Nope. You have no clue how LLM training works. I do.
You'd make a more compelling case if you adopted the words used by people in the industry, people in academia, rather than thinking you know something because you read a dictionary definition at some point.
The AI industry uses the term learning and training. No one gives a fuck what you think about it.
Genuinely curious: Why do you switch between transformative use and fair use? Do you consider the first a subset of the second?
Not the person you're replying to but one is a subset of the second. Fair Use covers a wide variety of cases, one of which is if a copyright work is used in a transformative way.
They are storing the training set, the transformative use is only relevant in that it's fair use. It was deemed legal (for non pirated content) because format shifting for a fair use goal was also fair use.
Fair use is the only defence.
Yes that was my point, it's not a copyright issue both because the use is fair use and because the act of using it necessitates non-fixed copies which is covered under the exception for normal operation.
No, because I'm paying a license (subscription) to Netflix to watch that movie and Netflix is paying a license to the producer.
Correct and thanks for emphasising my point. When you license the use of something you end up licensing all normal operation of that use. Storing things in cache or RAM for the purposes of its use is not copyright infringement, even if technically a "copy" does exist. And the process of training an AI model has been ruled fair use, so no copyright isn't an issue here.
This case can have merits if it argues the specifics of how the copies were acquired.
I'm not sure what other advice to give.
How about Bartz v. Anthropic? Not only did this cover that specific point, but it also involved one of the companies and the exact same process being discussed here, except for books instead of video.
For the record Anthropic won all accounts except for one, turns out they pirated the books used for training. Training models is transformative, if I buy a CD and use it to train an AI I'm good and there's nothing the music industry can do to stop me. If I download a bittorrent through then they may get me for copyright infringement (and Anthropic settled that class action for $1.5bn because they very much did that).
As usual, you are without insight. "Training" and "learning" does not exist in machines.
And as usually you're Dunning-Krugering your way though yet another discussion. Yes both "training" and "learning" not only describe the precise process for developing weights of neural networks in a continuously adjusted model, they are the actual terms used in the industry. You can either adopt them or continue to display your ignorance on the topic.
All that exists is data processing.
But I'm here to help, so let me point out to you that data processing isn't what is being described as raw data processing does not dynamically feedback and adjust algorithms or weights of algorithms. Data processing is deterministic. The training of AI models is not. That's an important distinction you should use as a starting point if you actually feel like you want to learn something about this topic rather than continuously spouting ignorance such as pretending the terms used by the very people developing those algorithms aren't the terms used.
Actually is that even ignorance? Or are you just gaslighting. Either way, be better.
We don't know how the brain works so no it isn't modeled on the brain.
That is stupid. We don't know "exactly" how the brain works, but we do know that they learning and strengthening of skills is the result of training neurons. We know how the brain is made up and we literally modelled AI after it, and even gave the type of process a name named after our biological components: neurons, neural network, ring a bell?
Just because we don't know something 100% doesn't mean we didn't model something else after how we perceive it to work, and perception is the only thing under discussion here since the GP was arguing semantics about language. Right now the process for training AI is the same as we understand it to happen in our brains. Whether we're right or not is completely irrelevant.
Blimps and drones are two very VERY different beasts. One of them is highly subject to winds, and slow moving with high inertia taking a long time to react to any external change in force.
The other is small, zippy, and can hold insanely steady with rapidly changing conditions.
Also these drones will be local launched, they aren't flying long distances high in the sky and won't be subject to the same winds as aircraft or blimps. The bigger question is one of landing areas.
Wait until you hear that there was only an anticipated market for 5 computers globally. Yeah numbers are always stable, and there's no expectation that anything will grow.
Judging a project's success based on it's very early scale out it silly. No one invested this money planning to stop at 1million drone deliveries / year.
Even with this big rollout targeting nearly 500 communities, the operational reality dictates where these things can actually land:
This is America we're talking about. The first commandment has solved this: "I am Ford your car you will worship no other gods before me" has resulted in a metric fuckton of accessible widely sprawling asphalt. Just use one of the endless number of parking lots to drop your parcel.
Also I don't think anyone read the summary, saw the term City and assumed it was the downtown apartments being discussed here. When someone does talk about that area they normally explicitly state it. For common language an American "city" is a vast sprawl of suburbia. This isn't just company talk.
But all you can say is if you don't like it leave?
He didn't say you don't like it leave, he was pointing out something fundamental to the country's founding, namely that it's not communist, while taking a dig at the OP's historically stupid posts.
No rsilvergun spouted rubbish. Capitalism in the US doesn't need a referee because of RAM prices caused by *checks notes* non-US companies.
I bought a MiniPC from China instead
Instead? It probably means you bought the wrong thing originally in the first place. SBCs and MiniPCs are a different product class with different purposes. Anyone who suggests using a MiniPCB instead of a small SBC shouldn't be touching SBCs.
And that is relevant to their current policy matching that of the Google Pixel how? I get it, you're burned and have a jaded view of the company. Don't buy from them. All I'm saying it that currently Samsung's support policy matches that of Google. It wasn't always the case, but it is their official position now. Companies do change (including Google as well).
Nothing is certain. Oh except my watch. There's a reason I wear a mechanical one.
Many people write memos to tell you they have nothing to say.
