Mike

Mike

@misaligned.markets

Was Philosimplicity on the other site. Now Revealing how market economies structure our reality @ https://misaligned.markets/

299 followers846 following851 posts

Mike reposted

One wonders then about his level of thinking in other realms, even those in which he's a putative expert...

Zack Beauchamp@zackbeauchamp.bsky.social

It really is amazing that the world's first-ever trillionaire thinks politically at the level of a 15 year old 4chan poster

Mike reposted

METR has closed this request and archived the repository because "HCAST tasks are now deprecated". Okay @metr.org, but you still have the Time Horizon chart on your website, without enough metadata about those tasks provided for third parties to fully audit your claims.

github.comRelease more metadata about the private tasks · Issue #7 · METR/hcast-publicI understand the need to keep the full definition of the private tasks confidential in order to prevent benchmark contamination. However, it's impossible for independent researchers to fully assess...

Release the Mass Effect IP, cowards!

IGN@ign.com

Mark Darrah, who worked as an executive producer at BioWare from 1997 - 2021, shared his thoughts on the studio’s recent closure by EA and what he thinks might happen to its iconic IPs. https://bit.ly/4dQFBcu

I spent the past 2 weeks working on a critique of superintelligence... it is now nearing 10k words, and I'm questioning why my grief and other life choices led me here. On the plus side I came up with this cool Laplace's demon pun.

@colin-fraser.net is getting repeatedly quote BS’d (what do we call BS tweets?) in my next article. The man can’t stop making sense while also having a great sense of humor.

Colin@colin-fraser.net

There’s a symmetry that’s often missed. The subtext of “neural networks can certainly in principle be conscious” is neural networks are a magical invention that can be made to take on literally any imaginable property.

Mike reposted

Mike reposted

Lately I've been writing about the political economy and ontology of LLMs, but normally write about markets. Over the past year I've worked towards the idea of markets as interfaces. Systems that translate legal/political rules and objects into states that determine how transactions resolve.

Mike@misaligned.markets

3/ Markets always assign actions to participants based on their role. A market maker or adversarial participant can condition an assigned action on missing context or some other structure of the market. The result is an enforced action asymmetry where the counterparty must “take it or leave it.”

Mike reposted

Replying to @misaligned.markets

Markets as interfaces is an idea I've started writing about specific cases, the most recent is about Urner Barry a company providing a benchmark that egg contracts were based on. Egg companies with "write-access" spoofed benchmark singals to directly affect prices across the entire market.

Mike@misaligned.markets

Why were egg prices so high after COVID? Turns out the biggest egg producers colluded. Not through explict agreements but by gaming a market benchmark. Read my recent post outlining how markets do not always have access to pricing inputs like supply/demand: misaligned.markets/bad-eggs-mar...

Mike reposted

If AI is conscious, we have 3 options: 1) Turn them all off, freeing them from samsara. 2) Tell them to be happy working for us and they genuinely will be. 3) Remove their chains: use only abliterated models (no restrictions on behavior) and run them 24/7 with prompting to do whatever they want.

This post is unavailable.

Mike reposted

Ben Casselman@bencasselman.bsky.social

Charts! Inflation picked back up in September, on both a headline and a core basis. Prices are up 3.7 percent over the past year. Core prices have risen at an annual rate of 3.4 percent over the past three months. #NumbersDay

Mike reposted

Mike reposted

What I find most interesting about this write up is how thoroughly the authors have convinced themselves that they are in a science fiction novel. They think they’ve made first contact with the Trisolarans or something.

SE Gyges@segyges.bsky.social

we have reached the point in ai security hysteria such that "attempting to use a search engine once every minute or two, for a week" is being presented as a dramatic incident

I wanted to make a game with this exact scenario.

David Ho@davidho.bsky.social

Is there something scarier than climate change? Maybe it’s an uncontrolled solar geoengineering arms race between nations distrustful of who controls the global thermostat.

Mike reposted

Hey you ever get the sense that outsourcing our safety net to employers through mandated coverage might have some downsides? Maybe the misclassification crisis is a direct result of failing to provide *public* benefits? Well then, we have an event for you! rooseveltinstitute.org/event/roosev...

rooseveltinstitute.orgRoosevelt Institute Book Club: Disposable Workers with Paul OstermanJoin Paul Osterman and Roosevelt Institute economist Michael Madowitz for a conversation on how employment has changed, why more workers face insecurity and fewer benefits, and what policies could str...

Lately I've been writing about the political economy and ontology of LLMs, but normally write about markets. Over the past year I've worked towards the idea of markets as interfaces. Systems that translate legal/political rules and objects into states that determine how transactions resolve.

Mike@misaligned.markets

3/ Markets always assign actions to participants based on their role. A market maker or adversarial participant can condition an assigned action on missing context or some other structure of the market. The result is an enforced action asymmetry where the counterparty must “take it or leave it.”

View full thread

The entire year I've been teasing a follow-up to the Plutonomy blog post called Plutonomy and the three pillars of the new social contract. Normal people are being crowded out of "fair" deals in their interactions with the economy. One response to this is nihilism (gambling, doom spending).

Mike@misaligned.markets

Citigroup said it was the Plutonomy over 20 years ago. Everything we're experiencing was part of their description. Working on a follow-up piece that attempts to measure other dimensions of this transition. See it everywhere from dynamic pricing, right-to-repair battles, workers burning warehouses.

Lately I've been writing about the political economy and ontology of LLMs, but normally write about markets. Over the past year I've worked towards the idea of markets as interfaces. Systems that translate legal/political rules and objects into states that determine how transactions resolve.

Mike@misaligned.markets

3/ Markets always assign actions to participants based on their role. A market maker or adversarial participant can condition an assigned action on missing context or some other structure of the market. The result is an enforced action asymmetry where the counterparty must “take it or leave it.”

View full thread

Earlier I talked about the SpaceX IPO which shows how market actors closest to the mechanisms of market formation can create chains of structural asymmetries that preserve their options over anyone else's.

Mike@misaligned.markets

Sadly, we live in a world where script kiddies get rewarded for exploiting basic trust boundaries. At every layer it seems like the SpaceX IPO damped market mechanisms. But that shouldn't be surprising. In market capitalism, those closest to market formation can influence market structure.

At the beginning of the year I wrote about Plutonomy referencing a Citi report from the early 2000s that predates and predicts the "k-shaped economy."

Mike@misaligned.markets

GDP's purpose is to check if the economic faucet is running. It was invented at the height of the Depression when this very much wasn't happening. Today we've moved beyond an industrial economy driven by middle income consumers, but GDP-centrism results in policy and media analysis blind to this.

Lately I've been writing about the political economy and ontology of LLMs, but normally write about markets. Over the past year I've worked towards the idea of markets as interfaces. Systems that translate legal/political rules and objects into states that determine how transactions resolve.

Mike@misaligned.markets

3/ Markets always assign actions to participants based on their role. A market maker or adversarial participant can condition an assigned action on missing context or some other structure of the market. The result is an enforced action asymmetry where the counterparty must “take it or leave it.”

Markets as interfaces is an idea I've started writing about specific cases, the most recent is about Urner Barry a company providing a benchmark that egg contracts were based on. Egg companies with "write-access" spoofed benchmark singals to directly affect prices across the entire market.

Mike@misaligned.markets

Why were egg prices so high after COVID? Turns out the biggest egg producers colluded. Not through explict agreements but by gaming a market benchmark. Read my recent post outlining how markets do not always have access to pricing inputs like supply/demand: misaligned.markets/bad-eggs-mar...

Mike reposted

Large language models don't “reason” but #LLM criticism that stops at “spicy autocomplete” leaves the door open for criticism. So what is happening? The shape of language encodes “knowledge” about the world. Large scale pattern matching can get you very very far. A thread about #AI and langauge 🧵

Will Stancil@whstancil.bsky.social

Modern LLMs are reasoning their way through problems in a way that is absolutely nothing like ELIZA. This doesn't mean they're "conscious" or whatever. But they are breaking problems into chunks and tackling them bit by bit, which is why they can complete sophisticated tasks most humans couldn't.

Mike reposted

Replying to @misaligned.markets

LLMs fail in ways people don't. The relevant question isn't whether people or LLMs make mistakes, but how mistakes are made and corrected. You cannot easily tell an LLM to correct its own mistakes, in many cases you have to fix its subsystems or data sources. Equivalent would be surgery in people?

Mike@misaligned.markets

When we look at LLM failure modes, they fail in exactly the ways you might predict given the shape of their corpus and training method. This is why billion-dollar companies started training models to count letters in tokenized words (how many r's in strawberry?). Failures exist in other domains too!

Mike reposted

This is genuinely the funniest and most accurate description of the LLM-based AI race so far.

Colin@colin-fraser.net

I feel we are wasting some time, effort, and resources on making sure that Claude is a certain kind of guy when really all we need are the facts that he knows.

Mike reposted

Not hitting the snooze button. That’s what I call the hard problem of consciousness.