@@skolarii If I'm not mistaken this was because the image(s) in question were "manipulated", that is they were quite substantially enlarged in order to be able to "discern" the actions/behaviour of the defendant. And pictures were still not at all "clear", and since they were ""Zoomed in and enlarged"" naturally a certain ratio of the pixels were NOT captured by the camera (drone video), but rather were "created by the interpreting software". And "in a nutshell" it was this that made the judge "throw out" SOME of the "photographic evidence", though some was allowed. At least that is how I understood it, but if You know differently I'd be interested to take part of Your information and knowledge about it. Best regards.
@@acberkowitzIt was really nothing. The judge was just dumb & the lawyers involved were just dumb. The prosecutors presented the "zoomed in & enlarged" version of the photograph, the defense objected & the judge accepted the objection because apparently, "zooming in" on an image makes it "manipulated". They basically said if you can't see something without zooming in, you can't use it. Like, seriously!
Cameras are not allowed because Samsung's facility is full of dismantled products from competitors. Every product category bearing an uncanny resemblance to the competition that preceded it doesn't come for free.
I love it when you guys just go off the rails. David and Ellis were so stressed about what is a photo and then there's Andrew cracking jokes and teasing David hahaha never change guys
27:05 That Do it! part was the funniest thing I have heard in a long time.. The wait, the tone, the slight hesitation, the delivery were incredibly good I laughed outloud multiple times. Thank you Ellis..
@@josh44026 Yes. He said that he was exhausted that week because of the Vision Pro videos. And loosing some trivia questions to David and Andrew didn’t help either 😊.
I just keep seeing these colour wheels on my phone feed and scroll past, wondering WTF is all this about??!! Didn't know it was some new RU-vid masterplan to f*ck with our feeds even more!! 😜
The photo not photo debate reminded me of the "ceci n'est pas une pipe" artwork form Magritte 🇧🇪 (1929) where the painting is just a drawing of a pipe with a writing that says "this is not a pipe". Because it's actually a 2D drawing representation of a pipe....
I watch a lot of comedy videos, but haven't laughed as much as I did during your conversation about 'what a photo is' in ages. Thanks for the great laugh!😂
Hey. Philosopher here. I've never researched the topic. But if you're interested. You can watch me postulate on photos... A photo is real insomuch as it is an accurate representation of a moment in time. Everyone has a different perspective of the moment and each lense, sensor, person, or film has a different reproduction of that moment in time. But no matter what. That moment was captured. And the quality of the photo can then be assessed. More post processing brings that photo further away from the original. And/or sometimes depending on the photographer they can edit the photo so that it looks even closer to the way the photographer saw the moment from their perspective. But no matter which way you process the photo. There once was a moment in time that was captured and that is the essence of a photo.
Why is the RU-vid thumbnail color sorting a thing before being able to sort your RU-vid Music playlists by artist, title, etc. Very backwards priorities on RU-vid/Google's part
Andrew, Ellis, and David going on a pedantic rant with the “what even is a picture?” tickles my brain so much! “Color isn’t real.” I see you, David! Bluejays aren’t blue 😉
It would be cool if Gemini becomes capable enough to just let you decide what you want the trigger name to be. So you could say Hey "Google","Bard", "Gemini", etc. or whatever you'd like and it can still trigger the action.
I gotta give it to you Marques, you’d make a really good scientist and/or epidemiologist. Your analysis of the RU-vid’s new multi thumbnail metric and the its pitfalls is really critical and well thought through 👍
I think the new metric for VR will be “ppd” or pixels per degree. So a combination of field of view and resolution. By that metric an iPhone’s Retina display is much higher (as it, in fact, appears to be).
Light hitting the sensor when it's off still CREATES an image, but that image is not picked up/recorded by anything. Just like a pinhole camera, where the "sensor" is just the back wall of the box, there is an image there, but it's not captured permanently. The key here is the lens (optical element) in the process, that creates the picture (eyes, camera lenses, or the hole in the pinhole camera), that is what renders the image, after that is just a capturing process, be it film or a photo-diode, and that's where different methods come into play, so even a film stock is not the original image, because each type uses a slightly different chemical composition, altering the colours and exposure already.
I totally agree with that Samsung guy. The moment a modern digital camera captures a photo it's heavily processed: - Picture profiles - Auto white balance - Active D-Lighting (Nikon) - Portrait impression balance (Nikon) - The way skin colors are handled in green colored environments (leaves, grass, trees) (was he running or was he sick?) - Surely Canon, Fuji, Sony have similar image improving steps that alter reality. More so concerning smartphone cameras: - Each night shot is a composite of multiple shots - Heavy noise reduction maybe deletes small objects like letters or hole writings on a sign or a wall. - Nobody knows how a smartphone or AI handles different parts of the photo individually after hitting the shutter button. With 10 different cameras you will end up with 10 different pictures and you won't be able to say how the scenery's lighting really was if you weren't there yourself.
For VR headsets, you want the Pixels Per Degree metric _(PPD)._ Quest 2 had 20, Quest 3 has 25 and iFixit estimates the Apple PPD at 34, so a pretty huge boost of 36% improvement, but it's not orders of magnitude different. The OLED displays + PPD improvement really separates it from a lot of other headsets from a visual quality perspective.
I have Gemini advance and you can set the new app to replace your Google Assistant but it doesn't necessarily replace your Google Assistant. It just when I trigger my assistant between many different ways that you can accomplish that. It goes through Gemini first and then It determines does it need to handle what I'm asking or does it need to delegate that to the assistant. It's awesome!
And it's horrible at delegating. I tried it and it was super glitchy. Uninstalled the app and went back to Assistant. I'm on the Pixel 8 Pro too. Kinda disappointed because I was wanting to have a conversation with a LLM with the convenient integration of Assistant.
@@CapCannoli what do you mean It's horrible at delegating? If I ask it something like set a timer for 25 minutes It sets a timer for 25 minutes, but it shows a little flag icon or toast notification if you will showing that the assistant did that. If I ask it something like a complex math problem or pretty much anything else that just requires an answer rather than a task it needs to complete on my phone the answer is given to me from Gemini and if it's a task that needs to be completed on my phone like set a timer or a calendar thing or other things like that the Google Assistant handles that. It does it beautifully.
@@CapCannoli also, I believe it's the pixel 8 Pro that is glitchy or buggy not necessarily integrating Gemini with the assistant. I'm using the Google pixel 7 with AP21.240119.009.
@@bl5843 I asked it to play a song and to launch an app. Neither command worked. Both times it told me that it couldn't control my devices. It also wouldn't listen after I said "Hey Google." It would launch at the bottom like Google Assistant but stop listening immediately. It would only work if I hit the microphone button to talk into it, then launch into the Gemini app. Also, never had any glitches with the phone before Gemini, and now that it's gone assistant is working fine.
I'm using the Galaxy s20 ultra and I have Gemini and it has replaced the assistant app. I got to say it's delegating well on my phone and has worked just as well if not better than assistant. I am really loving it so far and so excited to see how far it goes. I live my life with Google Assistant helping out and it's so far NOT a downgrade
Purist agree with David on photos are film old school. I think photos also include digital as long as it only processed to give you the best version of what you are seeing. Any altering or post process photoshopping, AI manipulation or filtering that fundamentally changes the image from reality are a different category of photo. Filters like black and white and sepia are okay but manipulations, art work, and other types of photos are not real because they are not what was observed. We still call them photos because they started as this but they should be called altered images. It is all just down to technicalities, definitions, and philosophy.
49:48 Shared experiences already exist in Apple Vision Pro. It's left with app developers to integrate those experiences. Anyone who have watched some of the developer videos on Apple Vision Pro released in June 2023 knows what they are.
Not only can you not control home devices with Gemini, you also can't use Gemini if you use Google Assistant instead. I just want to be able to ask assistant to turn on my lights AND ask Gemini questions when I want.
Clearly I am the only person in the world who does not care what a thumbnail shows, the video description is what makes 99% of my decision the thumbnail is literally just a reference point in my brain as have a I seen this video before
Not sure if maybe mine is different, but I installed the Gemini app, and saying 'Hey Google' activates Gemini and also allows for all my google assistant tasks (smart devices, routines, etc.) to work. It seems to me Gemini can be a full replacement of Google assistant
I feel like veritasium tends to undersell his thumbnail. For exactly his last video was just "blue LEDs are hard to make" but ended up being a very interesting video (as always)
Regarding that “new” RU-vid thumbnail colour sorting feature, I had access to it for probably the second half of last year but it recently disappeared. I’m in Canada so I wonder if we were among the first to test the feature.
The Samsung exec maybe trys to say that all the photos you take with electronic sensors never show reality how it actually was. Our eyes can perceive more brightness levels than any display can show (thats why tonemapping is a thing). Since no photo you see on a display is even close to reality and only an approximation of reality I can understand where the thought "there are no *real* pictures" might come from.
I think the "real photo" discussion confounded two concepts: 1. "Real" photos as photos do "exist" 2. "Real" photos as "the contents of the photos match the reality"
29:14 I like Wikipedia’s definition: “A photograph (also known as a photo, image, or picture) is an image created by light falling on a photosensitive surface, usually photographic film or an electronic image sensor”. To Marques’s “if the sensor is off, nothing is created” argument I would say that if the sensor is off, then it is no longer a photosensitive surface. Hence, the definition stands.
A photograph is what we consider a captured moment in time, whether the image is processed, altered or left untouched. It is a graphic, created by the absorption of photons and then made into something we can practically use.
Explainer for 21:29 - DeepMind's Chinchilla was a "compute-optimal" language model. Before Chinchilla, LLM Scaling was heavily inclined towards increasing the 'model size' (i.e. no. of parameters) and relatively less inclined towards increasing the dataset size used for Training the model. So, basically if you have 4 times the amount of 'compute' resources, it was like multiply the model size by ~3.5 and dataset size by ~1.2. With Chinchilla, they corrected this by balancing the model-size v dataset size choice. So, post Chinchilla it was proven that if you have 4 times the amount of compute, you could 2×modelsize and 2×dataset size and achieve a more optimal result. Conclusion: LLM performance improves as we increase model size, dataset size and amount of compute used for training. For optimal performance, all 3 factors must be scaled up in tandem.
Perhaps since this is a death of Bard episode, you should have asked Gemini "what is a photograph". "A photograph is a type of picture created by capturing light with a camera. Light reflects off objects and is focused by the camera's lens onto a light-sensitive surface, either film or a digital sensor. This process creates a chemical or electronic record of the scene, which can then be reproduced as a visible image. Photographs can be realistic or abstract, depending on the photographer's intent and technique."
Summary of the podcast: RU-vid is messing around, Google messes up their naming again, Top samsung ex and biggest tech youtube talk about what photos are, more apple vision pro stuff and car stuff
“Push to Pass” is something they use in Indycar. Every driver gets about 150-200 seconds per race. F1 uses DRS, or also an overtake button that allows the driver to use battery energy to overtake or defend a position. Cool to see Porsche using something similar!
I feel like the categorization of videos by RU-vid is a psychology thing. Personally I'm more aligned to clicking a video with a red /orange thumbnail. So, I think they were actually uo to something. Not that they were bored as you guys say Viewer's choice to click on a video is heavily influenced by the thumbnail Color is definitely major influence.
49:31 pinning windows is great but also, imagine making pinned windows public for other vision pro users. Like, at the studio, shared pinned windows floating with chats, announcements for others in specific spots. Can be a great way to reward perks.
Honestly, the right question isn’t “what’s a photo” but “what is an edited photo / an acceptable level of editing in photos”. I think we know what photos are. What we can’t agree, is what is an acceptable level of manipulation in photos, and at what threshold we should be informed that the photo is edited.
Why doesn’t Apple allow users when prompted to say “hey Siri” to use any phrase that they would like to use.. That would also function a bit like a personal password.
@32:00 @Marques The point being made for a photo definition was before digital photos. The light landing on paper or film brings it back to the 35mm days. When the light hits the film it has the original photo, and from there you can edit it how you wish.
The RU-vid color feature is really helpful when you are bored af and not sure what to watch. I have been using it for a couple of days, and it looks fun
"Bard" felt really good as a brand name, as it evoked something you could have a conversation with, that told you new things "Gemini" felt good for their knowledge model, as it evoked the idea that it was different versions of the same model. And it could change and adapt to Gemini 2, or another new model name, without muddying up the branding for their conversational endpoint
The thing that's incredible about the Samsung exec's comments is how casually Samsung will embrace nihilism to keep selling phones as easily as possible.
