4 ms·

Uninformed take. How long didn’t it take for ML to actually create meaningful results? Just because a space is riddled with 80% grifters during bull markets doe

by dsco 4y ago

Uninformed take. How long didn’t it take for ML to actually create meaningful results? Just because a space is riddled with 80% grifters during bull markets doesn’t mean you get to discard the 20% who are building decentralized file storage, lending protocols, stablecoins (Dai in particular) etc. Also after so many people trash crypto on HN daily - why’d you think it was even worth our time reading your post? At least some folks are funny with their takes.

LarsDu88 4y ago
Virtually immediately. The most rudimentary Yann Lecunn neural networks have been reading your snail mail addresses since the 90s, and AlexNet derived image recognition has pervaded image recognitions tasks after 2012. Transformers have only existed for 5 years and within literal months became the backend for Google Translate. Within 2 years, the first gpt was created
peoplefromibiza 4y ago
> How long didn’t it take for ML to actually create meaningful results? Very little time. ML has been creating meaningful results since the 1960s. > Just because a space is riddled with 80% grifters It is a strong indication that nothing useful will come out from the remaining 20%, which is probably wishful thinkers. But then again, hope is not proof, lines of code is not a product, hours spent tinkering is not quality, big ideas is not real world impact. Right now, and for the past 10 years, this space has been a lot less useful than what came out from imagining FTL travel or colonizing Mars. But costed nearly as much (actually, more: going to Mars costs around ~2.7 billions, while This year, more than $3.8B has been poured into startups in the Web3 space) If I believed in conspiracies I would say that the criminal cartels of the World have found the perfect tool to wash their dirty money. It works much better than pumping art.
Arubis 4y ago
> decentralized file storage, lending protocols, stablecoins (Dai in particular) I mean...yeah! Go builders! I wish you success. But I have my doubts around these ones. Decentralized file storage seems maybe to be better addressed by IPFS, which uses cryptographic protocols but isn't cryptocurrency. Stablecoins are mathematically and conceptually fascinating (like most of this space!), but what do they get you that you don't get with fiat currency? The answer I've actually seen applicable here (aside from "enable black markets") is to enable fairly paying people across borders that live in places with absurdly distorted exchange rates. So that's a win as a transfer medium; less so for storage (Cf. all the stablecoins that've unpegged recently). I freely admit not being sufficiently educated around the current state of lending protocols in the cryptocurrency space to have an informed opinion.

by dsco 4y ago

LarsDu88 4y ago

Virtually immediately. The most rudimentary Yann Lecunn neural networks have been reading your snail mail addresses since the 90s, and AlexNet derived image recognition has pervaded image recognitions tasks after 2012. Transformers have only existed for 5 years and within literal months became the backend for Google Translate. Within 2 years, the first gpt was created

peoplefromibiza 4y ago

> How long didn’t it take for ML to actually create meaningful results? Very little time. ML has been creating meaningful results since the 1960s. > Just because a space is riddled with 80% grifters It is a strong indication that nothing useful will come out from the remaining 20%, which is probably wishful thinkers. But then again, hope is not proof, lines of code is not a product, hours spent tinkering is not quality, big ideas is not real world impact. Right now, and for the past 10 years, this space has been a lot less useful than what came out from imagining FTL travel or colonizing Mars. But costed nearly as much (actually, more: going to Mars costs around ~2.7 billions, while This year, more than $3.8B has been poured into startups in the Web3 space) If I believed in conspiracies I would say that the criminal cartels of the World have found the perfect tool to wash their dirty money. It works much better than pumping art.

Arubis 4y ago

> decentralized file storage, lending protocols, stablecoins (Dai in particular) I mean...yeah! Go builders! I wish you success. But I have my doubts around these ones. Decentralized file storage seems maybe to be better addressed by IPFS, which uses cryptographic protocols but isn't cryptocurrency. Stablecoins are mathematically and conceptually fascinating (like most of this space!), but what do they get you that you don't get with fiat currency? The answer I've actually seen applicable here (aside from "enable black markets") is to enable fairly paying people across borders that live in places with absurdly distorted exchange rates. So that's a win as a transfer medium; less so for storage (Cf. all the stablecoins that've unpegged recently). I freely admit not being sufficiently educated around the current state of lending protocols in the cryptocurrency space to have an informed opinion.