Podcast notes – Hasu and Mike on MEV (Bell Curve) – “If a single regulatory regime can make rules in crypto, then crypto has just failed”

Hosts: Hasu and Mike
Hasu – advisor to Flashbots, Lido

MEV value chain
-money from reordering / censoring transactions
–any value a privileged actor can extract – eg, Central Bank printing money, can be considered MEV
–

People use crypto to escape MEV in real world

Should build crypto systems resilient to MEV

Principles in reducing MEV
-more competition = lower fees, less MEV
-more private = harder to extract MEV
-more user control

MEV is invisible – even looking at transaction data in Etherscan, won’t see sandwich attack

Parties:
Users
Wallets
Searchers
Builders
Relayers
Validators

MEV schools
1. Democratizing MEV – hard to minimize MEV, isolate builders role, make it competitive
2. Minimize MEV –

User/wallet layer – order flow auctions – users don’t send to public mempool or block builder, auctions off right to execute your transaction, if there are competing bidders, the price rises, and value goes to user (instead of to MEV capturer)

Mike: “Payment for order flow” – Robinhood offering zero fees, selling order flow to Citadel / hedge funds
Mike: In past, equity brokerages would charge you for trades – now people have opted for free trades / invisible fees (eg, Robinhood)
We can do better in Defi – especially the transparency

World of Cosmos and Ethereum are converging – ETH community has been better at executing
Hard to say in future if X project is ETH or Cosmos project – there’s increasing convergence

MEV accrues to whomever gets to order the transactions
Mike: MEV will accrue to execution layer

L2 sequencers today are centralized – with plans to decentralize – will eventually face same MEV problems as ETH L1
L2s all need PBS (proposer builder separation)

Sequencers today in L2 does 4 things
-receive transactions
-decide on ordering of transactions
-give user a receipt
-send order batch to data availability layer — that’s what creates finality

MEV should not be counted towards security budget — that’s how core devs think about it, want to minimize and not enshrine it
Minimum security should be paid from inflation + base fee

“MEV is very hard to track”

Different forms of MEV
-arbitrage – different prices on different exchanges, or underpriced asset
-sandwich attacks – buy before a user, then sell it to the user at higher price
-liquidations – searchers typically do this

Uniswap V3 – concentrated liquidity – reduced sandwich attacks

Statistical arbitrage – take balance sheet risk, small period of time where you have to hold asset before selling it

Many top Defi traders are also block builders – want to maximize inclusion guarantee, greater control over trading strategy – can make trade at last moment, can see all other transactions and order / cancel them

In systems we build, must make sure they’re not sensitive to latency — otherwise there’s incentive to colocate near each other, more centralization
Phil Daian post on this: https://collective.flashbots.net/t/decentralized-crypto-needs-you-to-be-a-geographical-decentralization-maxi/1385

Turn latency into price / auction, auctions are generally more fair, and price (ability to pay) is easier to decentralize than geographic proximity

Users love Robinhood because good feature is very visible (free trades) and bad feature is very invisible (selling user order flow)

Mike: Optimism and Arbitrum have very different approaches to MEV

“Solana is case study for why to not build low latency blockchains”
1 of 2 Solana block builders is operating liquid staking protocol
If you don’t have robust mempool and fee market design, get a lot of spam
58% of Solana transactions are failed arbitrage transactions

What’s novel in Cosmos —
-Osmosis doing something very interesting – onchain block building and searching
-Noma (sp?) & Penumbra – intent based transaction framework

Mike: Cosmos has very different opinions, diversity of ideas
Hasu: Big drawback is everyone has different validator sets, but as shared security grows, what compromises will be made?

How does regulation bump into MEV?
Crypto is about fair and equitable markets for users with less manipulation and exploitation
Execution on public blockchains is continually improving
Regulators are largely pragmatic

“If a single regulatory regime can make rules in crypto, then crypto has just failed”

Nat Friedman – Some things he believes: “Energy is a necessary input for progress”

Lovely list of perspectives and insights he’s gained over the years, at nat.org

A few of my favorites (copied verbatim):

—

The efficient market hypothesis is a lie
-In many cases it’s more accurate to model the world as 500 people than 8 billion

We are often not even asking the right questions

Where do you get your dopamine?
-The answer is predictive of your behavior

Going fast makes you focus on what’s important; there’s no time for bullshit

Enthusiasm matters!
-Energy is a necessary input for progress

—

Added to my personal bible

Two new Degens episodes with Imran from Kyber, and George and I shootin’ the crypto breeze

Lucky to have Imran Mohamad, Kyber’s marketing lead, to discuss in his words:

And in a separate episode, Jorge and I catch up on all the latest crypto shenanies and macro hankies and bank bailout pankies. Yup.

Podcast notes: Sam Altman (OpenAI CEO) on Lex Fridman – “Consciousness…something very strange is going on”

// everything is paraphrased from Sam’s perspective unless otherwise noted

Base model is useful, but adding RLHF – take human feedback (eg, of two outputs, which is better) – works remarkably well with remarkably little data to make model more useful

Pre training dataset – lots of open source DBs, partnerships – a lot of work is building great dataset

“We should be in awe that we got to this level” (re GPT 4)

Eval = how to measure a model after you’ve trained it

Compressing all of the web into an organized box of human knowledge

“I suspect too much processing power is using model as database” (versus as a reasoning engine)

Every time we put out new model – outside world teaches us a lot – shape technology with us

ChatGPT bias – “not something I felt proud of”
Answer will be to give users more personalized, granular control

Hope these models bring more nuance to world

Important for progress on alignment to increase faster than progress on capabilities

GPT4 = most capable and most aligned model they’ve done
RLHF is important component of alignment
Better alignment > better capabilities and vice-versa

Tuned GPT4 to follow system message (prompt) closely
There are people who spend 12 hours/day, treat it like debugging software, get a feel for model, how prompts work together

Dialogue and iterating with AI / computer as a partner tool – that’s a really big deal

Dream scenario: have a US constitutional convention for AI, agree on rules and system, democratic process, builders have this baked in, each country and user can set own rules / boundaries

Doesn’t like being scolded by a computer — “has a visceral response”

At OpenAI, we’re good at finding lots of small wins, the detail and care applied — the multiplicative impact is large

People getting caught up in parameter count race, similar to gigahertz processor race
OpenAI focuses on just doing whatever works (eg, their focus on scaling LLMs)

We need to expand on GPT paradigm to discover novel new science

If we don’t build AGI but make humans super great — still a huge win

Most programmers think GPT is amazing, makes them 10x more productive

AI can deliver extraordinary increase in quality of life
People want status, drama, people want to create, AI won’t eliminate that

Eliezer Yudkowsky’s AI criticisms – wrote a good blog post on AI alignment, despite much of writing being hard to understand / having logical flaws

Need a tight feedback loop – continue to learn from what we learn

Surprised a bit by ChatGPT reception – thought it would be, eg, 10th fastest growing software product, not 1st
Knew GPT4 would be good – remarkable that we’re even debating whether it’s AGI or not

Re: AI takeoff, believes in slow takeoff, short timelines

Lex: believes GPT4 can fake consciousness

Ilya S said if you trained a model that had no data or training examples whatsoever related to consciousness, yet it could immediately understand when a user described what consciousness felt like

Lex on Ex Machina: consciousness is when you smile for no audience, experience for its own sake

Consciousness…something very strange is going on

// Stopped taking notes ~halfway

Some excerpts from the excellent interstellar scifi Children of Men

Absolutely brilliant book and the first one I’ve read in many moons that I would prefer over Netflix as part of my evening wind down.

Amazon link: https://www.amazon.com/Children-Time-Adrian-Tchaikovsky-ebook/dp/B07DN8BQMD

A few highlights below to give a flavor of the writing and story:

—

For we are gods, and we are lonely, so we shall create …

Conversely, the Spitters are well aware of the danger that Portia’s kind poses. The two species have clashed over untold generations, each time with more understanding of the enemy. Now both recognize that the other is something less than kin but something more than prey.

He saw Guyen, looking more alert than anyone else there, and guessed the mission commander had ordered himself to be woken early, so that he could assert his bright, brisk dominance over this room full of zombies.

For so long, scholars had taught that the further the ice receded, the better for the world, and yet nobody had guessed what poisons and sicknesses had been caught up in that ice, like insects in amber, the encroaching cold protecting the shivering biosphere from the last excesses of Empire.

Fabian explains that there were once several warring colonies here, but one has become dominant. Instead of driving its lesser neighbours to extinction, the ruling colony has incorporated them into its own survival strategy, permitting their continuance in return for making them into extensions of itself, utilizing food that they gather and technologies they have developed. It is this world’s first superstate.

It is, however, quite true that packs of females—especially younger ones, perhaps newly formed peer groups seeking to strengthen their bonds—will descend to the lower reaches of the city and engage in hunting males. The practice is covertly overlooked—girls will be girls, after all—but overtly frowned upon.

She is also a recluse, wanting little more than to get on with her experiments—a common trait for those driven to build on their inherited Understandings—

Of course the Messenger is waiting for their reply. This was heresy such a short time ago, but Portia has since looked within herself. Why should we be made thus, to improve and improve, unless it is to aspire?

The male creeps out, to be pinned by their collective gaze. Fabian has given this moment some thought, based on his earlier failure with Portia. He will not ask for too much. He will show, rather than tell. He will woo them, but as a female does, with success, rather than as a male, with flattery.

“Traitors,” Guyen repeated, as if savouring the word. “In the end, they got what they deserved.” The transition from earnest, martyred leader to raving psychopath had simply happened without any discernible boundary being crossed.

It was too much. It had been too much. He, who had translated the madness of a millennia-old guardian angel. He who had been abducted. He who had seen an alien world crawling with earthly horrors. He had feared. He had loved. He had met a man who wanted to be God. He had seen death.

“Fuck,” she said expressively, and then repeated it a few more times, as if taking strength from the word.

The giants must live their lives amongst these rigid, unvarying right angles, entombed between these massive, solid walls. Nothing makes any attempt to mimic nature. Instead, everything is held in the iron hand of that dominating alien aesthetic.