Rich people? Lol. At least people are transparently saying their hateful nonsense these days. Yes, it’s people with large incomes that are making people take fentanyl, shit on the streets and stab people to death. If only rent was 50% cheaper, that guy would be a productive member of society! Hey, I have a great idea. Let’s have the government step in, take that stupid rich guys money, and give it to the drug addict. That’s never been tried before and I’m sure will solve the problem completely…
English is my first (and only) language and I’ve never heard of this. I would say that it not only doesn’t sound weird, but is totally normal. Who told you this isn’t correct?
I’m confused—you’re claiming the above are radical leftist policies? They seem like (mostly bad) moderate left policies to me.
Here are things I think are radical, and my gauge for that is asking if it could reasonably appear on a BLM demands list. Pushes to defund police, stop enforcing laws (or as Chesa did, enforce weakened versions for fear it would result in “dangerous” deportations) or direct repeat criminals to programs that allow them to avoid jail/records, giving homeless cash and a place to do drugs “safely”. Not to mention the new ask for $5M per black resident for reparations, with some claiming it doesn’t go near far enough.
The idea that these problems are a result of high cost of living and NIMBYism are absurd. When progressive policies fail, the left still find ways to blame it on “the rich”. Marxism 101.
Really? So there isn’t open and pervasive drug use? All these videos of herds of people robbing stores, and homeless people straight up saying they can use/sell drugs and commit petty theft without consequence are just lying?
They are one and the same. Chesa Boudine was largely responsible, and ran on the platform, of essentially not enforcing laws on drugs and theft. His parents spent decades in prison for murder as a result of their radical leftists activism with the Weather Underground. SF is basically the poster child for putting radical leftists in charge.
Are you being serious? SF is as much, or more, known for radical left wing politics than it is tech. They basically give homeless people free money, make it legally hard to clear camps, and don’t enforce laws of open, hard drug use and petty theft, leading to the obvious consequences. Basically they make it easy to be a drug addict on the street, and hard to be a police officer, and then wonder why crime and homelessness flourish.
One question that’s always top of mind when I hear about the insanity of SF: why on Earth are people with radical, proven ineffective policies voted in time and time again? Whenever I hear from tech people, they think it’s crazy. So if tech has overrun SF, why haven’t the politics become more moderate? At least one of my assumptions must be wrong.
I don’t think that’s quite right. Lower classes also have debt, and inflation effectively reduces your debt. In that case, it’s a direct transfer of wealth from lenders to debtors.
Anyone that has paid any attention to ChatGPT, or LLMs in general, knows that this is absolutely par for the course. There’s no reliability that you get correct answers with current training methods. They are optimized to create plausible (from a word-statistics perspective), human sounding sentences. They’re also overfit on an incredible amount of data, which is why they are so large—-they memorize web scale data. But we can’t tell the difference between a correctly memorized fact and an incorrectly conjured one. Until someone figures out, and implements, a good way to wire these up to an external fact data store, I don’t see a way around it. But I also don’t think that’s a very hard problem to solve and will likely result in far smaller models. I suspect companies launch these products knowing these deficiencies because right now it’s about market share and UX, and the technical ML problems will be confidently solved. Especially if you already own a search engine.
Yes, you’re right and I’ve been obscene and annoying. I’m sorry. I have kids and know many parents. I think it’s odd to use these ancillary reasons as motivations for having kids because the change to your life is so huge and profound (and the implications of creating a new human are even bigger!), so that the idea that a person would embark on that as a mechanism to solve their social problems is a baffling way to look at it to me. Words can’t even describe the profundity of the impact. It would be like joining the military to get green clothes. Also, it doesn’t even work. Sure, you’ll talk to other parents at bbqs and kid birthday parties. But you mostly won’t have time for a meaningful social life, because it’s work, kids, bed for many years. And the social events you do have are severely constrained by the presence of kids. You just wait to find out how much nap and eating schedules dominate your life, and how meaningful of a conversation you can have with these parents while you’re both parenting. Not only do you have little social time, but you will not having YOU time and will lose a big part of your current identity as a result. A major part of your life is redirected into serving the needs of your kid, and you and your pleasure, and your social needs and whatever else go on the back burner. So forgive me for being snarky, but this strikes me as overwhelmingly absurd. And also I probably have low emotional energy due to taking care of kids after working all day…. But yes, fast forward 10 years and I’m sure it’s a different picture. But that’s quite the investment. I’d just join a trivia group or something.
What “data driven” normally looks like: business people ask questions that are not answerable, but the DS team often goes off and tries to answer them anyway. I’ve been in so many meetings, in multiple organizations, where a senior leader asks a question like, “why did this number go up? We need an owner to deep dive this.” Then, someone disappears for a few hours or days and comes back with some nice narrative that more or less might make some sense, and is directionally consistent. Then, best case, they go “hmm, ok” and worst case, they say it’s an interesting view and we need to add it to the tracking. Then, move on to a new question + action item, rinse and repeat. And then, 6 months or a year later, some brave soul goes, “geeze, there are way too many things to look at here, can we streamline this reporting to focus on the handful of most important metrics, and then assign owners to review the others and surface anything interesting?” And then all those things people asked for over time get stuffed into an appendix or deleted, and the whole thing starts all over again.
I don’t recommend the kids one. You might get to talk to some people, but only about kids, and only in 10 second intervals before being interrupted by said kids. Also, when you go back home, you still have to take care of them.
How? Stable Diffusion v1 uses, for example, the off the shelf CLIP model. The hard part is getting the dataset and something that’s functional, and then the community takes over and optimizes like hell to make it way smaller and faster at lightning speed.
The same will probably happen here. Set up the tools. Get the dataset. Sew it together into something functional with standard building blocks. Let the community do its thing.
No, that’s not the argument. I’m calling out your false comparison that positions it as if modern tech workers are in some sort of proletarian struggle against their oppressors. I’m sorry if being in one of the most lucrative and cushy jobs in history still leaves you feeling like there’s a big hole in your life. That’s an existential crisis. That’s not corporate oppression. You’re not in the same situation as oppressed workers of the past.
Barf!!!! To compare a 2023 Google employee to any population that’s had “boot stomping on their faces” is so out of touch with historical reality that it’s disturbing, disgusting and laughable all at the same time. We’re not talking about kids cleaning chimney sweeps here, or even modern slaves digging in cobalt mines. But Google employees, where 6% were allowed to go find another lucrative position and are sacrificing their massage services? Tone down the hyperbole sir.
He seems like the slimiest CEO imaginable. I have no data to back this up, just his constant woke clamoring, before woke dominated the culture, and the way he looks. My gut just instantly sets off alarm bells.
I recently spoke at a conference, and had some interested audience members approach me after to discuss more. One said he really valued the content and asked if I had published a paper outlining the same. I replied that I had not, but I did write about it in a blog. He said, “Ah, that’s a shame,” and walked away.
I think the point is that it’s hard to scale this. I’ve worked in the fraud space for awhile and a big problem is sophisticated actors that scale their operation and cause a lot of damage. If you can make it more expensive to scale, you can change the economics. For example, if it cost 1 cent to send an email, you would probably solve a big chunk of the spam problem.
I would be cautious about going down this road. For context, I have YT channel with >12k subscribers that is also for technical content. When I was similarly early on, I found myself in a similar place, thinking a lot about growth and metrics. I came to realize that I was playing the game that YT wanted me to play, to their benefit.
Since chasing "more" never ends, and every surpassed hurdle results in yet another one in the distance, it can quickly lead to burnout and asking yourself questions like, "why am I even doing this?". Worse, focusing too much on the metrics can cause you to do things that are ultimately counter-productive. For example, making a ton of videos optimized for clicking so that I could get 1M subscribers and monetize with ads is clearly a thing YT would like me to do. But to do that, you lean a little more into click-baity things, and start choosing topics based on their likelihood of being successful from a metrics perspective. And since frequency matters, you maybe cut corners on videos because it often makes more sense to put out more of lower quality than less of high quality if the objective is maximizing metrics. But then what are you left with, and are you proud of it?
If I could do it again, I would spent more time on answering the "what's it for?" question, and then ignore the metrics unless your answer ends up being, "To become the Mr. Beast of X". In my experience, the best benefits have been 1/ the serendipitous view that impresses someone and leads to a real-world opportunity and 2/ pride and reputational benefits in having the best video on a particular topic that then becomes the standard, and adds lots of value to lots of people. Those benefits are realized by taking care and making the best videos you possibly can, not by optimizing the metrics.
There is plenty in this article that is just wrong.
1. GNNs are no more "sequential" than CNNs and are therefore just as parallelizable in this respect (caveat below). A single GNN layer simply aggregates the features of the connected neighbors, just as a CNN aggregates the values of a nearby pixel. This can be parallelized across the nodes/pixels. The next layer depends on the output of the previous layer and is sequential in that sense, but that's true of all forms of neural networks. If other architectures have "won the hardware library" relative to GNNs, it's because GNNs depend heavily on sparse*dense multiplication. The real thing that makes it hard to parallelize is that you have to partition the graph intelligently when splitting across machines because there's a computational dependence between nodes, and you don't want connected nodes to be on different devices. In the metaphor with CNNs, that would be like needing to split a single image across multiple machines and still carry out the convolution operation.
2. It's not true that pre-training doesn't work. It's very common to use unsupervised/self-supervised pre-training to e.g., get node embeddings, which are then fine-tuned on a down-stream task.
3. It's true that the naive application of deep GNN architectures leads to problems like over-smoothing and the information bottleneck, but there are known solutions to each, and it's just rarely the case that you reasonably want/need information from far away in the graph except in special applications. In those cases, you likely want a different graph representation of the data rather than the perhaps obvious one.
4. It's true that GNNs improperly applied to problems, whether choosing the wrong graph representation, pathological architecture, or simply a problem that doesn't have dependence between the data points, will have poor performance. But I don't think that's surprising and I'm sure that there are many problems where simply throwing a CNN at the data doesn't help as well. Obviously, the modeling approach needs to fit the inductive priors of the problem.