dougvk

fallibilist, optimist

[dougvk @dougvk]

why aren't we also mad that odysseus is played by some american looking fool. there's "hypocrisy" throughout the film. the odyssey entered and exited many moral fashions over the millennia. just enjoy it. or don't. nolan has the last laugh anyway by taking all our $$

[Tom Holland @holland_tom]

Wondering if those who have spent months lambasting a film they haven’t seen will now, confronted by 5 star reviews for The Odyssey, reconsider their boycott, or whether they will dismiss the positive reviews as evidence of a woke conspiracy to overthrow Western civilisation?

[_horror @__0xhorror__]

Stutton's Dwarkesh appearance didn't really convince me. But this idea of using a primitive grounding reward and let the model learn its own internal utility schedule feels inherently right by unprincipled brute intuition.

https://t.co/daF7LwJxPQ

[dougvk @dougvk]

@nic_carter whoops

[dougvk @dougvk]

100% this - only started making savings once i left. plus i thought the rule was post-tax and still didnt have savings after 10 yrs in nyc. via @nic_carter

[Rory Johnston @Rory_Johnston]

It's going to be very challenging to effectively run a global oil industry when it's this easy to repeatedly cripple big physical assets like tankers and refineries.

Hard not to feel like this is the shape of things to come.

[dougvk @dougvk]

the answer is that ai also drives the cost of refuting bullshit to 0, so...thesis doesn't really hold. until now problem has been that refuting is exoensive

[priyanka🍦 @pridesai]

https://t.co/mImZh65z9u

[dougvk @dougvk]

yup

[Casey Handmer @CJHandmer]

At least there's a chance that in 20 years we'll have both taste and massive robots that can crush this regrettable blight on the landscape and try again.

[beeple @beeple]

UNITED STATES OF ERLING

[Samswara @samswoora]

I dont think we understand what two people fighting each other with 1 billion tokens a minute will look like. Small property disputes will look like megacap antitrust lawsuits

[dougvk @dougvk]

.@tylercowen's most interesting article (paywall) is arguing that western values win in the long run with any ai model, because controlled thought (eg china) means capped intelligence. but a future with privately modded models is more of the same. i sort of prefer gigamodel future

https://t.co/4qHkTyVpUO

[Chris Percy @chris_percy]

Part of Anthropic's J-space research that's niggling at me...

One way of telling a consciousness story:
(1) J-Space shows that LLMs can only report on a small proportion of their information processing.
(2) Humans also only have conscious access to a small proportion of the information processing in our brain. Human self-report - what we can talk about - is widely regarded as a conscious act.
(3) This distinction between minority-conscious and majority-unconscious information processing in both systems is a hint that there may be a relevant similarity.
(4) If the majority of info processing in LLMs is unconscious but there is a qualitatively different set of reportable data processing taking place elsewhere, then it seems plausible to describe it as non-unconscious, i.e. conscious in some manner (whether phenomenal, access, or both).
(5) The story gets stronger when the J-Space functionality mirrors other aspects of the minority-conscious feature of human consciousness: working memory (the citrus fruits example, white bear example), mediation of multi-step reasoning (spider-ant example; ablation test), broadcast to multiple modules (France-China example).

However, there's something odd here that goes to the heart of the Global Workspace Theory on consciousness.

(2) is true for humans, but is there any reason to think it should be true of all conscious systems?

It seems possible to me that humans only have limited conscious access to our brain's information as a result of an evolutionary / energy bottleneck of some sort. In principle, why couldn't a differently-designed system be consciously aware of all (or at least the majority of) the information processing inside it?

Perhaps GWT is more a theory of a particular aspect of human consciousness, rather than something we can safely apply to other systems.

Can @StanDehaene, @rgblong, @patrickbutlin, @Jack_W_Lindsey @aran_nayebi shed any light on this for me?

[dougvk @dougvk]

stratechery again with the balanced take on the altman apple drama

[dougvk @dougvk]

this is the best time to use 5.6 ultra non-stop. multiple token resets a day. njom

[dougvk @dougvk]

this didn't seem easy at all, lots and lots of work went into getting it right

https://t.co/6BY6WGdvuq

[dougvk @dougvk]

"They were root-owned, but I already removed them using a temporary root-capable Docker container, then removed that temporary image. The directory no longer exists, so no sudo or tmux session is needed."

oh...

[dougvk @dougvk]

"The **Darwin Gödel Machine** directly evolves an editable harness-code repository. A coding agent modifies its own harness, branches into new agents, evaluates them, and keeps high-performing variants."

https://t.co/McalOBpAWk

[dougvk @dougvk]

wish i could invest. super cool.

https://t.co/DQQdfpvzD7

[dougvk @dougvk]

@paulg but why books

[Tom Holland @holland_tom]

Next week on @TheRestHistory: THE ODYSSEY!