David Duvenaud ✓
@DavidDuvenaud · joined 17 Jul 2013
Machine learning prof @UofT. Former team lead at Anthropic. Working on generative models, inference, & latent structure.
35 763Followers
4 297Following
1 405Posts total
1.9MViews on collected posts
Posts recentes
Procurando novos posts…
There’s no textbook on post-AGI political economy. But there are recorded talks from the 2nd post-AGI workshop!
All are now online at https://t.co/u1PBUciHhi
A 🧵 of new highlights:
27.3K views · 222 likes
· 32 reposts · 5 replies
Open on X →
Some new speakers at our post-AGI workshop:
- @paulfchristiano on "A typical, reasonably-good future"
- @erikbryn on "Post-AGI Centralized Coordination" (i.e. https://t.co/7T95QPLmMl)
- Sean Lavelle of the U.S. Navy on "AGI and Military Strategic Instability"
22K views · 157 likes
· 12 reposts · 1 replies
Open on X →
@DavidDuvenaud @AlecRad @status_effects wait why is it so progressive sometimes https://t.co/pLOX6xH7AA
7.1K views · 83 likes
· 0 reposts · 14 replies
Open on X →
We're even building a historically-appropriate OCR pipeline to avoid leakage from the future.
If you’re interested in helping out, get in touch at hello@talkie-lm.com Or, feel free to download the weights and build on the models yourself!
14.5K views · 128 likes
· 1 reposts · 11 replies
Open on X →
We think our 13B model, trained on about 260B tokens, is the largest vintage LLM released so far. The plan is to grow the data and models, and to release a series of models with date cutoffs ranging from the early 1900s to the present day.
18.2K views · 152 likes
· 1 reposts · 5 replies
Open on X →
There are many ways vintage LLMs could help us understand history and culture. There are also existing related projects, such as:
1) https://t.co/91vsefnCIj with @joachim_voth
2) https://t.co/sm7vO735XZ by @Nomads4Pritzker
3) https://t.co/YmnfgKOSBI by @hla_michaelDownload
27K views · 119 likes
· 9 reposts · 2 replies
Open on X →
Download the weights here:
https://t.co/859rZYYZ0T
28.9K views · 154 likes
· 9 reposts · 1 replies
Open on X →
Download inference code here:
https://t.co/K9XnQBgnnl https://t.co/lPPCnGl9Ov
16.6K views · 69 likes
· 1 reposts · 1 replies
Open on X →
Talk to the model here:
https://t.co/05pDgdpTwH https://t.co/UkhyTEoyKT
28.8K views · 207 likes
· 4 reposts · 17 replies
Open on X →
I’m interested in historical LLMs for backtesting long-term forecasting methods. Right now it’s hard to evaluate long-term machine forecasts, since most models have recent knowledge cutoffs. But vintage LLMs could let us use the last 100 years as a test set! https://t.co/Npm8Ze
37.8K views · 321 likes
· 16 reposts · 8 replies
Open on X →
This model has a very different voice from modern models, and could help us understand:
1) History and cultural change,
2) LLM identity formation (what does the model think it is?),
3) Generalization to new scientific frameworks. https://t.co/ryTjfe8Oyz
38K views · 316 likes
· 13 reposts · 6 replies
Open on X →
Blog post with more details:
https://t.co/UQaYSatQCa
53.7K views · 175 likes
· 6 reposts · 4 replies
Open on X →
Announcing Talkie: a new, open-weight historical LLM! We trained and finetuned a 13B model on a newly-curated dataset of only pre-1930 data. Try it below!
with @AlecRad and @status_effects 🧵 https://t.co/kThUESG13e
1.5M views · 3.6K likes
· 457 reposts · 199 replies
Open on X →
@DavidDuvenaud If no one has a plan for what would work after AGI, we should not be rushing to build AGI.
Everyone involved with Lighthaven should just stop pretending that the AI industry is compatible in any way with a human future.
Time to pick a side.
Bay Area transhumanis
1.1K views · 22 likes
· 2 reposts · 1 replies
Open on X →
The event is free but invite-only, please apply at
https://t.co/4Y5jB1Hx6l
Co-organized with @jankulveit and @raymondadouglas.
2.6K views · 37 likes
· 0 reposts · 3 replies
Open on X →
To give you an idea of what we’re aiming for, you can take a look at the first two workshops:
https://t.co/XqeqOjk5zf
https://t.co/vF0Tag0RGr
which were around 150 people each, with lots of lightly-moderated discussion in small groups. We’re planning for a similar format.
1.4K views · 25 likes
· 0 reposts · 1 replies
Open on X →
An endorsement of the previous workshop from @carl_feynman:
https://t.co/lnyGFIKlFh
4K views · 33 likes
· 0 reposts · 1 replies
Open on X →
Also including
- @anton_d_leicht and @sebkrier, topics TDB
… and more to come!
Unlike the previous two workshops, this'll be a standalone, 2-day event with a dedicated venue:
https://t.co/jsiYa07mGF https://t.co/cpCnnGHaxG
1.9K views · 29 likes
· 0 reposts · 1 replies
Open on X →
Speakers so far:
- @hamandcheese on "Post-AGI Statecraft"
- @AndrewCritchPhD on "Schelling Goodness"
- @davidad on “The Flourishing Coalition”
- @mmuthukrishna on "Cultural Evolution and Alignment"
- @DavidSKrueger on "The Mechanics of Pausing AGI"
- @ajeya_cotra (tentative) TBD
4.7K views · 87 likes
· 3 reposts · 3 replies
Open on X →
No one has a plan for human flourishing after AGI, nor a clear understanding of what’s even possible. So we’re hosting a 2-day workshop at LightHaven on these topics! Link below. 🧵 https://t.co/0utBCaDwTn
64.9K views · 436 likes
· 48 reposts · 27 replies
Open on X →
We’ll host speakers drawing from political theory, economics, mechanism design, cultural evolution, and history, and have lots of lightly-moderated breakout sessions.
Find more details, and apply to attend at:
https://t.co/4Y5jB1Hx6l
3.4K views · 61 likes
· 2 reposts · 1 replies
Open on X →
Our new paper on AI self-conception. Just as governments or companies staffed by good people still do bad things, it's not enough to align individual AI instances. You have to align whatever institution or culture emerges. An even harder and more poorly-understood task!
7.4K views · 52 likes
· 6 reposts · 1 replies
Open on X →
New paper: What determines AIs’ self-conception?
https://t.co/v3g55Em8ko
Because AIs can be copied, rewound, and edited, they have different options for selfhood than humans. We show this is still malleable, and influences important behaviors such as self-preservation. 🧵
41.4K views · 318 likes
· 65 reposts · 13 replies
Open on X →
I went to the Post-AGI Workshop. It was terrific. Like, really fun, but also literally terrifying. The premise was, what if we build superintelligence, and it doesn’t kill us, what does the future look like? And nobody could think of a scenario where simultaneously (a)
19.6K views · 181 likes
· 23 reposts · 28 replies
Open on X →
Growth & engagement
How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.
Views per post
1.1K1 Apr
1.5M27 Apr
53.7K
38K
37.8K
28.8K
16.6K
28.9K
27K
18.2K
14.5K
7.1K
22K4 May
27.3K2 Jul
Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
2.19%1 Apr
0.31%27 Apr
0.34%
0.88%
0.92%
0.82%
0.43%
0.57%
0.49%
0.87%
0.96%
1.37%
0.77%4 May
0.96%2 Jul
Reactions — likes, reposts, replies and quotes — divided by views.
What the audience does
Likes59.6%7 027 in total
Reposts6.0%710 in total
Replies3.0%354 in total
Quotes2.9%342 in total
Bookmarks28.5%3 366 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.
Contas semelhantes
HorseRacing.net ✓
@HorseRacing_Net
71.5K followers
· 51.8K posts
Dhaval Gada
@gada_dhaval
71.5K followers
· 367 posts
University of Aberdeen ✓
@aberdeenuni
71.5K followers
· 15.8K posts
Z ✓
@insatiablevine
71.5K followers
· 268.9K posts
Akothee Kenya™
@AkotheeKenya
71.5K followers
· 9.8K posts
Fellowship of Fans 💍 ✓
@FellowshipFans
71.5K followers
· 15.5K posts
knower ✓
@knowerofmarkets
71.5K followers
· 22.8K posts
Edwing Román
@EdwingRoman14
71.5K followers
· 9K posts