Another example of the same thing: he says that it doesn't make sense to use LLCs, which are built for the physical world, in the digital world. He says that the digital world can use DAOs. But, of course, "Stratechery" isn't a DAO - it's an LLC. Meta is an LLC. LLCs are already non-physical, even if they're not purely digital. That's precisely why LLCs were invented - to decouple capital from physical ownership. Why do we need DAOs when we already have LLCs? Is it just that you don't need a lawyer to make a DAO?

Ben spends a lot of time in this article arguing that blockchain can be better than centralized services for some purposes - but then fails to argue what those purposes are, in my opinion.

To me, conflating the metaverse/VR - which I think is overhyped but has some obvious value - with blockchain - which is a lot harder for me to understand - is a mistake.

I agree with you that it's definitely about who you know - if you go to Stanford, you can get better jobs out of college compared to someone with similar skills who didn't. But I worked at a FAANG (or whatever it's called now) company and many of the senior engineers we hired were from small/less prestigious companies. These were people with ~10 years of experience who had become great software engineers but had never considered themselves in a class to work in big tech. Then, when given a chance, they got an offer and often accepted it because it paid so much better than their old job. If there weren't such high demand for software engineers, the company would just have hired from other big tech companies since it's safer, but they were forced to try to find people in unexpected places.

On the other hand, these people were still negatively impacted by their backgrounds: they hadn't had the chance to join fancy tech companies out of college, which meant they missed out on several years of career opportunities. They were more likely to be hired at a Senior Software Engineer level, for example, when someone similar from a big tech company might be hired at Staff+.