Steal shit, get shit stole.
The thieves are crying foul because their stolen goods get stolen yet again
“if we paid for every piece of data then we can’t have LLMs!”
“they’ve paid for their LLM usage but doing it in a way I don’t like!”
If the Alibaba linked accounts paid for the tokens, then how they used said tokens should not matter. These so called “distillation attacks” aren’t really attacks if they are paying for the compute time.
Your plagiarism machine copied my plagiarism machine!
Let me get my Planck-length violin.
That’s rich.
Oh no, this evil company stole our stolen data!
Its the 7-step AI leadership program:
1. Create an AI model that can generate data.
2. Brazenly and illicitly steal terabytes of copyrighted data to build a usable product.
3. Publish said model on the internet.
4. Get rich of stocks and user fees, without paying a dime to the actual data-owners.
5. Discover that your competitors brazenly and illicitly stole terabytes of your generated data to build their own product at a fraction of the cost.
6. Watch them get rich of stocks and user fees, without paying you a dime.
7. Cry deeply.
Oh the irony!
It’s funny because Alibaba’s open source model Qwen is one of the best open models that are approaching the paid service’s capabilities.
<Greedily licking up Anthropic’s crocodile tears>
So?
Pot. Kettle. Black.
That image looks like temu jerry seinfeld telling a joke.
