Fabio Manganiello on Nostr: Funny. Training and running ChatGPT alone pollutes as much as a small developing ...
Funny. Training and running ChatGPT alone pollutes as much as a small developing country, but Microsoft can't afford to give more than 20GB of cloud space to researchers (who btw are still mandated to keep all the results of their studies for 5 years) because "data centers are destroying the environment and cloud space should be used more responsibly".
Why not start with pulling the plug from all the AI hype that is literally boiling the oceans, and maybe pair it with a firm condemnation against cryptocurrencies still running on PoW? Or maybe think of how to use resources more efficiently - do we really need to destroy and recreate containers that often?
We all have our share of guilt and sins, but I don't think that the biggest culprits of the unsustainability of our tech stack are the poor researchers who dumped more than 20GB of raw data on their Microsoft cloud storage.
20GB of data, if not more, is a likely weight for an average data-intensive research, especially if some large ML models were generated/involved or if large raw datasets were used. What researchers will do is probably store this kind of stuff on their own USB sticks like it's 2003, if given no alternatives. So much for security.
It's probably time for Microsoft to GTFO academia and institutional places. A single dumb decision by a single dumb executive (which may be motivated by marketing, public perception, or any other dumb short-term hype that the jerky corporate world loves so much) is enough to jeopardize the way millions of researchers operate.
I'm wondering what's preventing more academic institutions from plugging their own Nextcloud instances, save money, have full control over their data and scale hardware as they need, other than Microsoft sharks that always manage to get into any room where any deal about cloud infrastructure is discussed.
https://scholar.social/@researchfairy/111778617625312456
Why not start with pulling the plug from all the AI hype that is literally boiling the oceans, and maybe pair it with a firm condemnation against cryptocurrencies still running on PoW? Or maybe think of how to use resources more efficiently - do we really need to destroy and recreate containers that often?
We all have our share of guilt and sins, but I don't think that the biggest culprits of the unsustainability of our tech stack are the poor researchers who dumped more than 20GB of raw data on their Microsoft cloud storage.
20GB of data, if not more, is a likely weight for an average data-intensive research, especially if some large ML models were generated/involved or if large raw datasets were used. What researchers will do is probably store this kind of stuff on their own USB sticks like it's 2003, if given no alternatives. So much for security.
It's probably time for Microsoft to GTFO academia and institutional places. A single dumb decision by a single dumb executive (which may be motivated by marketing, public perception, or any other dumb short-term hype that the jerky corporate world loves so much) is enough to jeopardize the way millions of researchers operate.
I'm wondering what's preventing more academic institutions from plugging their own Nextcloud instances, save money, have full control over their data and scale hardware as they need, other than Microsoft sharks that always manage to get into any room where any deal about cloud infrastructure is discussed.
https://scholar.social/@researchfairy/111778617625312456