Most OLEDs today ship with logo detection and will dampen the brightness on static elements automatically.
While it isn’t a silver bullet, it does help reduce burn in since it is strongly linked to heat, and therefore to the pixel brightness. New blue PHOLEDs are expected to also cut burn in risk. Remember that LCDs also used to have burn in issues, as did CRTs.
I’ve been using Nvidia under Linux for the last 3 years and it has been massive pita.
Getting CUDA to work consistently is a feat, and one that must be repeated for most driver updates.
Wayland support is still shoddy.
Hardware acceleration on the web (at least with Firefox) is very inconsistent.
It is very much a second-class experience compared to Windows, and it shouldn’t be.
Linux and Nvidia really need to sort out their shit so I can fully dump windows.
Luckily the AI hype is good for something in this regard, since running gpus on Linux servers is suddenly much more important.
Humans are intelligent animals, but humans are not only intelligent animals. We do not make decisions and choose which beliefs to hold based solely on sober analysis of facts.
That doesn’t change the general point that a model given the vast corpus of human knowledge will prefer the most oft-repeated bits to the true bits, whereas we humans have muddled our way through to some modicum of understanding of the world around us by not doing that.
But the most current information does not mean it is the most correct information.
I could publish 100 papers on Arxiv claiming the Earth is, in fact, a cube - but that doesn’t make it true even though it is more recent than the sphere claims.
Some mechanism must decide what is true and send that information to train the model - that act of deciding is where the actual intelligence in this process lives. Today that decision is made by humans, they curate the datasets used to train the model.
There’s no intelligence in these current models.
A token is not a concept. A token is a word or word fragment that occured often in free text and was assigned a number. Common words, prefixes, and suffixes are the vast majority of tokens, and the rest are uncommon pairs of letters.
The algorithm to generate tokens is essentially compression, there is no semantic meaning embedded in them.
Copilot is GPT under the hood, it just starts with a search step that finds (hopefully) relevant content and then passes that to GPT for summarization.
Every billion parameters needs about 2 GB of VRAM - if using bfloat16 representation. 16 bits per parameter, 8 bits per byte -> 2 bytes per parameter.
1 billion parameters ~ 2 Billion bytes ~ 2 GB.
From the name, this model has 72 Billion parameters, so ~144 GB of VRAM
Role of thumb is an employee costs roughly twice their base salary, as the employee still needs to cover insurance, taxes, sick time, and other benefits.
That leaves an average salary of 190K for the 50 employees. That isn’t much for tech.
AND you’re assuming youtube wants to continue the already unsustainable ad-based model at all
No, I was explaining how people who do not watch ads are still valuable to YouTube today. It doesn’t matter if they want to move away from serving ads in the future or not, the points above are still valid.
Netflix is actually a great parallel. They need people to watch the shows and buzz about them to draw in more subscribers. YouTube is the same way, they need people sharing videos and funny comments to scrape attention away from other bits of entertainment.
Further, this isn’t a binary outcome. Each time YouTube makes it a little harder to block ads, a slice of people who don’t want to put in the effort will start watching them. It is trivial, on the software side, to fully block a video from playing if the ad is not served. To date, they have not done that, and I sincerely doubt they ever will - because ad-free viewers are still valuable.
Yes, they would prefer if everyone watched ads. But they would still prefer ad-free viewers to watch YouTube and add to the network effect than to spend their time elsewhere.
‘Those people’ are still incredibly valuable for YouTube.
They watch content, and interact with creators which increases the health of the community and draws in more viewers - some of whom will watch ads.
They choose to spend their time on YouTube, increasing the chances they share videos, talk about videos, and otherwise increase the cultural mindshare of the platform.
Lastly, by removing themselves from the advertising pool, they boost the engagement rates on the ads themselves. This allows YouTube to charge more to serve ads.
Forcing everyone who currently uses an adblocker to watch ads wouldn’t actually help YouTube make more money, it would just piss off advertisers as they would be paying to showore ads to an unengaged audience that wouldn’t interact with those ads.
Explaining what happens in a neural net is trivial. All they do is approximate (generally) nonlinear functions with a long series of multiplications and some rectification operations.
That isn’t the hard part, you can track all of the math at each step.
The hard part is stating a simple explanation for the semantic meaning of each operation.
When a human solves a problem, we like to think that it occurs in discrete steps with simple goals: “First I will draw a diagram and put in the known information, then I will write the governing equations, then simplify them for the physics of the problem”, and so on.
Neural nets don’t appear to solve problems that way, each atomic operation does not have that semantic meaning. That is the root of all the reporting about how they are such ‘black boxes’ and researchers ‘don’t understand’ how they work.
In the language of classical probability theory: the models learn the probability distribution of words in language from their training data, and then approximate this distribution using their parameters and network structure.
When given a prompt, they then calculate the conditional probabilities of the next word, given the words they have already seen, and sample from that space.
It is a rather simple idea, all of the complexity comes from trying to give the high-dimensional vector operations (that it is doing to calculate conditional probabilities) a human meaning.
No, it isn’t. The key conceit is they are removing water from the river and evaporating it.
The water isn’t ‘lost’ it is still part of the hydrosphere, but it is made non-local. That water goes into the air and will go on to be rain in some place far away from the community where it was sourced. This will absolutely contrubute to local droughts and water insecurity.
The porn bit gets headlines, but it isn’t the core of the issue.
All of these models retain a representation of the original training data in their parameters, which makes training a violation of copyright unless it was explicitly authorized. The law just hasn’t caught up yet, since it is easy to obfuscate this fact with model mumbo-jumbo in between feeding in voices and generating arbitrary output.
The big AI players are betting that they will be able to entrench themselves with a massive data advantage before regulation locks down training and effectively kills any future competition. They will already have their models, and the worst case at that point is paying some royalties to people whose data was used in training.
There are really only 3 search providers, Google, Bing, and Yandex.
All others will pay one of these three to use their indexes, since creating and maintaining that index is incredibly expensive.