AlexanderTheGr8

The github repo has > 3k stars so does the person still have an incentive to plug their solution everywhere?

I am calculating vector embeddings on 60 sec intervals with 30 sec overlap window. So overall, I only get 20-30 embeddings for a video. So a linear lookup works fine.

Considering that such a huge amount of information exists in videos compared to texts, I think that we are missing out on all that information from our current search engines. I hope that this can be used to bridge that gap.

shameless plug : http://semanticvideosearch.com. Thoughts?

I noticed you are using pbkdf2 to convert password to key. It's recommended to use ARGON2 nowadays.

From some googling, pbkdf2 increases CPU usage to make brute force harder, scrypt increases memory usage to make brute force harder. ARGON2 increases CPU and RAM usage to make brute force harder.

There are other advantages as well, and it is considered the best key derivation function right now.

any link on stanford's report? When will it be published? It sounds pretty interesting

I thought google's captcha (the tick captcha) used mouse and keyboard movements to detect humans. If a bot is using the web, their mouse movements must be either zero or very precise. So an erratic movement can be a signal of human user.

Is this actually how it's done? If it is, how can an AI beat that? Thanks!

I agree how today's parents slow innovation in their children or in some cases stop it directly \s

does that help you?

Yes, thanks!

Do you think I should add Youtube?

No, you are right, youtube is very popular.

As for the circles, honestly, I don't know. Since nobody else mentioned it, it could be that only I have that problem. And in that scenario, you can safely ignore it. If more people complain, maybe they will have the perfect solution.

Is there any way to get google cards? Google's cards are so useful because they answer 80% of queries without the user having to go into the website (which usually takes 2-3 seconds + ads)

But I understand if google doesn't want to allow devs to get access to their cards.

As a picky user, I would like to have the option to remove a box of any website. E.g. I don't like youtube thumbnails (and don't think of youtube as a good resource), so it would be nice to have the option to remove it.

As for implementation, it should be easy to allow users to remove any boxes, right?

PS : As for the clickable round circles, if there is a list with some black circles and others white, it just feels very intuitive to click on a black circle to make it white. It could be that I don't get this because I don't use iOS.

2 months before deciding to reneg due to too many fake people in Montreal. He accuses the City of Montreal of hiding the true number of thugs and criminals.

It's very good (will definitely consider moving). Some user feedback:

1) Dark Mode please

2) Toggleable websites; I saw that the websites I was seeing were black-circles in the panel below the search box. I instinctively tried to click on youtube to not see youtube box. It would also be nice to have a global panel where users can choose if they never want to see a certain website.

But these^ are just me being super picky. Overall, a great UX.

Fascinating. I wonder if this much funding will change the incentives to not release models in the future. (A lot of proprietary companies started with the open source promise)

Although this might incentivize more people to create open source models to capture capital. OpenAI went closed-source due to their 1B$ funding by microsoft. Stability showed that they could have made a lot more money by going open source.

Curious how AI models will progress. There are incentives for both open and closed source rn.

Can you link to the hundred-page document? I believe you are talking about prompt engineering, and I would love to get more information about it. I am struggling with figuring out good prompts.

Not trying to throw ML at the wall, but could it be used for this problem (considering all other options seem to be failing)?

As far as I understand, after MV3, ad blockers won't be able to use a long list of ads to remove them.

How about using a simple ML algorithm to detect whether the request is a genuine one or an advertisement? I am sure that getting training data wouldn't be too hard (all the ad lists that will get useless after MV3 are good data).

I don't make chrome extensions so I don't completely know how MV3 will cripple ad blockers.

Any feedback would be appreciated!

This seems to be a very big fine ($4.12B), much bigger than any "cost of business" I can think of. What are the chances that google will actually have to pay it?

They could re-appeal [0] or play some other judicial trick. Or maybe they say "enough is enough" and don't pay it.

They have been appealing the €2.42 billion fine from July 2017 even now[1]. So I don't think that they have any intention to pay this much larger fine any time soon.

Google's stock is already doing bad this year and this will likely make it worse (although the early trading today hasn't shown any indication of a change.)

[0] "The company can still appeal the ruling in the EU’s highest court." - https://www.cnbc.com/2022/09/14/eu-court-backs-antitrust-rul...

[1] https://curia.europa.eu/jcms/upload/docs/application/pdf/202...

As pointed out by user "space_fountain" below, this test was done by a competitor and may not be using autopilot or self-driving beta. My apologies. I should have pointed that out in the title.

Out of curiosity, how does someone possibly get 100k+ IP addresses? I had enough trouble getting 1 public IP address.

Why is talking to all authoritative servers so hard? As long as you cover all the cases such as CNAMES and have the IPs of the roots, you seem to be good to go.

To further support your point, AI is 90% efficient/good at driving. But the last 10% is so hard that some think it's impossible to have 100% self-driving cars.

Interesting point. I guess one way to avoid that could be making them believe that you are on their (govts) side? Intentionally click on pro-govt websites, etc.

If there is a conflict and you don't know who will come out on top, either support both of them (one actual support and one fake) or don't support either of them.

Either way, I believe that this topic of giving incorrect information is worth looking more into.

Very interesting. Thanks for sharing. There was another tool called something like "noisify". It intentionally visited random websites or clicked on random links to trip up any algorithm tracking you. That idea was fascinating.

I believe there's some misunderstanding. I meant letting them know who you are, but giving incorrect information about your beliefs and thoughts. I still believe that giving incorrect information can be very effective.

Even if Google/NSA knows your name, they can only see the online activity that you choose to do. You can intentionally choose to give them very incorrect online activity.

For example, intentionally visit websites (manually or automated) that you don't believe it. This will trip up their algorithm. How are they supposed to know what you actually think or believe in?

I have a different perspective and a question for HN. We always ask how to prevent corporations from surveilling us. How about letting them surveil BUT giving incorrect information? Intentionally do things (manually or automated) that gives them completely incorrect information about you.

Avoiding surveillance is becoming very hard. But giving incorrect information is very easy.

If he actually said this in 1971, hats off to him. That's an amazing prediction. Internet and the personal computer (information-rich) became popular after 1970s, so predicting attention scarcity before that is very prescient.

The cynic in me says that hundreds of people guessed about computer's future in 1970s that some of them are bound to be right. But Simon's prediction seems genuinely amazing.

Is C's std impl slow because of calling strlen() or due to converting into a dummy FILE* object or due to calling fscanf?

Does anyone know why the EU is doing this? Everyone knows that the children thing is a pathetic excuse, but why do they want surveillance? I highly doubt that messaging is a big risk to EU stability. So why make such an effort?

I feel the exact same way!! Throw RL at the wall for any problem, and it will magically solve it. In the back of my mind, I know that it's not true, but I always want to throw RL at any problem. Right now, I am considering throwing it to make TCP/other-networking-stuff more efficient...

If you run image recognition algorithms on this, will they recognize the outer image or the inner image or both or none?

Btw this is fascinating!