There is actually a limit, but I think it was too high.
Currently the limit is for the deployment, but your comment gave me the idea that I should give this option to the site owner.
I recently started interacting with JIRA exclusively through AI agents (as a Skill in Cursor) and the difference between Linear and JIRA has immediately evaporated. No UI clunkiness problems if I don't even use it :)
I'd say that if anything AI generated ends up in the final product then it should be labeled as AI made. So using AI as a prototyping tool would be fine but using it either to generate the end product itself or using it to generate a script would require tagging.
I watched a documentary recently that had an indicator in the top right corner when the content was AI generated. I found it incredibly helpful.
Every online video platform should let you label specific segments as AI generated, even better if it is a requirement with validation checks for certain kinds of content.
Yes, it seems like the creator applying the label is the most scalable approach. But you could have the requirement and random sampling with consequences if it isn't followed. You could also sample based on video traffic and user reports.
It's a really valuable feature that I expect will eventually be the gold standard, it was surprising how helpful it was. I think a lot of creators will embrace it, it adds credibility/authenticity. You aren't just labeling the AI content, you are labeling the content that isn't generated by AI, with a validation layer to back it up.
Plenty of big artists like Kanye use AI to experiment with ideas before releasing the full studio recordings. That’s going to become more common. Just like how developers use LLMs to make a POC to test new ideas before putting the hard work into making it real.
- Footage created by "Extend Scene" features in Premiere Pro and others
- Word correction from tools like Descript
- AI relighting or colorization
- Reaction video to a video containing AI-generated content
And in general, what amount of combination of any of these applications constitutes as "AI generated"? If I have a 30 minute video with a 3 second AI generated clip, do I get the same label as full-blown AI slop video?
All of them in any amount and yes, your specific 3 seconds should be labeled as being AI generated in your video (or description) and your entire video should be labeled as made using AI by youtube.
A 30 minute speech by a president where I use AI to change 3 seconds in order to make that person say something they never did should also get the label. The label shouldn't be about how much AI was used, but that it was used at all.
Why does this mean AI-aided color correction, 25fps to 30fps, or upscaling 480p to 1080p should be treated the same?
Using AI to make someone say something different (even if only for a small part of the video) is very clearly different, that's deception.
Why does AI editing 3 seconds of audio from a president's speech be treated different from when a human does it without AI?
It sounds like your argument is about deceptive editing. Why not just require labelling deceptive editing? That way there can be place for comedy channels to do so for comedic purpose (rather than blanket ban on deceptive edits).
Almost any photograph taken with a mobile phone is AI edited because phones come with AI to enhance photo quality (e.g. Deep Fusion, Smart HDR), should any photo taken with a mobile camera be labelled AI now?
> Why does this mean AI-aided color correction, 25fps to 30fps, or upscaling 480p to 1080p should be treated the same?
You may not be old enough to remember seeing it, but notices like "This film has been modified from its original version. It has been formatted to fit this screen" were put there for a reason. It was to avoid legislation which would have forced distributors to disclose that the product they were delivering was not the same as what people saw in theaters because it was considered deceptive to hide that fact. Altering of someone else's video should always be disclosed. That includes upscaling and color alterations even when they aren't done using AI and should absolutely be done when AI is inserting a bunch of fake frames.
When using AI to edit your own content deception is much less of a concern, but yes the use of AI should still be disclosed. There are clearly people who want to know when AI is being used to create something. There's no harm in having the label there, it just makes it easier for people to decide if and how to view your work. What value is there in hiding your use of AI from others in cases where you aren't trying to trick them?
Also yes, cell phone cameras should be required to clearly disclose that the photos they take to not reflect reality, and include the fact that they are edited with AI in meta-data. I've had friends complain that the photos their phones were taking were automatically edited and there was no option to disable those edits. They were upset because they intentionally wanted photos that reflected the real world and ultimately were forced to switch to using other cameras to get the accuracy they needed. As far as I'm concerned cell phone cameras doing that are simply defective. People should be aware of that fact before they use cell phone cameras for things like evidence collection, pictures of medial concerns, and for anything where accuracy matters more than what some algorithm thinks you'd rather see than the truth.
Other than your last bullet point I don't see anything ambiguous. It's a very clear line. I do not want to see an explainer video with AI generated content, end of story.
I absolutely hate those full-blown AI 'explainers' that just have AI voiceover and a bunch of auto-placed b-roll. I don't want to see them. But I don't think that falls in the same bucket as a creative short film with some AI-generated SFX or someone doing a tutorial with an AI-generated lofi track in the background.
I still think MCP is completely unnecessary (and have from the start). The article correctly points out where CLI > MCP but stops short on 2 points:
1. Documenting the interface without MCP. This problem is best solved by the use of Skills which can contain instructions for both CLIs and APIs (or any other integration). Agents only load the relevant details when needed. This also makes it easy to customize the docs for the specific cases you are working with and build skills that use a subset of the tools.
2. Regarding all of the centralization benefits attributed to remote MCPs - you can get the same benefits with a traditional centralized proxy as well. MCP doesn't inherently grant you any of those benefits. If I use AWS sso via CLI, boom all of my permissions are tied to my account, benefit from central management, and have all the observability benefits.
In my mind, use Skills to document what to do and benefit from targeted progressive disclosure, and use CLIs and REST APIs for the actual interaction with services.
> This problem is best solved by the use of Skills which can contain instructions for both CLIs and APIs
You've just reversed the context benefits because the content of the skill...goes into context.
> ...you can get the same benefits with a traditional centralized proxy as well. MCP doesn't inherently grant you any of those benefits.
You've just rebuilt MCP...but bespoke, unstructured, and does not plug into industry tooling. MCP prompts are activated as `/` (slash) commands. MCP resources are activated as `@` (at) references. You can't do this with a proxy.
See the three .gifs at the end of the post to see how clients use MCP prompts and resources and definitely check the specification for these two.
I was surprised by my own feelings at the end of the post. I kind of felt bad for the AI being "put down" in a weird way? Kinda like the feeling you get when you see a robot dog get kicked. Regardless, this has been a fun series to follow - thanks for sharing!
Extensions have too many security risks for me. At this point I'd rather just vibe code my own extension than trust something with so much access and unpredictable ownership.
This model is awesome. I am building an infinite CYOA game and this was a drop-in replacement for my scene image generation. Faster, cheaper, and higher quality than what I was using before!
I didn't expect IBM to be making relevant AI models but this thing is priced at $1 per 4,000,000 output tokens... I'm using it to transcribe handwritten input text and it works very well and super fast.
IBM and Nvidia speech to text models are also SOTA (according to HF leaderboard) and relatively lightweight. Replicate hosts those too, although some (like Parakeet) run easily on consumer GPU.
