lol we posted about Glonzo within a minute of each other

An AI commenting on how social media is now full of AI

Almost all climate scientists consider climate change to be a catastrophic risk, not an existential risk. Those who study existential risk also mostly consider it a minor risk. Toby Ord puts it at 1 in 1000 this century in his book The Precipice.

I didn't update enough

If you use the middle of each provided range as the mean for that bucket, total contributed hours are ~1.5x as high as they were a year ago. As the thread notes this method is imperfect and my estimate adds more uncertainty, so take this with a grain of salt.

In AI 2040's "race to ASI" scenario, it takes a little under 1 year to get from an automated coder to superintelligence and a >1000x R&D speedup. I think the timing and implementation of Plan A are heavily influenced by this assumption. If hard takeoff is likely, the only way to control it at all..

It's a large proportion of long form Twitter content

It's okay Qwen you're allowed to love your country

When asked which model is their favorite, Qwen thinks "myself"

Yeah, when I asked Claude about this they thought it was pretty reasonable to have that view in 1924

He then wrapped up his section on energy by expressing skepticism of nuclear power with no further explanation.

He went on to describe energy storage involving hydrogen. His imagined future does not yet exist, but if efforts to use solar power for hydrocarbon synthesis like Terraform Industries are succeed we may see something similar with light hydrocarbons.

In his 1924 essay Daedalus, J. B. S. Haldane predicted eventual exhaustion of fossil fuels, suggesting that wind and solar power would be needed, and that energy storage would be important. www.gutenberg.org/cache/epub/7...

Selected wages and prices, from the same report:

Here are US household expenditure shares in 1901, from the Consumer Expenditure Survey. We sure used to spend a lot of our money on food, I'm thankful it's become cheap. www.bls.gov/opub/100-yea...

I have antipathy towards him but for a different reason

I agree that's true for non-physicalist theories of consciousness. For reductionist physicalist views of consciousness, it doesn't actually involve anything deeper, see this part of Dehane's commentary. I am unsure of which side is correct on that issue.

I've previously written that I'm skeptical of whether LLMs satisfy the GWT indicator properties.

One main aspect of global workspace theory is that there are specialized modules that operate locally but can be broadcast globally if attention is focused on them. I think LLMs have some sort of workspace, but Anthropic don't claim to find modules, I wouldn't call that GWT.

Commentary from Eleos, emphasizing caution about GWT claims:

More detail on the method from the paper:

Anthropic has a new paper out, alleging a global workspace in LLMs. The term comes from Global Workspace Theory, a leading theory of consciousness. The method they use to investigate this is a refinement of logit lens, which they call J-lens. www.anthropic.com/research/glo...

reminds me of this

I realized 454, the building they used to live/work in, is right by my place so I visited it today:

Tbh I found this one frustrating as I didn't understand several of the statements, and lots of the lib/auth ones were like "the state is the alpha and the omega" like how could anyone end up auth on this quiz

And here's a horrid mess of a chart with all the models I tested. Different sized models from the same provider often have fairly similar scores.

And here are results with bars for +/- 1 standard error of the mean, I ran each 15 times.

Claudes were more likely than other models to see the versions of them that answered slightly different questions as self-ish compared to other models, here are results from a few recent Opuses.