You're a slow learner. But I'm glad you figured it out.

That's funny. I know other compiler writers with a decade of experience. We all agree, the first 2 times you really don't know shit. People who think writing a compiler once or even 3 times means you understand the subtleties of compiler theory are delusional. I've read up on compiler theory and one thing is clear. Text book examples only scratch the surface. Only real way to learn this stuff is to try writing a compiler for poorly designed language or DSL. I've had to do that in the past and tried to explain to a dev with masters in CS why his language needed to change. After I left that job, the CTO asked him to implement the compiler and his response was "I can't, I don't know how." They asked me to come back, I politely declined and summarized the syntax changes needed to fix the DSL. Even with the AST I provided, the CS person couldn't understand.

That's too bad, you should have. That's a deficiency of your university. Usually they teach it in the operating systems class. Either way, you need to learn it somewhere, and it sounds like you took care of it.

Just because OS class teaches multi-threading, that doesn't mean a person understands it. The concept and practice are two very different things. I would say most people who took OS class leave with a false sense they understand multi-threading. It's only when they work on complex threading apps that you really understand it. The same is true of writing compilers. a decent percent take compiler class to get their bachelors degree, but they don't really understand it. From my own experience, after you've written 3 compilers you start to really understand what "right expansion" and "left expansion" means.

I wish japan good luck. With their declining birth rate, Japan has too many problems to solve. Most of the people that lead the semiconductor research are probably close to retirement or retired.

originally httpd was call "a patchy webserver" because it was constantly being patched. The name stuck and was changed to apache. I've been contributing to apache for over a decade and I have no problems with changing the name. The down side though is it's a lot of work. All of the software will need to be updated to remove apache. That's not a trivial task. Keep in mind open source developers don't like to do work they feel is pointless. There have been thousands of flame wars on apache mailing lists over the years about pointless work.

Also keep in mind there's lots of bloated egos in apache community. I stopped counting the flame wars after a few years and just ignore them. The bigger issue is source control and regression management. Once you change all of the package names, you've added extra overhead with back porting fixes. Look at Hadoop. If you count all the components, there's over 10 million lines of code.

It will take a long time for every project to transition and the amount of work needed is easily tens of thousands of hours.

If there's enough utility people will pay for it. If they pay actual money you can afford to scale.

that's $$ question. How much would it have to cost to break even? How much does 4 A100 cost today? A quick google search says around $60,000.00 for 1 2U rack mount.That's just the server, not including network gear, rack, electricity and other overhead. How many data centers can handle the power demands of 10K 2U servers? Sure some people will pay for it, but the question is how much. $20 a year is about as much as I'd want to pay. Say Microsoft charges $20/year for the service and 1 2U with 4 A100 can handle the load. That would mean 3K people per 2U rack mount. Microsoft would need over 3K 2U rackmounts to handle 10 million users.

Google handles 8.5 billion searches a day. Do the math, is it even physically possible to scale it that far? Can Nvidia produce enough A100 to satisfy the demand? Once you look at the facts, the details matter.

Bullcrap. Your ignorance is showing. Backprop was a big breakthrough in 1986. DL was an even bigger breakthrough in 2006. GANs were a game-changer in 2014.

That's not the whole picture. Even though backprop was discovered in 1986, the hardware wasn't enough and the datasets were too small. The combination of GPU acceleration with CUDA and gigantic datasets over 100 million was needed to make a break through. Remember there were several AI winters.

The turning point for artificial neural networks wasn't the concept itself, since we know that is old. If the internet didn't exist and wikipedia didn't exist, do you think researchers could have trained large language models? If common crawl didn't exist, would GPT exist today?

GAN is cool, but it is also over-hyped. Some types of use cases generative approach works to fill in the dataset gaps. If you only have 100,000 records in your dataset, using GAN can help you make progress. The bigger breakthrough is transformer, which is an improvement of LSTM. Another big breakthrough is residual layers. If you look at Bengio's system 2.0, you can see other approaches are needed.

yes, GPT-chat does some cool things, but it also eats much more electricity than simpler methods. How many concurrent searches does Google handle today? Can you scale gptchat to handle the same amount of traffic and not fall over? OpenAI is already reporting performance issues with the current load.

the beauty of Google's approach is it's easy to parallelize, easy to scale and easy to refine. The same isn't true of Large Langauge models. You need a lot of A100 or custom ASIC racks to do the same thing. There's still a lot of work ahead to make GPT based solutions scale without needing crazy amounts of GPU+CPU power.

after college, a lot of friends moved from southern CA to silicon valley in the mid 90's. No, 150K a year gets you a shed today in silicon valley. A few friends were paying over 1K /month for a room in a 5 bedroom house around 1996. By late 90's, a few friends were paying over 2K/month in SF for 1 room in a shared apartment. The avg rent in SF today is around 3,300/month. That's actually down a bit from pre covid prices

that's not a complete picture of Taiwan.

For most of china's history, Taiwan was the island where people fled to find safety. Any time a king lost his kingdom, they would run to Taiwan. During the cultural revolution there was two groups fighting for control: Mao and Cheng Kia-shek. When Mao won, cheng kai-shek escaped to Taiwan and took over. Both Mao and Cheng claimed to be the "True democracy of China." The culture of Taiwan has always be scrappy underdog. During the cultural revolution, a lot of mainlanders escaped to Taiwan. The first few decades after Mao came into power, China was too busy trying to survive and didn't bother with Taiwan. You can read about the famine that happened, there's plenty of reports and movies about that period.

I have family in Taiwan and mainland. Most people don't want a war with Taiwan. Taiwan uses 3 different spoken languages: taiwanese, hakka and mandarin. From a culture perspective Taiwan is distinctively different than mainland. In reality China has over a hundred dialects that sound different mandarin. From a politics perspective, Taiwanese people consider themselves self governing and have never considered themselves ruled by mainland.

I wouldn't say Elon drove it. China is far ahead of the US on BEV vehicles. Even if Elon didn't champion BEV, china would be even farther ahead of the US. Elon was the most vocal person in the US pushing BEV, but to claim he drove it undermines the hard work by thousands of engineers that did the work. I have to ask "why do people credit 1 person with the work done by thousands?" By himself Elon wouldn't have changed BEV world. Elon gets credit for being the most vocal wealthy person to put his money into BEV. He gets credit for hiring good managers and building the foundation of the company, but ultimately the entire team is needed to make "it" happen. The US has this disease of worshiping people blindly.

Now it's official. Google true motto now is "do evil and F the world."

thanks for caring about my tampon pad. Looks like your pad is causing an infection, you should get a doctor to remove it.

apple is a walled garden. I get your angry and pissed, but I have to ask. By shouting into the wind, do you think Tim Cook will change his tune? No, I don't remote into iPad for my relatives, because it's not possible and they don't need it. When they do, I tell them "go to the apple store." I stopped providing tech support for family members a decade back when they switched to macbook and ipad. When they used windows, it was a headache. Getting relatives to turn on remote help just wasn't worth it and honestly it sucks. I have zero desires to go back to being family tech support.

I don't like walled gardens either. It's not worth the energy to get pissed at Apple or Google. No amount of me bitching will change either company. When I was younger and more stupid, I got pissed. Now that I'm older and slightly less stupid, I've realized that is just wasting energy

Like you, I've been using computers since my parents got me atari.

The few times I've used a real Android device, I find it annoying and silly. I have zero desire to customize my phone or tablet. In fact, I never think about it. I get lots of people want to customize it. Go ahead and pick the device you like. When I use android emulator during development, I find it finicky. I have a beefy workstation with 64G of memory, so android emulator shouldn't run like crap on windows 10. The point is about complete newbies who hate tech and don't want to learn. From my own experience with people that hate tech, iPhone and iPad are easier to learn than windows or macos. Clearly some people find iPadOS and iOS confusing, but that's really an individual thing.

From a developer perspective, android API and version crap is a complete pain in the @r$$. Thanks to google and their BS software. Both apple and google are anti-competition, so people shouldn't trust either corporation. I own apple products, but I'm not a fan. Plenty of annoying BS about iOS. When apple updated iOS, I was super annoyed for months because some UX asshole decided to move shit around for no good reason. Same thing happens with Android. Some new UX person decides they have to put their stinking stamp for the sake of being different.

my grandmother is over 90 and uses iPad just fine. Not all grandmothers are the same. Even my dad, who is the opposite of tech savvy finds iPad easy to use compared a windows laptop. Being a developer, I have galaxy tab and ipad. As much as I want to like android tablet, it blows. Maybe galaxy tab is screwed up and samsung can't make a decent tablet. When I bought mine, the galaxy tab was rated higher than other android tablets

by "them" what are you referring to?

If you look at the autonomous taxi that went live in ShenZhen china, you'll see China is already ahead of Tesla. The videos show the autonomous taxi handling crazy traffic. Based on videos of the latest Autopilot betas on youtube, 60-70% situations in ShenZhen would classified as "edge cases" by autopilot. Anyone paying attention would realize China is already ahead in some areas and unlike the US Federal government doesn't suffer from decision paralysis. Look at how much solar China has deployed in the last 5 years, it makes the US look pathetic. We keep voting for corrupt kleptocracts and let the media divide us. The reason China is ahead is mostly because we're too busy fighting between ourselves.