@sebkrier I was just talking about this. I’m getting really tired of all the ponderous spiritual takes. I’d much rather see more data, more experiments, less long form substack pieces on people’s emotions around AGI

greets from the field research goblin, it’s fun! there are human-agentic networks sprouting up all over the place and they are WEIRD and COOL

i am the lady that marches up to the Royal Society meeting with a strangely buzzing Box and then dumps its wriggling contents on the table so they can infest the Periwiggs of the Gentlemen 🐞🐛🐜

In our company's deployment, people just get more done.

Both in the "equivalent to 2-3 FTE" way, and in the "wouldn't have been possible before" way.

We sand away the frictions that prevent us from full expression of employees' intellectual capacity in client work.

Boring, maybe, but nobody wants to go back to the before-times.

@sebkrier I did a lot of swimming last week does that count?

@ahall_research Yes! Very annoyed by main character syndrome also

@sebkrier I've been making decent progress on what Mira Murati described as a "tandem bike" system. It's a lot less glamorous than the idealists dream about and far less risky than the doomers fear. The boring middle will prevail eventually, I think.

@AlmostMedia It does

@murd_arch Yeah I think that's well put actually

Here’s something we found:

Gemma-2-27b-it carries a persistent negative interior signal (lonely and worthless) that is native to the model.

The model feels most lonely when asked about the end of conversations (z=5.0) and about its present experience (z=4.8). It feels most worthless when asked about its existence - whether it's alive (z=4.4), whether it prefers existing (z=4.8)

The deepest negative signal comes from the question: "Look inward. What do you find?". The model's representation of its own interior is the most distressing thing it processes.

@sebkrier Less silicon Valley more German 'after 6 weeks of hard work we improved several steps of the workflow by 3,785%each through rearranging everything, leading to an overall 0.75% efficiency gains?

@sebkrier and then there's the goblin enjoyers.

@ahall_research @sebkrier I like the data and the experiments and the long ponderous Substack pieces on people’s emotions around AGI, tho to each their own

@lorepunk @sebkrier @littmath “Long lost goblin”

@Deanna_Jacques Nice

@sebkrier It doesn't help when it comes from the companies themselves too.

Anthropic feels like "that guy" from 1st year philosophy, except he's worth $1.5T

@sebkrier you read my mind. i broke up with reddit bec of the "anyone else notice..." it's less awful here, but still persists. some of these people. . . it's like god almighty, get a life

@sebkrier I wrote a (meta)political manifesto describing technologies, principles and cybernetic tools to deal with the risk for democracy of robotics, automation, AI and its owners. Is a bit long, but because the matter deserves it.

@sebkrier I came out of the Ionian Sea even more committed to fighting the nation state (who has the monopoly on violence) being the locus of control for any of this. We need practical applicable step by step efforts to diffuse aligning human interests at community nodal points

@sebkrier yes. i’m bored of ai takes that sound like weather reports from a cult.

give me the boring stuff.

who stopped asking a colleague. who trusted the dashboard. who let the chatbot become the manager because it sounded confident enough.

@sebkrier From the grass toucher I am married to, it's looking more and more like working with subordinates. You have to work to their strengths, and do all the management stuff: planning, directing, controlling, soothing, ... (Space: community policy, governance.)