mgpc

I don't think there are actually major legal concerns. Copyright protects reproduction of a specific image. Looking at an image and producing something in a similar style is not copyright infringement, it's called being an artist. The law on this seems pretty clear.

The UK recently announced plans to make this completely explicit, to remove any remaining doubt: "For text and data mining, we plan to introduce a new copyright and database exception which allows TDM for any purpose. Rights holders will still have safeguards to protect their content, including a requirement for lawful access."

https://www.gov.uk/government/consultations/artificial-intel...

Google's arrogance here is astonishing. The police say no fault, and it's the subject of an NYT investigation, and they still won't restore the account. What hope do the rest of us have?

I've been an Android user for a long time, but I think this might finally push me to switch to Apple. I'm just disgusted by this.

Completely agree.

Some of the tools for interpretability can be useful (particularly for debugging), but I think the broader idea that we always need to be able to understand our own models is basically wrong.

For example, if you want AlphaGo to explain why it made a particular Go move, what kind of an explanation is possible? In many cases the only explanation may be that the move leads to a higher probability of a win. There simply may not be a more “compressed” or “high level” answer. Even human Go players often cannot explain why they choose particular moves, other than references to shape and feel, which is basically another way of saying their evaluation of the move leads to a higher win probability. There are a lot of domains where we may just have to accept that that _is_ the explaination.

To zoom out a bit, our greatest discoveries have historically been about finding the rare places where the universe is computationally compressible. Boiling a kettle is almost unimaginably complex to describe in terms of the interaction of elementary particles. But you can make very good predictions about that process using an equation that fits on a cocktail napkin. There may be other areas in which the universe is compressible only to a lesser extent. The parameters of AlphaGo are vanishingly small compared to the size of the Go game tree, but are very large compared to the equation we can use to predict the kettle. There may be many problems where the best descriptions lie in this intermediate domain, a domain which we have never really had access to before (except via biological brains).

So if learned models give us access to some truths without access to their (human intelligible) explanations, I think we need to just embrace that. If you allow yourself a new way of seeing, you can see new things.

I recognize what you're saying is partly true, but I still have an almost irrational burning hatred for AMP.

Is the email address something someone else might enter by accident? I constantly get emails intended for someone who has the same surname and first initial as me (and presumably a very similar email). She signs up for all sorts of things and accidentally enters my email (it's quite annoying). I even got her plane tickets once. And I can see loads of FB advertiser matches that are clearly her rather than me.

Why would this help them? Facebook already lets you target by geography, they could just target all US residents much more easily.

As a European start-up founder, I agree with this. The unintended consequences of the GDPR will be to (1) cement the power of Google and Facebook, since they have logged in consent (2) kill third party retargeting companies, ironically French company Criteo being one of the promient examples (3) generally make life harder for Euppean companies, and sharpen the incentive to ignore the EU market where possible.

As a user I do feel some privacy regulation is a good thing. There are certainly good things about GDPR (e.g. a lot of third party ad tracking deserves to die). However, I am far from convinced that the GDPR is the saviour some people think it is.

The DGraph team have done some very interesting work. The recent addition of transactions is an extremely impressive technical accomplishment, and a very good long term move. Excellent to see a production-ready 1.0 release.

Alternative perspective:

Like many universities, the value of YC to founders is mostly signalling and network. What you actually learn at YC is a minor component of the value (since you could learn most of it just by reading or watching videos)

As YC batches get larger, the signalling value of being a YC company goes down. If everyone is special then nobody is.

As YC batches get larger, I suspect the value of the YC network also goes down. This is maybe a little bit counter-intuitive, but I think a large influx of less highly-selected startups will dilute the network and prevent it from functioning effectively.

Unfortunately YC is not completely aligned with founders here. There is a strong temptation for YC to increase batch size. YC gets 7% of a company for far below market price. So increasing batch sizes makes short-term economic sense. However, if YC increases batch sizes too much and harms its brand, it will eventually break its own model. So increasing batch sizes could be seen as form of spending down the brand equity.

This all hangs on a few premises that may or may not be true: 1) You can't have larger batches while keeping quality constant. 2) YC doesn't directly influence the success of participants that much. 3) If the value of YC drops, good founders will eventually stop applying

I suspect an argument like this has been going on inside YC for quite a while. And the blog post was written by the original founders to try to settle the dispute in favour of the pro-scale faction.

What you're describing is called email. Seriously - it is exactly this, right?

Paper from a few years ago comparing Google's OCR system to commercially available benchmarks:

http://www.educatingsilicon.com/wp-content/uploads/2013/10/p...

A lot better for text in photographs. Comparison might be different on dense document text though.

The same thing happened to my grandmother. She went in for a minor hernia operation, and came out a different person. She never recovered. It was very dramatic and very sad.

Nice.

If we a seriously believe AI is an extinction level risk, and getting there more slowly might improve our chances, then it seems like we should try anything with a good chance of working. I think this has a better chance of working than regulating AI research itself, since chip manufacturing is centralized and highly visible, whereas AI research could be advanced in secret by a few people in a room. Much easier to successfully regulate the former than the latter.

I think @sama's plan isn't very likely to work. Once strong AI is possible, there's no regulatory structure that will keep the genie in the bottle for a meaningful amount of time.

The only path I see with some prospect of success is to limit the total amount of computation available on Earth. (Basically like Vinge's slow zone). If we could engineer a pause in Moore's Law at just the right point, it would buy us time. Maybe we should put a drastic tax on computation at some stage.