Here's an EIP for adding a separate gas type to calldata, and putting the three gas types into a single harmonized system:

https://github.com/ethereum/EIPs/blob/4020064d72da32adc691cf1906f466a2c03b00f5/EIPS/eip-7706.md

Hopefully it can make the discussion on what "multidimensional gas" might look like more concrete!

Moving from downtown SF to a suburb of SF does not count as "leaving San Francisco".

You're still using the same airport. You're still able to have in-person meetings with the same people. Travel time from SF downtown to your new place is significantly less than between two opposite ends of Beijing or Seoul.

I did this last year and couldn't agree more. I meet more people in SF now than when I was living in the city.

I'm in San Mateo now which has a very central location, 10min from airport and ~30min away from pretty much everywhere in the bay area. So I also +easily meet people in Palo Alto, Oakland, and NYC now :)

d/acc

"What if some actor that you don't trust. If they're the ones who have the super strong AI, whether it's some like other government that is an opponent of our country or some company that you don't trust. 

I think that's potentially a much bigger risk" - Zuck

Zuck is /dacc 

Talks about infosec and biosec in latest pod

We just published a new ROP (RIG Open Problem). The first one focused on blockchain infrastructure for AI

We already have one amazing team working on routing/MEV and hope to fund other relevant projects

DM if interested

We just published a new ROP (RIG Open Problem). The first one focused on blockchain infrastructure for AI

We already have one amazing team working on routing/MEV and hope to fund other relevant projects

DM if interested

https://efdn.notion.site/ROP-12-Credible-LLM-Markets-and-AI-MEV-1ce249f3492e48a488f69d02fb18e154

I realize PeerDAS is not nearly ready yet, but I hope we see it fast-tracked for inclusion in the not too distant future. 384kbps of blob space per block isn't going to cut it for long and we need to get to the point where every validator doesn't need to download every single bit of chain data.

Without safeguards in place, language models will always attempt to promote the ideological interests of their creators

Think of this, but personally tailored to hundreds of millions of people 

https://youtu.be/ZggCipbiHwE

The capacity of a blockchain is equal to it's direct capacity plus the capacity of all additional layers under its security

Base is Ethereum

So are all other systems that settle or are secured by Ethereum

It's a diverse and ever expanding ecosystem yet cohesive

Nothing seen before in crypto or otherwise

Live it!

Are validators in the same cluster (eg same exchange, same user) more likely than unrelated validators to miss attestations at the same time? If so, can we tweak rewards to favor decentralized staking?

Possibly yes.

https://ethresear.ch/t/supporting-decentralized-staking-through-more-anti-correlation-incentives/19116

Not going to beat centralized AI with more centralized AI.

All in on #DecentralizedAI 

Lots more 🔜

https://stability.ai/news/stabilityai-announcement

@emostaque resigning from Stability AI is great news for /dacc

Not your models not your mind

#decentralizeAI

Regarding Base transaction fees: they are high because the chain is at its target capacity (2x the throughput of the L1), but demand has increased dramatically. Data fees remain insignificant thanks to EIP-4844 blobs which are still far from capacity!  So while "it could be worse..." what can make it better? (cont)

Gud thread to understand the problems rollups are going through as they significantly scale tx volume post 4844

And their proposed solutions in the short and medium term

Ethereum protocol R&D. Admirer of emergent order.

This is a crucial fork in the road, because it significantly alters the way we should govern the L1 EVM:

If L2s will over time diverge from L1 (towards a more advanced "L2 EVM"), L1 EVM can start to ossify.
If L2s and L1 however share a VM forever, we will need to build a shared EVM improvement and governance process.

There is also an option 3: multi-fast-evm superchain

It's been silent here! Linking the initial analysis of stake distribution from @julianma.

Analysis of different types of stakers is important when discussing potential changes to issuance.

It's been silent here! Linking the initial analysis of stake distribution from @julianma.

Analysis of different types of stakers is important when discussing potential changes to issuance.

https://ethresear.ch/t/initial-analysis-of-stake-distribution/19014

Dencun went live, we've had blobs land in blocks on mainnet .oO 🎉

Being able to train a 70b model on 2x consumer GPUs seems like a big deal and important milestone towards a diversity of actually user aligned personalized AI systems that can power tools for info/cyber defense.

https://www.answer.ai/posts/2024-03-06-fsdp-qlora.html

jeremy howard continues to do amazing work!

we need to get him on farcaster and /dacc

In my latest post I make the case that the integration of blockchains with AIs will give rise to an Internet of Agents and discuss a path towards it.

Based on research done over the past year & examples from innovative projects.

In my latest post I make the case that the integration of blockchains with AIs will give rise to an Internet of Agents and discuss a path towards it.

Based on research done over the past year & examples from innovative projects.

https://davidecrapis.notion.site/The-Internet-of-Agents-23aa09799b9c4620a1a287926bcfd6af

Finally got around to uploading the corresponding video: https://www.youtube.com/watch?v=l8uRkvrnc_c

I love the "computer in the sky" metaphor. Great content from Tim!

I'm speaking at a Decentralized AI event at Stanford today. With some super interesting topics and speakers.

Will be posting my content and my general takes in the coming days.

I'm speaking at a Decentralized AI event at Stanford today. With some super interesting topics and speakers.

Will be posting my content and my general takes in the coming days.

https://lu.ma/opensourceaisummit

Andrej K makes an interesting point and Autism Capital's dismissive reply is wrong.

It is rational to be more afraid of man than of nature, because it is rational to be afraid of powerful optimizers.

This is also often why we are more suspicious of for-profit versions than non-profit versions of the same activity.

https://ethresear.ch/t/economic-analysis-of-execution-tickets/18894

We published the first economic analysis of Execution Tickets. First step towards designing robust pricing and allocation mechanisms.

Reach out if interested. We have more in the workings.

ethereum protocol research /ethrd + Crypto x AI /dacc 

Davide Crapis

@davidecrapis