To be blunt, the current economic value of the humanities is pretty low, just like STEM as developed by humans will be in the future. I think universities may play a role in the future, but more as something people do for fun as a hobby, and not mainly for its superior teaching capability, but for the social experience.
The whole education system and its purpose is crumbling. Not just the universities but all the way down to preschool. It started with the internet, gained steam with top notch content being available on YouTube, and it's properly dying with the advent of AI.
If the purpose is to learn, you can do it better with YouTube and AI. If the purpose is to have fun, socialize and network, that is better done elsewhere, doing sports or other hobbies with other people. If it's babysitting you need, there are cheaper, better and more fun ways for the kids to spend their days. If the purpose is to learn a job, again, education is a terribly wasteful way of achieving that.
Then there's the fact that we'll all soon have to come to terms with, which is that most people are already barely able to contribute value to a white collar job, and in 20 years I'm pretty sure that number will be down 99%.
"If you do something you love, you will become good at it much faster than something you don’t love, and when you are good at something you will have some opportunity to commercialize it."
This sums up my life philosophy very well at this point. It also applies to kids and education, as I encourage my boy to do the things he loves (happens to be Minecraft and its thousands of mods, commands and other things to learn), and gently guide him over time to a place where he might be able to commercialize his skills in the future. It's amazing to see what a child is able to learn when they are obsessed.
Exactly. I think the people who oppose this do it for sentimental reasons. Either because they want the prestige of a place like Stanford on their resume, or they enjoyed the social milieu. A social network of similarly ambitious people is also arguably more important than the actual knowledge. The knowledge is already available for free online, and like Bryan Caplan points out: at many universities you can just go and sit in class, but nobody does it because what you actually want is the credentials, not the knowledge.
How would it make people dumber? Students provably forgets 90%+ of what they learn, and education is massively inefficient today, which can also provably be improved by virtual tutors (see Bloom's two sigma problem). Your reaction seems knee jerk instead of thinking through this from first principles. Something like this is likely to happen whether you like it or not.
Last summer I put together a flag quizzing game with my niece, who's a geography nerd: https://flagsaroundthe.world/
I think minimally contrastive examples work well for learning, i.e. you want to first learn to distinguish between broadly different categories, and focus on more and more similar examples over time. This little app doesn't have scheduling at all, it just samples examples based on previous responses, so sampling examples you got right exponentially less often as you keep answering correctly. I think sampling has some interesting advantages to SRS scheduling actually. Add new "items" becomes a straight sampling problem, where you can set a constant probability for each item that hasn't been seen yet.
Xi has been moving in that direction yes, but generally that has not been true. Also, it sounds like you greatly sympathize with Chinese authoritarianism, maybe you need to do a reality check. You can always go live in China if prefer that. I've been many times, for months at a time, since I have in-laws there. Lately I would describe it as a high-tech North Korea.
I was thinking in the context where you'd like to potentially monetize the side project as some point. But definitely, I spend a lot of time supporting my son's Minecraft obsession, and since he's already expressed great interest in building mods I think that's something we'll do together that could be considered a side project.
In fact... I see a possible future where I have a collaborator worth his salt :)
I could keep up side projects before I had kids, but now... I've learned you have to focus on your health with whatever time you have left in the day/weekend, no matter how excited you are about some idea.
No, we can hate Musk all we want, but we shouldn't rewrite history because we despise him. He is a complex individual, and he still deserves the credit for founding SpaceX and being chief engineer.
I think what we actually want is to be the equivalent of a gentlemen scientist supported by family wealth or a patron, sitting in our study doing what interests us, corresponding with other smart people interested in the same topic, going for long walks thinking about hard problems. Even though we shouldn't complain, we can still strive for this.
Even if it's not hell, it could be so much better. It could be a place that kids actually look forward to going every day. Instead, we put them through 12 years of mandatory low grade torture where nothing they do is connected to the real world, their interest or curiosity, and when they're done they're launched into a world of AGI and ASI where none of what they learned is remotely enough for them to contribute to society in any way.
It's the same kind of people who claimed human flight would not be possible for 10,000 years in 1902. I just can't understand how narrow your mind has to be in order to be this skeptic.
I think it's really underestimated how much "frivolous" obsessions lead to rapid learning. I think you can take any obsession and guide it down a productive path. Played a lot of Counter Strike back in the day, and went to school with two brothers who built a betting site for CS esports games (not real money) and it blew up. I think they're rich now. Final Fantasy may seem like it doesn't teach you much, but it taught you reading, probably inspired many people to learn about Japan and the Japanese language, build websites about it etc.
Agreed on Roblox. BUT it has Roblox Studio, which seems like a Unity-light. I taught my son to add different shapes, move them around, scale and rotate etc, so he could build his own obstacle courses. I think if you go down the Roblox route, make sure to focus on creation.
Yes, first of all just try the game and see if it sticks. Once there is obsessive interest, you can try to guide it towards mentally stimulating ends. I don't know if I'd recommend YouTube for english native speakers, but for kids like my son, learning English from YouTube is a plus. There are lots of shitty YouTubers just goofing around, but there are some good ones like Pixlriffs (Minecraft Survival Guide), Shalz and Mumbo Jumbo that focus on building things and exploring all aspects of the game and different mods. As an example, my son learned about crop rotation from Pixlriffs, because apparently its a mechanic in the game. Can't be mad at that!
There are two versions of Minecraft: Java Edition and Bedrock Edition. Java Edition is only on PCs/Mac, while Bedrock is on phones, tablets and consoles. Java Edition is the original and I think more customizable in general, a bit more hard core. We just have both.
We use a special launcher for Java Edition called "Prism Launcher" which lets you handle different versions of the game, installing mods etc. It's really taught him important concepts like "if you install 134 mods, incompatibilities will break the game", and general debugging, like, install one mod at a time and revert it something doesn't work. I prefer PCs in general because it's more likely to lead to an understanding for computers.
I can attest to letting my kid play Minecraft, install whatever mods he wants (at 6-7) and watch related tutorials has catapulted his English (not his native language) to insane levels. Playing with the Create Mod is basically just engineering. I do feel conflicted about how much screen time he gets but compared to how I spent my time as a kid, watching shitty soap operas on TV in the afternoons (because that's what was on), this is infinitely better. How can I be mad when he's throwing around words like "non-euclidean"? He's learning WAY more from Minecraft than school at this point. As a bonus, we're practically neighbors with Mojang, and visited them on a sick day once.
You have to remember what came before 2012: SVMs, Random Forests etc, absolutely nothing like the brain (yes, NNs are old, but 2012 was the start of the deep learning revolution). With this frame of reference, the brain and neural networks are both a kind of Connectionism with similar properties, and I think it makes perfect sense to liken them with each other, draw inspiration from one and apply it to the other.
V13.2 is probably close to 1000 miles per critical intervention now. Geofence it, avoid stupidly dangerous UPLs and fix some map issues and they're already on par with Waymo. Next version has 3x the number of parameters and 3x context length. If you know anything about scaling laws you wouldn't be betting on Waymo right now.
Corollary: do a lot of different things, activities, jobs, hobbies. I need to get better at this.
It's amazing how many different kinds of businesses there are. I once talked to a Dutch guy who designed and manufactured machines that clean astroturf in stadiums. He was rolling in it, small company. So many industries like this you have no idea existed.
That's just cope. With that savings rate he can save for 10 years and spend the next 40 years doing his own thing. As long as you keep your standard of living the same.
Would love to know the details of their situation. I'm guessing this is a case of hedonic treadmill and keeping up with the Joneses, or just generally extremely high expectations. And even if you have a high income that doesn't mean you can have everything you want right away.
They train an end-to-end model to drive based on 8 camera streams and recorded input from human drivers, training on tens, (if not hundreds now) of millions of 30 second clips from their consumer fleet. That's why they're bought one of the largest GPU clusters and making their own chips and transport protocols.
It's not widely known, but Tesla probably has one of the largest training cluster, because practically all the GPUs they buy go towards training, while most of GPUs for e.g. OpenAI go towards inference. Tesla does inference in the car.