When I was at Google, I was asking WHY they were interested in building THE LARGEST model. Why?
Why were you going for size?
What function are you trying to achieve?
Why was the thing you were interested in having THE LARGEST model?
Why is the thing you were upset about that you didn't have THE LARGEST model at the time?
They responded by doing me the favor of firing me 🤷🏼
=> More informations about this toot | More toots from timnitGebru@dair-community.social
@timnitGebru I have a theory that the costs of building large models create a barrier to entry, and so help corporations maintain market dominance.
=> More informations about this toot | More toots from PeteBleackley@wandering.shop
@PeteBleackley @timnitGebru It also makes business for their overpriced data centers and lets you launder money that way.
=> More informations about this toot | More toots from dalias@hachyderm.io
@dalias @PeteBleackley @timnitGebru when you have an oversized datacenter hammer, everything starts to look like a nail.
like some AI version of Conway's law
=> More informations about this toot | More toots from risottobias@tech.lgbt
@timnitGebru in the current paranoia, they're all going to lose to solutions that you can run locally and don't have to ship to someone else's black box at 100x the cost.
Although Google is at least paying lip service to that with some of their Pixel features.
=> More informations about this toot | More toots from craignicol@glasgow.social
@timnitGebru Twenty-ish years ago, I was teaching CS at a small state college. One weekend the student ACM chapter organized a LAN party in the student union building. Another professor and I went to check on things that evening and found one female student sitting amidst a large group of male students. She had a small 15 inch monitor, while most of the male students had large 20+ inch monitors. She gave us a mischievous grin and said “I'm the only one here not compensating for something.”
=> More informations about this toot | More toots from mgrinder@mastodon.social
@timnitGebru you women just don't understand the importance of having the largest.
=> More informations about this toot | More toots from hllizi@hespere.de
@timnitGebru
I'm only just realizing it's a dick thing!
=> More informations about this toot | More toots from Edelruth@mastodon.online
@timnitGebru my personal theory is that doing the same thing with more number is a moderately successful strategy for mercenary-minded individuals to climb the ladders, so the entire bay area has self-selected until it's full of the stupidest yes men with PhDs.
=> More informations about this toot | More toots from elrohir@mastodon.gal
@timnitGebru If it t had been called "long language model" it would have been too obvious
=> More informations about this toot | More toots from harry_karadimas@toot.community
@timnitGebru you need to own a penis to understand
=> More informations about this toot | More toots from enkiusz@is-a.cat
@timnitGebru if you don’t go along with the groupthink you never get to climb the greasy pole.🤷
=> More informations about this toot | More toots from holdenweb@freeradical.zone
@timnitGebru 🤏 🍆
=> More informations about this toot | More toots from dalias@hachyderm.io
@timnitGebru How else to show off THE LARGEST ego? Always a shame how that affects others.
=> More informations about this toot | More toots from simondassow@masto.ai
@timnitGebru my name is ‘Bigus Dickus’
=> More informations about this toot | More toots from fletch@mapstodon.space
@timnitGebru It is what we in the industry call a "dick measuring competition"
=> More informations about this toot | More toots from jmaris@eupolicy.social
@timnitGebru Everytime someone goes for the bigger, I assume it's a classical toxic masculinity manfiestation ...
=> More informations about this toot | More toots from Riduidel@framapiaf.org
@timnitGebru yeah it seems like the dumbest thing to just throw resources at it. I've had interactions with people at Microsoft who are convinced ChatGPT will always outperform any more focused form of AI. Rather than build machine learning models tailored at specific problems (taking a scientific approach to evaluating them) they built a generic LLM believing it will somehow spontaneously spawn 'general intelligence' or 'super intelligence' or whatever. It's just magical thinking at scale.
=> More informations about this toot | More toots from stew_sims@mastodon.social
@timnitGebru sad for them! Any ideas why the largest now?
=> More informations about this toot | More toots from efialto@mastodon.online This content has been proxied by September (3851b).Proxy Information
text/gemini