It does make me think that this decline in nutrient values isn't so big a problem though? Since diversity and quantity makes up for that decline

When I visited the site I was expecting to see people having more nutritional problems as a result of the decline in the nutritional value of food, but that wasn't the case.
Though you're right that they're different, so can't really draw conclusions either way from that data.

I checked ourworldindata.org/micronutrien... to see if it corroborates. Frankly, I left with more questions than answers, but seems that for most of the world nutrient intake has gotten better over the last ~20 years.

Is it not possible for a mind to recognize a good quality about the culture from a certain past period without yearning for that culture wholesale?

Sora is here and will only get better. Video provenance will become weird. LLMs for coding are imperfect but are serious productivity boosts. These are just the first things to really matter

hey that worked, thanks

Looks really nice. How do I sign up? Clicking the sign-up button takes me to bluesky

(This view is also consistent with some AI policy folks' observation that China's industry is building products rather than AGI. Of course this may change)

4. the moat is everything surrounding the model (infra, agent scaffolding, product design...) and the ability to integrate them. despite the marketing, the model is seen as just another software component rather than the secret to their success

a few guesses:
1. open weight isn't open source, releasing their model weights won't enable their competitors to learn their edge
2. the moat is the data center, so releasing models is fine
3. the moat is also the talents in the team. releasing open models attract talents instead

They may be treating language models similarly. Fine-tuning and applying the open source models contribute to the companies' knowledge.

Now the surprising part is that these models seem like the main product of these companies - what enabled their moat. Why would they destroy their moat via OS? 1/n

I've worked in Chinese tech startup environment. One observation is that open sourcing libraries and tools allow more developers to find bugs and debate / develop features, benefiting the companies themselves.

Well said.

One can also pay attention to the fact that it's pre-trained on public data, and its coding ability comes from retrieval of these code. Many AI researchers have proposed that the verbosity and strange designs are artifacts of its training data.

Let's try it in a month :D

none of the US commercial lab models (gemini, claude, chatgpt) understood the joke

ordinary human wins

good choice

The original post got deleted. Does anyone know what happened?

📌