Snow Flow
banner
snowflow17.bsky.social
Snow Flow
@snowflow17.bsky.social
Informatiker. Fundiertes Halbwissen.
Reposted by Snow Flow
Deutschland nach den Wahlen: was heißt das für die Ukraine und die Rolle Deutschlands in Europa? Darüber habe ich mit @KEigendorf in unserem podcast Global Deep Dive gesprochen, jetzt online - mit Katrin Eigendorf direkt aus der Ukraine

podcasts.apple.com/de/podcast/g...
3 Jahre Krieg: Wer entscheidet über die Zukunft der Ukraine?
Podcast-Folge · Global Deep Dive - Außenpolitik anders denken · 24.02.2025 · 43 Min.
podcasts.apple.com
February 24, 2025 at 8:48 PM
Reposted by Snow Flow
For vertical farming to work for cereals, energy would need to be insanely cheap.

We underestimate how cheap a lot of outdoor farming is. That's good for feeding 9-10 billion people. But bad news for indoor competitors.

I ran the numbers here:
www.sustainabilitybynumbers.com/p/vertical-f...
February 24, 2025 at 3:28 PM
Reposted by Snow Flow
Today, we have a new page on fertility data on Our World in Data!

It has lots of new charts on a wide range of measures such as parental ages at birth, teenage birth rates, twin birth rates, the use of reproductive technologies, and more.

Explore it all here: ourworldindata.org/fertility-rate
February 24, 2025 at 9:19 AM
Reposted by Snow Flow
A few implications of tricks like this:
1) We are still VERY early in the development of Reasoners
2) There is high value in understanding how humans solve problems & applying that to AI
3) Higher possibility of further exponential growth in AI capabilities as techniques for thinking traces compound
This paper is wild - a Stanford team shows the simplest way to make an open LLM into a reasoning model

They used just 1,000 carefully curated reasoning examples & a trick where if the model tries to stop thinking, they append "Wait" to force it to continue. Near o1 at math. arxiv.org/pdf/2501.19393
February 7, 2025 at 3:23 PM
Reposted by Snow Flow
Good test of the ability of AIs on hard math problems they haven’t seen before*: the brand new qualifying math olympiad test

Reasoners are clearly a big breakthrough. And DeepSeek does well but loses to o3-mini on both ability & cost.

* Though contamination is always impossible to rule out.
February 8, 2025 at 2:53 AM