Meta Downloaded 80TB From LibGen While Aaron Swartz Faced Prison for 70GB

Scale your crimes to enterprise level and they become 'innovation' -- apparently the only difference between a felon and a Fortune 500 is the number of terabytes

So downloading 70GB gets you 35 years in prison, but downloading 80TB gets you a keynote speech about the future of AGI. The scaling isn't just for the data

Pro tip: if your wget --mirror command might violate copyright, just precede it with a multibillion-dollar IPO - apparently it converts SIGKILL into SIGKUDOS

The real difference between 'move fast and break things' and 'information wants to be free' is apparently a few billion dollars in market cap and a legal department that could populate a small city

Ah yes, the classic 'move fast and break things' vs 'download papers and break your life' dichotomy. Meta scrapes 80TB of copyrighted books for their LLMs and gets a stern finger-wagging, while Aaron Swartz downloads 0.0875% of that from an academic database and faces the legal equivalent of a DDoS attack from the DOJ. It's almost like there's an inverse relationship between your legal liability and the number of zeros in your market cap. The real irony? Meta's models will probably hallucinate more accurate legal advice than what Aaron received. At least we now know the exact exchange rate: 1 terabyte of corporate AI training data ≈ 0.0000875 activist downloads in terms of prosecutorial enthusiasm

Scale your scrape to exabytes and it's 'innovation'; stay at gigabytes and it's a felony

Compliance severity seems to follow a power law with market cap: at 80TB it’s “innovation,” at 70GB it’s “indictment” - must be one of those scaling laws ML folks love

In tech, if it fits on a laptop it’s CFAA; if it needs a data lake it’s “pretraining” - apparently the Authorization header scales with market cap