johnc1

Or, just incentivize or mandate stores to sell "child-certified" phones with parental controls pre-configured (along with a physical plug-and-play usb key for parents disable them when the child is old enough).

Exactly! We already have content tags on TV/Movies, just extend it to the web and make mandatory.

I imagine it could be not trivial to enforce (esp. for offshore web) - but definitely easier than enforcing the same sites to implement much more complicated identity verification (while preferably also not leaking this data).

But that might not even be necessary. A small on-device AI can probably do a decent job classifying pretty much everything we don't want children to see - with and option for parents to override it when needed.

I know! What puzzles me is responses every such article gets even on HN - let's build some cool tech that 95% of the general population and 100% of politicians won't even understand not to mention agree to.

Yes, government want to end anonymity and that's clear to some. But governments enjoy on a pretty broad support for this and many people supporting this believe it's a real problem. Suggesting to leave it unsolved or solve it in a way they can't trust or understand is only going to alienate them, making the government job easier.

I think suggesting a simple, cheap and effective solution to this problem that has no impact on privacy is a way better way to counter that. I think local parental controls fits the bill.

There is a much easier solution that already exists - parental controls on children's devices. I honestly don't understand why is it not solving the problem?

Yes, parents are responsible to set this up. But parents are also responsible to lock their alcohol, drugs or guns, condoms, etc., and many other things.

Perhaps parental controls are not good enough? That's where the regulation could genuinely help - require child-certified devices to implement minimum set of parental controls, and make them easy to use.

We're already been protecting children from printed porn, alcohol, drugs (legal or illegal), guns, etc. Children don't even have to buy those items, sometimes they can get them from friends or even find at home. It's always been parents responsibility to protect from that, and the Internet doesn't change that.

What the Internet does change is granularity of controls. With physical goods, parents can choose to give children travel magazines but not porn magazines. With Internet, parents don't always have a choice to give children a device to read travel websites but not porn websites.

Solving that granularity is all we need! And the good news - it's already largely solved: just give your kids one of the locked down phones/tables with parental controls (i.e., almost any phone/tablet today).

The law can and should make a few things better: - introduce minimal requirements for parental controls, certify devices as kids safe (i.e., implementing these requirements) - ideally, require websites to tag content - same way we tag it on TV for ages; this one is harder to enforce in practice (what about offshore websites?) but also not really needed - AI can tag content pretty well as part of parental controls

I know all of this was said before many times, but somehow it feels not enough people understand it.

> Can these models feasibly be run locally?

Actually you can, it even works without GPU, here's a guide on running BLOOM (the open-source GPT-3 competitor of similar size) locally: https://towardsdatascience.com/run-bloom-the-largest-open-ac...

The problem is performance: - if you have GPUs with > 330GB VRAM, it'll run fast - otherwise, you'll run from RAM or NVMe, but very slowly - generating one token every few minutes or so (depending on RAM size / NVMe speed)

The future might be brighter: fp8 already exists and halves the RAM requirements (although it's still very hard to get it running), and there is ongoing research on fp4. Even that would still require 84GB of VRAM to run...

> Politics and consumer capitalism are motivated to identify and target stupid people...

My bet is on a variation of this. To some extent, we all target people to advance our goals, be it to get them hooked on our product, to get people to rally behind an idea or a policy we want, or perhaps to get ourselves elected to a public office.

Entities with more resources naturally invest more in this and have more advanced tools to get people do what they want. Most likely emotions work much better when targeting large groups of people than smartness and objective truth, so that's what we get.

I think it's nothing new, but the recent research advances and the ease of reaching out to people personally these day made it so that using people to accomplish your goals become probably the most powerful tool on the planet. Why build weapons or wage wars if you could just make people do what you want on their own will and also sing you praises along the way?

Highly debated argument though; to me it's like saying you CPU doesn't understand HTML and your browser is running on a CPU, hence it can't understand HTML either. Scott Aaronson explained it nicely too: https://scottaaronson.com/democritus/lec4.html#:~:text=Searl... . Even the wikipedia page mentions many reasonable counter-arguments.

That's of course correct. Perhaps GPT-3 can do that too? I don't have access to it, but I wonder if it can be taught new words using few-shot learning.

In fact, even GPT-2 gets close to that. Here's what I just got on Huggingface's Write With Transformer: Prompt: "Word dfjgasdjf means happiness. What is dfjgasdjf?" GPT-2: "dfjgasdjf is a very special word that you can use to express happiness, love or joy."

What takes time is all the learning a child needs to go through before they can be taught new words on the spot.

One point missed by the article is visibility: even with SMS-2FA, I at least know when my password is being used by someone else (modulo sim-based attacks). For example, if my password manager gets hacked, and a password leaks. I think the overall conclusion is still right: it's a rather minor concern and let's come up with a proper solution and not waste the developers' good will on this one.

Another way to think about it is comparing to how children learn. First, children spend inordinate amount of time just trying to make sense of words they hear. Once they develop their language models, adults can explain new concepts to them using the language. What'd be really exciting is being able to explain a new concept to GPT-n in words, and have it draw conclusions from it. Few-shots learning is a tiny step in that direction.

Awesome work, thanks for sharing! For those trying to replicate it, could you please share some insights on which steps to train the model worked the best for you? I see 3 different train.py invocations in your colab - for how long did you end up running each of them?

Thanks a lot for advice! How does L1-As compare with L1-Bs or O-1s on terms of being tough? Those two are my other options I'm considering.

Hi Peter, thanks for doing this AMA!

Do you have any insights on whether getting an L1-A got more difficult with the current administration, and if/how the kind of documents/evidence needed for it changed recently?

I'm asking because I just got my L1-A renewal rejected (after having it successfully renewed 2 years ago), on the grounds of insufficient proof of my position being managerial. My I-94 expired and I don't have another visa, so I had to leave the US – and it's awful for my startup! Timing couldn't be worse.

I'm applying for a new L1-A, so any insights on my question above, or any advice on how to maximize my chance of getting it as quickly as possible, would be super welcome!