kalal

To some extend I agree, that if you are professional in some are, current LLMs may not solve the issues you are facing. But, while I consider myself as "expert" in one very specific area, and don't find huge help from ChatGPT in that domain, I don't live in a bubble. I have to sometime touch on completely orthogonal skills where I am not expert at all. And in these areas, I get huge benefit even though LLMs give me "average" responses. It is still better than my own, below average judgements. That is very strong point.

Serious question. What is the chance this video has been deepfaked? https://www.youtube.com/watch?v=ENlL-Uru-cM&ab_channel=CNBCT...

Just recently I skipped buying Samsung TV because I feared to get unwanted advertisement in home screen. Is that really the case?

Bluetooth is great up to the point when you want to have video in sync with audio. For video/audio editing or even gaming it is a problem.

My background is in image processing and I can see quite clearly that the high frequency noise happens when you subtract two images, one of which is shifted by half-pixel for instance. What you are left with is the edges (high frequency) of the image. The same applies to 1D in my opinion. For this reason I am not convinced (1) is actually true. The argument about lack of low frequency is interesting, who knows what is happening there indeed.

I am not expert, but if you don't use noise cancellation you would probably do better with standard headphones where you pay for the sound quality instead.

I purchased Bose QC-35 II as well. I don't have chemical burns, but after a while of wearing my ears become really sensitive. I feel strange pain while wearing them. This applies only when the noise cancelation is ON. When it is off the pain immediately disappears. My explanation is that these phones actually generate inverse signal to the noise, but due to various factors such as delay, the cancellation is not perfect and what I am actually getting into the ear is the residual high frequency noise, which may in fact be quite dangerous. Anybody else experienced this? There might be some study about it ...

I am honestly shocked by that! When you google authors name, you get bunch of anime faces + GAN research papers. So clearly, this is not a joke, but his personal interest/fetish developed after long hours behind his PC. I believe he needs medical treatment and not propose "startup ideas"!

If you find this interesting, you may also consider canny edge detector.

To a colorblind person (me included), this discussion is absolutely irrelevant. Why not to focus on edges instead? A black and white picture would solve all such nonsense.

Does the author refer to the whole AI community, when using the word "we"?

This is important point indeed. Here is what helps myself, when thinking about pixels.

- pixels are point samples in 2D space - their position is exact - position of of top-left point is (0,0) - position of bottom right is (cols-1,rows-1)

This way all math work (subsampling, affine or perspective warps, lens distortion, or even warping between image and any layer neural network). Failure to do that, will cause subtle issues that will degrade your performance. These will become quite important when working with pixel accurate methods (3d object tracking, object detection in tiny resolutions and 1-1 mapping between OpenGL and Neural Network). So I have to agree, pixels are not tiny squares, they are dots in 2d space.

I was wondering, how much different the simulation would be from stock models from Turbosquid: https://www.turbosquid.com/3d-models/formula-1-season-2020-3... The article mentions there were some problematic areas in the extracted model. So I am quite skeptical that the model is indeed more accurate than what you can easily get from legal sources.

Williams engine is supplied by Mercedes, tires are shared from Pirelli. But sure there are other parameters as well apart from aerodynamics.

> "I tried to match my lapsim program to telemetry from the 2020 Spanish GP, I arrived at CL = -5.421 and CD = 1.150"

This is very interesting. Does it mean that the author has access to 2020 Spanis GP telemetry from Williams?

Interesting point, which a little bit implies that Williams would have the best aerodynamics by now. But clearly that is still not the case. So I am quite skeptical, whether CFD time has indeed the effect it aims to have. We all know that knowledge which you build up over time as a developer pays in long term. So even thou teams like Mercedes may have shorter time for CFD, they have the knowledge base build up over time which they use heavily.

If you subscribe to F1TV you get timing data in form of additional screen which you can view. Apart of that, there were several attempts in the past to extract timing data from from live streams, which stopped around 2016 when the API was changes. So to my knowledge, there is no free API for that. If there was payed version, I would be also interested, but I don't know about it either. The thing is that F1 considers this traditionally very sensitive information and is unlikely to make it public any time soon.

Since Williams defines the very bottom of the race grid, they don't have much to hide in terms of technology. So releasing a 3D model of their car in this form is nothing that would harm the team at this time.

While I believe this is interesting project, I got really distracted by the PR around it. "scientist love flow chart" is like talking to babies and making fun of science at the same time. The fake scientist in all pictures ... probably I am not the target audience here.

What you can actually do with these extensions?

Yet another "which color is better" problem.

I personally use more or less 3 stages. 1. whiteboard and marker - explore ideas, big-picture thinking, fast iterations. 2. paper notebook - things that emerged from point 1 are more precisely recorded into a paper notebook. Things go much slower here, especially because the eraser or the paper burns out quite quickly and there is no point to keep pointless notes. The structure of the paper is quite important as well. I noticed blank paper does not work that well and I spent more space there with less structure. Linked paper is also not good. Square paper is better. But the best one for me is dotted paper. Ideal compromise between structure and freedom. 3. everything that survives stage 2 goes to real code and git logs. Here the structure is maximal. Keeping notes in the code is of course good way as well as extensive git comments.

Does the tone bothers anybody? After reading first couple of sentences I got the impression that the author knows it all and everybody else is just stupid... This may be cultural difference, or too much sensitivity on my side. I don't know.

What is DOP?

Well, that is possible. I have probably worked too much with trees and forests.

I am afraid that a decision tree HAS direct impact on the forest. Multiple trees averaged give the forest response. It's like saying, one vote in elections has nothing to do with the final decision.

Sure we have heard it many times that it IS a like a brain, and many times it is NOT like a brain. It really depends on your intended audience, but it is not worth picking on and it feels like the tabs/spaces problem. BTW: tabs of course!

MOTO AI is a project supported by EPIC MegaGrant. Our mission is to transform motorsport viewing experience. Our first use case is to augment the race with driver labels. This is achieved by 3D tracking of the cars in real time.

YouTube, it is your fault! It is too painful to watch such videos and knowing that the fix is possible. Efforts like YouTube-BB (https://research.google.com/youtube-bb/explore.html) clearly shows, that you have the capacity to process and annotate large quantities of videos manually. So hire annotators and manually check every video. Then stop the full channel.

Simple AI that detects absence of a face mask.

TLD Vision | Czech Republic | ONSITE | http://tldvision.com

TLD Vision an AI company focusing on tracking of objects in video. We are building a new tracking engine which combines latest advances from deep neural networks, real-time rendering and 3D computer vision. Our goal is to enable new kinds of applications in broadcasting and augmented reality which require extreme robustness and precision.

The following positions are currently available: C++ developer, 3D graphics developer, Visual tracking researcher, Deep learning researcher, Android developer

Apply here: info@tldvision.com