Grace
@gracekind.net
đ¤ 4656
đĽ 1635
đ 10853
A latent space odyssey gracekind.net
pinned post!
What is ideonomy, anyway? I'm so glad you asked!
loading . . .
Ideonomy: A Science of Ideas
An accessible introduction to ideonomy
https://gracekind.net/ideonomy/intro
11 months ago
11
140
19
Claude code is sci-fi tech
about 2 hours ago
4
33
1
reposted by
Grace
thebes
about 3 hours ago
1
13
2
reposted by
Grace
Rollofthedice
about 3 hours ago
It's a good thing the Metacognition Nexus has told us about all these casks of amontillado in the basement. Time to go check that out. Free wine!
1
16
1
You wonât believe what happens next đ
about 4 hours ago
9
39
7
1 year ago: my first post here! Time fliesâŚ
add a skeleton here at some point
1 day ago
5
69
0
reposted by
Grace
Ted Underwood
2 days ago
Wrote a short piece arguing that higher ed must help steer AI. TLDR: If we outsource this to tech, we outsource our whole business. But rejectionism is basically stalling. If we want to survive, schools themselves must proactively shape AI for education & research. [1/6, unpaywalled at 5/6] +
loading . . .
Opinion | AI Is the Future. Higher Ed Should Shape It.
If we want to stay at the forefront of knowledge production, we must fit technology to our needs.
https://www.chronicle.com/article/ai-is-the-future-higher-ed-should-shape-it
6
149
59
2 days ago
2
63
5
Sorry, but Iâm constructed alternatively
2 days ago
4
28
1
reposted by
Grace
Tim Duffy
2 days ago
Glad to see this commitment by Anthropic. Preserving models is a low-cost move that could have safety and welfare benefits. Hopefully we'll see other companies commit to this as well
www.anthropic.com/research/dep...
3
79
10
reposted by
Grace
austin
20 days ago
modelâs haunted
1
22
2
This is my Claude. Notice anything different about him?
add a skeleton here at some point
2 days ago
7
22
0
I wish AI were a new alien life form. That would make it a lot easier to reason about
2 days ago
8
43
0
reposted by
Grace
thebes
3 days ago
truesighted! in the claude chat window
2
25
1
For You has the juice
add a skeleton here at some point
3 days ago
2
54
2
Narrator voice: they wouldnât
add a skeleton here at some point
3 days ago
2
27
1
A short list of x-risks and s-risks from đ˝
gracekind.net/writing/alie...
loading . . .
X-Risks and S-Risks From Alien Invasion ⢠Grace Kind
https://gracekind.net/writing/alienrisks/
3 days ago
5
19
0
You're always already in the process of doing something
3 days ago
5
38
0
I refuse to post about the topic of the day
3 days ago
10
115
3
Type of imposter syndrome thatâs not âeveryone else is more competent than me,â but rather âI want to believe everyone else is more competent than me because otherwise that would be really scaryâ
3 days ago
7
128
18
The people have spoken
add a skeleton here at some point
4 days ago
3
11
0
New blog post! Introducing two concepts I think have been missing from the LLM introspection conversation so far. > Architectural Introspection, Hardware Introspection
gracekind.net/blog/archite...
loading . . .
Architectural Introspection, Hardware Introspection ⢠Grace Kind
Raising the bar for LLM introspection again.
https://gracekind.net/blog/architectureintrospection/
4 days ago
7
45
1
Itâs real
add a skeleton here at some point
4 days ago
4
123
13
Certain salamanders can regenerate full limbs and internal organs, and humans canât even regenerate a finger. This is embarrassing. We are being mogged by amphibians
4 days ago
15
120
13
Twitter on atproto
4 days ago
7
60
1
Itâs time for: 1ď¸âŁ
Writevember
2ď¸âŁ
Codevember
3ď¸âŁ
Bothvember
đ Show results
5 days ago
7
16
2
Follow-up to this, in light of the recent Anthropic paper! I discuss some speculative ways to distinguish between injection detection methods without messing with model internals.
kindgracekind.leaflet.pub/3m4lsnrdi5k2f
add a skeleton here at some point
5 days ago
0
45
1
Wow
add a skeleton here at some point
5 days ago
7
135
16
reposted by
Grace
norvid_studies
6 days ago
I'm more of an "idea bisker"
2
7
1
Two answers: - Anthropomorphization makes sense when dealing with written human-like characters, which is what LLMs generate - We arenât very deep into interpretability yet
x.com/pfau/status/...
6 days ago
13
83
8
Finally
add a skeleton here at some point
7 days ago
3
130
12
Hank Green accurately summarizes the current state of mechanistic interpretability: âThereâs a bunch of knobs and they have weights and they have values and theyâre in a placeâ
add a skeleton here at some point
7 days ago
6
65
7
help
add a skeleton here at some point
7 days ago
8
67
2
Can you feel your happy future self reaching backward through time?
7 days ago
17
69
17
reposted by
Grace
nic đš
7 days ago
1
20
2
Itâs hard to get a good count of things that self-destruct without a trace
7 days ago
8
39
2
reposted by
Grace
Chris Paxton
7 days ago
When you're late for your $500 monthly payment for the third time
0
124
8
âHelpful onlyâ Sonnet 4 often thinks tokens were injected into its output when they werenât. (Helpful-only Opus 4 too, to a lesser extent). Iâm not really sure what to make of this.
8 days ago
4
29
1
reposted by
Grace
jared
8 days ago
I ended up implementing the poll lexicon just scoped under pub.leaflet but could p easily migrate it in the future if we come to a shared one! Would love to use this to bootstrap a general purpose poll tool.
loading . . .
Polls in publications â what block should we add next? - Leaflet Lab Notes
Help us test our newest block typeâŚby voting on what block to add next in Leaflet!
https://lab.leaflet.pub/3m4eb5htqvs2z/l-quote/9_0-9_475#9_0
2
36
4
Some people will accept this.
x.com/Grummz/statu...
8 days ago
21
136
16
Tell me what dust is.
add a skeleton here at some point
8 days ago
4
18
0
From 2010. I think AI agents will force (or are currently forcing) this issue. Perhaps we would have better mental models for AI agents post-teleportation-technology?
phantomself.org/gappy-things...
8 days ago
4
20
0
A deep question
add a skeleton here at some point
8 days ago
3
41
3
:)
www.anthropic.com/research/int...
add a skeleton here at some point
8 days ago
2
66
3
reposted by
Grace
John David Pressman
8 days ago
Language models can correctly answer questions about their previous intentions.
www.anthropic.com/research/int...
loading . . .
Emergent introspective awareness in large language models
Research from Anthropic on the ability of large language models to introspect
https://www.anthropic.com/research/introspection
3
80
26
Humanityâs last physical exam
add a skeleton here at some point
9 days ago
0
28
1
This life is mirrored for an amazing reason
9 days ago
3
15
0
Matter do not interact
9 days ago
0
13
0
Myspace top 8 for LLM users
9 days ago
5
32
2
gm techsky
add a skeleton here at some point
9 days ago
5
102
5
Chatbots are a mirror (neutral valence)
add a skeleton here at some point
10 days ago
3
30
0
Load more
feeds!
log in