Tejas Subramaniam🔸 @TejasReal

AI strategy researcher and Emergent Ventures fellow, previously @Stanford. I tweet about economics, futurism, philosophy, and animal welfare. Views my own. tejassubramaniam.github.io Stanford Joined March 2017

Tweets

757
Followers

841
Following

466
Likes

4K

Tejas Subramaniam🔸 @TejasReal

2 days ago

@NathanpmYoung Sorry, I meant to write women following and men leading.

0 0 1 36 0

View Details

Tejas Subramaniam🔸 @TejasReal

2 days ago

@NathanpmYoung I’ve never partaken, but a friend has, and she says that women leading and men following is a Schelling point, and a large fraction of the preference it captures is wanting to dance with the opposite sex (and this is also why she defaults to being a follow). Thoughts?

1 0 0 77 0

View Details

Tejas Subramaniam🔸 @TejasReal

a week ago

@PradyuPrasad Jevons Paradox occurs when demand elasticity is greater than 1! (Whether this is a case of Jevons Paradox depends on whether they were discounted due to an increase in input efficiency, and if that caused input usage to increase.)

0 0 2 42 0

View Details

Tejas Subramaniam🔸 @TejasReal

a week ago

Having looked into this more, I feel more confident that “academic fieldbuilding” is not the most effective path forward in improving our understanding of wild animal suffering and ways to reduce it (for pretty much the reasons in the thread below).

Tejas Subramaniam🔸 @TejasReal

3 months ago

I’m pretty uninformed about this, but as someone who thinks wild animal suffering is plausibly among the most important issues in the world (and wishes to make donations to such work over the next few years), I have some worries about the agenda to build it as a research field 1/

3 1 13 2K 0

0 0 2 686 2

View Details

Tejas Subramaniam🔸 @TejasReal

a week ago

@peterwildeford Not necessarily only (terminally) reward! lesswrong.com/posts/pdaGN6pQ…

0 0 2 71 0

View Details

Tejas Subramaniam🔸 @TejasReal

2 weeks ago

@fixedfrontier @apokolokyntosis The argument goes through as long as the perpetrators of the injustice are not the majority of those paying the reparations directly (e.g., their descendants are paying on their behalf). Most reparations claims consider effects on current victims.

0 0 0 24 0

View Details

Tejas Subramaniam🔸 @TejasReal

2 weeks ago

@fixedfrontier @apokolokyntosis I don’t know about how to do the accounting for remuneration here, but I think the view that reparations are owed for historical injustice is generally implausible due to non-identity issues (jstor.org/stable/20014043).

1 0 0 33 0

View Details

Tejas Subramaniam🔸 @TejasReal

2 weeks ago

I’m not confident, but it seems plausibly useful (especially if you want to start seeing progress before transformative AI or an intelligence explosion) to do targeted work to make AIs better at basic research on wild animal welfare (e.g., designing evals, generating data).

Tejas Subramaniam🔸 @TejasReal

2 weeks ago

To be sure, I expect AI to also have substantial effects on the quality and tractability of basic research on wild animals. I think it’s plausibly worthwhile to make sure this is true (cf. the current work to make AI models better at conceptual reasoning, such as DTBench).

1 0 3 1K 0

0 1 9 1K 3

View Details

Tejas Subramaniam🔸 @TejasReal

2 weeks ago

By “basic research,” I mean into questions like invertebrate consciousness (and the consciousness of larvae), the population effects of human actions on small animals, what the experiences of wild animals are typically like, and so on.

0 0 3 66 0

View Details

Tejas Subramaniam🔸 @TejasReal

2 weeks ago

Third, I think smaller-scale, tractable interventions would also have substantial second-order effects (such as on the population of soil animals), which could easily dominate their first-order effects. Understanding them seems important.

1 0 2 88 0

View Details

Tejas Subramaniam🔸 @TejasReal

2 weeks ago

I think basic research that helps resolve some of the deep cluelessness about wild animal suffering is more valuable than research that aims to find concrete, low-risk interventions in the near term to help wild animals, for three reasons. (Confidence: ~60%)

1 1 15 1K 2

View Details

Tejas Subramaniam🔸 @TejasReal

3 weeks ago

@BasilHalperin How would you describe why your treatment of decision theory differs from son-of-CDT (lesswrong.com/w/son-of-cdt)?

0 0 0 18 0

View Details

Tejas Subramaniam🔸 @TejasReal

a month ago

The history of technology transfer never ceases to amaze (cc: @karthiktadepall)

0 0 16 2K 5

View Details

Tejas Subramaniam🔸 @TejasReal

a month ago

@dwarkesh_sp Think this is a misunderstanding (he thought you meant total revenue).

0 0 3 489 0

View Details

Tejas Subramaniam🔸 @TejasReal

a month ago

@karthiktadepall (Also, the ways in which it’s a bit worse for economics are less substantive and more “a lot of the articles are stubs/just a few paragraphs long.”)

0 0 0 30 0

View Details

Tejas Subramaniam🔸 @TejasReal

a month ago

@karthiktadepall I knew someone would say that, but I don’t think this is it (and why I specified “other subjects I think about”)!

1 0 1 50 0

View Details

Tejas Subramaniam🔸 @TejasReal

a month ago

Wikipedia remains quite underrated. A lot of my knowledge of the world comes from studying Wikipedia articles carefully (although, sadly, I think it does a bit worse for economics than for other subjects I think about). In the next tweet, I’ll link some of my recent favorites.