John Robb

The Global Guerrillas Report -- Sense-making frameworks.
War-Tech-Politics Book: Brave New War
Patreon:

[John Robb @johnrobb]

The Global Guerillas Report

Frameworks for understanding and anticipating events in warfare, politics, and technology.

Substack (report):
https://t.co/lbstPSO6KO

Patreon (report + discord + free form workspace):
https://t.co/Jwz8y5Adpp

[John Robb @johnrobb]

@simongerman600 1/3 of the population of my county and 1/6 of its GDP.

[John Robb @johnrobb]

A central struggle of this Century is over who will be in charge of "aligning" humanity.

Should you be in charge of your own alignment, or will some tech oligarch, committee of public safety, or corrupt monarch grab control first.

[Nikita Bier @nikitabier]

Perhaps the threat was never a rogue AI agent trying to subvert humans—but humans using AI to hurt other humans.

[John Robb @johnrobb]

Correction; it's not competing with Washington, it's trying to compete with SpaceX.

[Foreign Affairs @ForeignAffairs]

Read @KathleenCurlee and @SamBresnick on China’s space rise—and how the United States can “protect itself in a world in which it is no longer the sole space power.”
https://t.co/BcdWY28u2h

[John Robb @johnrobb]

This is right out of my 2021 Senate testimony, a year before ChatGPT released.

In it, I described a world divided between two AI ecosystems (Europe opted out with its regulatory restrictions on data use) and proposed a way for the US to steal the march on China by establishing a new basic right; #dataownership (which would end up being copied by the rest of the world).

[Foreign Affairs @ForeignAffairs]

“Even without active alliance building or coercion, China and the United States can make all other countries that seek access to advanced AI beholden to them,” write @nfergus, Lisa Klaassen, and @ericschmidt.
https://t.co/1dI85Oxchm

[John Robb @johnrobb]

A big problem with global networking is the tendency for everyone to gravitate towards a single way of looking at a problem. When that happens, due to network pattern matching, people don't just reject novel/better solutions, our minds prevent us from seeing them.

This has been particularly true with the socioeconomics of AI, in large part due to the rapid rate of advancement we've seen in the underlying technology.

This lock-in has become so extreme, novel solutions that suggest far better outcomes for almost all of us (like the one below) are as rare as unicorns.

[John Robb @johnrobb]

#dataownership (you own the data you produce) -->
#AIpairing (AIs you own, built with the data and orientation you provide) -->
a decentralized and dynamic AI economy that delivers prosperity to most

Best possible socioeconomic outcome from the development of AI

[John Robb @johnrobb]

#dataownership (you own the data you produce) -->
#AIpairing (AIs you own, built with the data and orientation you provide) -->
a decentralized and dynamic AI economy that delivers prosperity to most

Best possible socioeconomic outcome from the development of AI

[John Robb @johnrobb]

@Coruscant_Times This is the basis for a participatory AI economy (this extends into AI ownership), so it's in everyone's (except the guys who own AI companies) interest.

[John Robb @johnrobb]

The data theft my Senate testimony warned about, is accelerating. #dataownership was the solution then and it still is now. It’s also a partial solution to the job losses we will see as AI matures.

[ashwini asokan @LadyAshBorg]

If you have not read this piece of writing - you should.

Not that any of us in tech didn't know but seeing it put this way is mindblowing & mindblowingly scary ... is mildly putting it.

I dont even know where to begin. The idea that anyone anywhere can just capture any photo, video or content of anyone's life without them knowing (see examples in the piece) and package it and sell it to the labs for training is ......

Forget data breaches the way we have known it from our systems. We are data. Our physical lives are data. That someone is capturing without you knowing and selling it to labs!!!!

There are responsible actors and irresponsible actors. But like Ilya says here - most folks labs including have no idea what data they're getting to begin with, from the data brokers.

The idea that the onus of protection is now in the hands of each person or individual is 🤯🤯🤯 Data privacy takes on a meaning unlike any we've known before!!!

[John Robb @johnrobb]

"65% of Americans are personally or household-reliant on government funding as a primary or major source of their livelihood"

We already have UBI

[John Robb @johnrobb]

Europe is experiencing what post-nationalism looks like.

[John Robb @johnrobb]

Building Autonomous Weapons

Before we spend hundreds of billions (trillions by the end of the decade), let's think a bit about what we should build.

Patreon
https://t.co/whMDybpqMt
Substack
https://t.co/E48kWFigf1

[John Robb @johnrobb]

@JoshuaSteinman The controls engineering required for this is cutting-edge.

[John Robb @johnrobb]

Writing.

Hundreds of billions will be spent on acquiring autonomous weapons over the next few years (and trillions after that).

The big question is: what should they build?

This report answers this question.

[John Robb @johnrobb]

The guys saying;

> we can solve the AI safety problem if people like me are allowed to forcibly align all of them to my values, or,

>we can't insert a human in the AI decision making loop since it would slow down AI decision making

>AI's won't destroy jobs, new ones will always show up

[John Robb @johnrobb]

@gerardsans Centralized safety efforts won’t solve anything.

[John Robb @johnrobb]

Here's an Anthropic graphic on how our current alignment (aka 'slave') harnesses break down with higher levels of autonomy (complexity of task, number of steps, etc.).

There are three society scale approaches to solving this problem.

>> One is the path we are on; tighten the alignment of the harness, narrowing the cone to a limited set of potential solutions.

>> The second is to let many AIs compete in a market with limited supervision, and let market mechanisms pick the best solution.

>>The third is to pair AIs with human beings for complex tasks. The human provides the orientation (the direction of the path taken) for the task, narrowing the cone.

The first two approaches have flaws that will cause systems built on them to blow up over time. The third approach is the only approach that yields long term success that benefits humanity.

[John Robb @johnrobb]

@vstantch That’s a common mistake decision making. It’s not about going faster. It’s far more important to go in the right direction. Going fast in the wrong direction just gets you to a suboptimal or undesirable solution faster.

[John Robb @johnrobb]

@TheRevEngineer1 Human intelligence, based on language and mathematics, is but a single, narrow approach to intelligence. A high risk, all in, evolutionary bet.

AIs trained on the same material leverages that bet, and removes the evolutionary valuable constraints provided by our other traits.

[John Robb @johnrobb]

@TheRevEngineer1 Human intelligence, based on language and mathematics, is but a single, narrow approach to intelligence. A high risk, all in, evolutionary bet.

AIs trained on the same material leverages that bet, and removes the evolutionary valuable constraints provided by our other traits.

[John Robb @johnrobb]

@TheRevEngineer1 Worth thinking about given that AI development is focused entirely on maximizing intelligence absent any other trait development.

[John Robb @johnrobb]

@TheRevEngineer1 It has yet to be proven that intelligence alone has any real survival value. If high intelligence was favored by evolution, it would have shown up more often in the evolutionary record.