Typical ML argument: "If I can read something legally, why can't I train an LLM on it?"

Humans are capable of reading things and later writing a similar thing that is still a copyright violation. If I go and write a book that follows the plot line of Star Wars, that's still a copyright violation, even if no text is literally the same. If I play the melody to a song on my piano and release it without the appropriate mechanical cover license, that's also a copyright violation.

The reason this does not happen often is that, as humans, we are aware that that's plagiarism and there are rules. Sometimes it happens by accident, and people still get sued and lose.

LLMs have no such awareness and routinely output things which are blatant copyright violations when appropriately prompted. That means the model weights encode that work, and therefore, are themselves a derivative work.

Your brain encodes a massive amount of copyrighted information. You are not a walking copyright violation because humans aren't data, can't be copied and distributed en masse, have human rights, etc. This is why "mind reading machines" are a classic dystopian plot point (monetizing your thoughts etc).

An LLM is not a human, does not have human rights, nor human privileges. It is data, and if it encodes copyrighted information, that's a derivative work. If you aren't following the license of the training data, that's a copyright violation.

The quest to write something that is, perhaps, "non-traditional", or maybe "odd", while not coming off as an attempt to be "deep", "poetic", "pretentious" in any capacity. Just pure, boring, empty prose taking up space -- just like this drawing on whatever timeline/feed/context it is being consumed through.

#art #abstract #ink

Commission for one of my usual clients! I really liked drawing these two!

#PixelArt #Art #Commission #MastoArt #FediArt #CrossCode

We strongly oppose the Unified Attestation initiative and call for app developers supporting privacy, security and freedom on mobile to avoid it. Companies selling phones should not be deciding which operating systems people are allowed to use for apps.

https://uattest.net/

"The most wildly successful project I’ve ever released is no longer mine. In all my years of building things and sharing them online, I have never felt so violated."

https://beyondloom.com/blog/onwigglypaint.html

Edit: I am not the author of this. Please go check out https://beyondloom.com/ for more of the author's work.

You're paying AI companies a monthly subscription fee to be fingerprinted like a parolee.

I got bored and ran uBlock across Claude, ChatGPT, and Gemini simultaneously.

Claude:

• Six parallel telemetry pipelines.
• A tracking GIF with 40 browser fingerprint data points baked into the URL, routed through a CDN proxy alias specifically to make it harder to block.
• Intercom running a persistent WebSocket whether you use it or not.
• Honeycomb distributed tracing on a chat UI because apparently your conversation needs the same observability stack as a payments microservice.

ChatGPT:

• proxies telemetry through their own backend to hide the Datadog destination URL from blockers.
• uBlock had to deploy scriptlet injection — actual JS injected into the page to intercept fetch() at the API level — because a network rule wasn't enough.
• Also ships your usage data to Google Analytics. OpenAI. To Google. You cannot make this up.
• Also runs a proof-of-work challenge before you're allowed to type anything.

Gemini:

• play.google.com/log getting hammered with your full session behavior, authenticated with three SAPISIDHASH token variants, piped directly into the Google identity supergraph that correlates everything you've ever done across every Google product since 2004.
• Also creates a Web App Activity record in your Google account timeline. Also has "ads" in one of the telemetry endpoint subdomains.

When uBlock blocks Gemini's requests, the JS exceptions bubble up and Gemini dutifully tries to POST the error details back to Google. uBlock blocks that too. The error messages contain the internal codenames for every upsell popup that failed to load.

KETCHUP_DISCOVERY_CARD.
MUSTARD_DISCOVERY_CARD.
MAYO_DISCOVERY_CARD.

Google named their subscription upsell popups after condiments and I found out because their error handler snitched on them.

All three of these products cost money.
One of them is also running ad infrastructure.

Touch grass. Install @ublockorigin

#infosec #privacy #selfhosted #foss #surveillance

I don't like trump. I didn't vote for him. I disagree with many of the things he's doing. But attacking and toppling down the regime in Iran is 100% justified. It's about friggin time that an American president will take action until this brutal regime. It is incredible to see leftists like Hasan Piker and his ilk totally losing their minds over this, justifying Iran and taking the side of North Korea. Very revealing.

Free Iran.