QuadratureSurfer
@QuadratureSurfer@lemmy.world
- Comment on STALKER 2: Heart of Chornobyl Potentially Banned in Russia Due To Potential 'Justifying Terrorism' 5 days ago:
The physical copy comes with some goodies, or else you can pre-order through GoG, Steam, Epic Games, MS Store, or Xbox.
- Comment on Vehicle combat in Scaravan 66 looks nuts and I want it now 1 week ago:
Direct link to the trailer:
www.youtube.com/watch?v=InEY6skrfl8 - Comment on Why is Nintendo targeting this YouTuber? 3 weeks ago:
But he does have a guess as to why Nintendo targeted him. The first copyright strike landed on his video about the MIG Dumper and the MIG Flash, a pair of devices that let you turn genuine Nintendo Switch cartridges into digital files and then carry around an entire library of those ROMs in a special microSD-equipped flash cartridge for your console. I’ve watched the video, and while Crandall does explicitly take an anti-piracy stance, it’s easy to imagine these gadgets being used by bad actors, too.
“I think the first strike was simply due to the fact that they wanted to minimize attention around the MIG Flash cartridge and dumper […]"
Looks like he also mirrors his channel on Odysee, so the removed videos can still be seen here: odysee.com/…/mig-flash-v2-and-cartridge-dumper-re…
- Comment on AI Models Falter Answering Election Questions in Spanish. 3 weeks ago:
I understand, that definitely makes sense then.
- Comment on AI Models Falter Answering Election Questions in Spanish. 3 weeks ago:
This isn’t an article about mistranslations.
This is an article focusing on how asking about US election questions in Spanish will give you answers that are for the wrong country, or just wrong in most cases when compared to asking the same question in English.
One example is that, if someone in Puerto Rico were to ask ChatGPT 4/Claude/Gemini/Llama/Mixtral a US Election question, it would respond with information for Venezuela/Mexico/Spain instead.
- Comment on Concerns about medical note-taking tool raised after researcher discovers it invents things no one said — Nabla is powered by OpenAI's Whisper 3 weeks ago:
Same, I’d say it’s way better than most other transcription tools I’ve used, but it does need to be monitored to catch when it starts going off the rails.
- Comment on Researchers say an AI-powered transcription tool used in hospitals invents things no one ever said 3 weeks ago:
Whisper isn’t a large language model.
It’s a speech to text (STT) model.
- Comment on Researchers say an AI-powered transcription tool used in hospitals invents things no one ever said 3 weeks ago:
As someone who uses Whisper fairly often, it’s obvious that they’ve trained off of a bunch of YouTube videos.
Most of the time it’s very accurate, but there have definitely been a few times in long transcription sessions where it will randomly hallucinate that someone is saying “Don’t forget to like and subscribe!” When nothing was said anywhere near that.
- Comment on Anthropic's AI can now run and write code 4 weeks ago:
Ok, this is a bit more than what the title implies. This isn’t just outputting the code in text, but rather the ability to verify its own answers before responding to the user asking questions using code.
Claude could attempt these tasks before. But, because it lacked a mechanism to mathematically verify the results, the answers weren’t always incredibly accurate.
So now if you ask it a math question or for it to create some visual bar chart, it will actually do the work to verify that what it’s saying is valid.
I’m sure there will still be ways to trip it up, but this is a good step forward.
- Comment on World of Warcraft adds $90 mount to in game store 4 weeks ago:
It wouldn’t bother me if you were still able to earn the mount in-game.
- Comment on Dragon Age: The Veilguard | Official Launch Trailer 4 weeks ago:
Looks like they’re also planning on releasing without any DRM:
- Comment on 'It even breaks my heart a bit': Denuvo pushes back on its haters, says Steam forums are a 'very toxic, very hostile environment' 4 weeks ago:
What do you mean?
All I see are hearts and love ❤️❤️❤️❤️. /s - Comment on Tesla, Warner Bros sued for using AI ripoff of iconic Blade Runner imagery, despite the producers having previously rejected any association between their iconic sci-fi movie and Musk or his companies 4 weeks ago:
You’re right, whether it’s AI generated or not doesn’t matter.
This is a copyright infringment matter in which “Fair Use” will become a major factor. fairuse.stanford.edu/overview/…/four-factors/
In this case, if the courts rule in favor of Alcon there’s a danger that this expands how copyright law is judged and future cases can use that ruling in their favor. It would make it a lot easier for them to only prove that someone wanted an image that “looks like” even when the image wouldn’t normally be held to that level of scrutiny at face value.
You’re right that there are other factors at play here:
The “Hollywood talent pool market generally is less likely to deal with Alcon, or parts of the market may be, if they believe or are confused as to whether, Alcon has an affiliation with Tesla or Musk,” the complaint said.
They are absolutely concerned that Musk is trying to associate his product with Blade Runner and if the case hinges on the association rather than the image in question then I don’t see a problem with that.
But it’s very concerning that the image itself seems to be a major factor in this case, specifically that they are accusing “(WBD) of conspiring with Musk and Tesla to steal the image and infringe Alcon’s copyright”.
So you saying that anything AI generated that is similar to something else will get sued for copyright infringement makes no sense, unless you can already do that for hand drawn images.
Yes, you can already sue someone else for copyright infringment with hand drawn images. What matters for the decision are a number of factors (as listed out on that link to fair use) one of them being how closely your drawing resembles the copyrighted material. Here’s an article about a photographer who successfully sued a painter who plagiarized her work: boingboing.net/…/photographer-wins-lawsuit-agains…
- Comment on Tesla, Warner Bros sued for using AI ripoff of iconic Blade Runner imagery, despite the producers having previously rejected any association between their iconic sci-fi movie and Musk or his companies 4 weeks ago:
@kameecoding@lemmy.world exactly this.
In the U.S. we have what’s known as a legal" precedent". If a court case makes a decision on something, it massively increases the chances that other courts will use that same decision in similar future cases.
- Comment on Tesla, Warner Bros sued for using AI ripoff of iconic Blade Runner imagery, despite the producers having previously rejected any association between their iconic sci-fi movie and Musk or his companies 4 weeks ago:
The producers think the image was likely generated—“even possibly by Musk himself”—by “asking an AI image generation engine to make ‘an image from the K surveying ruined Las Vegas sequence of Blade Runner 2049,’ or some closely equivalent input direction,” the lawsuit said.
In my opinion, I hope that this lawsuit fails. I know that the movie industry already follows similar practices to what Musk has done. If a studio goes to a certain musician and the price is too high to include their music in the show, they’ll go to a different artist and ask them to create a song that sounds like the song that they originally wanted.
If this lawsuit succeeds it’s going to open the door for them to sue anyone that makes art that’s remotely close to their copyrighted work. All they will need to do is claim that it “might have been created by AI with a prompt specifying our work” without actually having to have any proof beforehand.
- Comment on Tesla, Warner Bros sued for using AI ripoff of iconic Blade Runner imagery, despite the producers having previously rejected any association between their iconic sci-fi movie and Musk or his companies 4 weeks ago:
- Comment on Here’s Gary Oldman fighting space Goombas in over an hour of new Star Citizen Squadron 42 gameplay 4 weeks ago:
Here’s a direct link to the video: www.youtube.com/watch?v=1H-0x4xk2Xk
- Comment on Many companies won't say if they'll comply with California's AI training transparency law 1 month ago:
It would be interesting to see just how often Instagram sold its users’ images to companies creating generative image models.
I see a lot of artists complaining that their work was used for training, but they also agreed to upload their photos to a company along with a royalty free license.
- Comment on YouTube apologizes for falsely banning channels for spam, canceling subscriptions 1 month ago:
Another reason why content creators should be mirroring/uploading their videos to other platforms like Nebula/PeerTube/Odysee/etc.
- Comment on Google is testing verified checkmarks in search 1 month ago:
And that’s when scammers/disinformation campaigns can really take advantage of this system.
People are more likely to trust a “blue check mark”, so finding every way to trick the system into either giving one, or making it look like they have one will become a high priority.
- Comment on California governor vetoes major AI safety bill 1 month ago:
What are you talking about? AI is way too broad of a term to be talking in generics like this.
Are you talking about ML models that help with diagnosing patients, find new cures, find new planets, upscaling, motion prediction, self driving cars, image generation, object detection, text generation (ChatGPT), audio generation (music, voices), etc?
If you’re talking about some “AI” that’s meant to replace a system where we already have a simple solution (like a rain sensor) then, yeah, training a model on that is not worth the effort.
- Comment on Palworld developer has no idea why Nintendo’s suing over its Pokémon-like game 2 months ago:
Though Nintendo’s not going after Pocketpair because lamballs look uncomfortably similar to wooloos, we do know the company is famously litigious in protecting its brand.
So no one really knows what exactly Nintendo is accusing them of yet.
Here’s an older video by an attorney that did a deep dive on Palworld and its legality: www.youtube.com/watch?v=ns3ATH556NU (Warning: long video).
- Comment on Violent racing simulator turns force feedback up to 11 with an ejection seat — sim project debuts in crash and smash video 2 months ago:
Direct link to the video: www.youtube.com/watch?v=TnkXpJCoY9c
but the Tank Simulator also looked great: www.youtube.com/watch?v=NXGwaJBB_J0
- Comment on Flappy Bird’s original creator says he has nothing to do with the new game 2 months ago:
Flappy Bird wasn’t the first to even implement a game like this.
I remember playing a very similar game with a helicopter and another one with a worm long before that app released for phones.
Flappy Bird is just the game of this type that ended up going viral in its time.
- Comment on Dell will continue to reduce its workforce amid push to focus on AI 2 months ago:
I think this title is being easily misunderstood.
Dell is not cutting jobs by having AI compensate for those positions, they’re cutting down on sales teams and creating a new team that focuses on AI products.
From the article:
The latest job cuts appear to be part of a broader reorganization of Dell’s sales teams, including the creation of a new group focused on AI products and services – an area where Dell plans significant growth. For instance, in June, Dell, alongside Supermicro, was selected to provide hardware infrastructure for Elon Musk’s xAI startup’s AI supercomputer.
- Comment on Would you recommend any of the Star Trek games? 2 months ago:
I see “TOS point and click” and can only think of “Terms of Service point and click”… what does this actually stand for?
The other side? TOp Side? It’s definitely not “Thoracic outlet syndrome”…
- Comment on The Irony of 'You Wouldn't Download a Car' Making a Comeback in AI Debates 2 months ago:
Ok, but the most important part of that research paper is published on the github repository, which explains how to provide audio data and text data to recreate any STT model in the same way that they have done.
See the “Approach” section of the github repository: github.com/openai/whisper?tab=readme-ov-file#appr…
And the Traning Data section of their github: github.com/openai/whisper/blob/…/model-card.md#tr…
With this you don’t really need to use the paper hosted on arxiv, you have enough information on how to train/modify the model.
There are guides on how to Finetune the model yourself: huggingface.co/blog/fine-tune-whisper
Which, from what I understand on the link to the OSAID, is exactly what they are asking for. The ability to retrain/finetune a model fits this definition very well:
The preferred form of making modifications to a machine-learning system is:
- Data information […]
- Code […]
- Weights […]
All 3 of those have been provided.
- Comment on The Irony of 'You Wouldn't Download a Car' Making a Comeback in AI Debates 2 months ago:
I don’t understand. What’s missing from the code, model, and weights provided to make this “open source” by the definition of your first link? it seems to meet all of those requirements.
As for the OSAID, the exact training dataset is not required, per your quote, they just need to provide enough information that someone else could train the model using a “similar dataset”.
- Comment on The Irony of 'You Wouldn't Download a Car' Making a Comeback in AI Debates 2 months ago:
I did a quick check on the license for Whisper:
Whisper’s code and model weights are released under the MIT License. See LICENSE for further details.
So that definitely meets the Open Source Definition on your first link.
And it looks like it also meets the definition of open source as per your second link.
Additional WER/CER metrics corresponding to the other models and datasets can be found in Appendix D.1, D.2, and D.4 of the paper, as well as the BLEU (Bilingual Evaluation Understudy) scores for translation in Appendix D.3.
- Comment on The Irony of 'You Wouldn't Download a Car' Making a Comeback in AI Debates 2 months ago:
The STT (speech to text) model that they created is open source (Whisper) as well as a few others:
