Benjamin Manning @BenSManning Twitter profile | Pikagi

Pikagi

Benjamin Manning

@BenSManning

743

Followers

578

Following

17

Media

272

Statuses

PhD student @MIT | AI, Econ, & Experiments

Cambridge, MA

https://t.co/CGdh0awKpK

Joined February 2022

Don't wanna be here? Send us removal request.

Pinned Tweet

@BenSManning

Benjamin Manning

2 months

fwiw, really appreciated the dialogue here, we are excited to discuss---both the positives and the critiques. There is so much unknown with these models and their potential to improve our methods for social inquiry, but we won't know whats good---or bad--- until we explore.

@emollick

Ethan Mollick

2 months

Yeah, gonna delete this. People are just insulting the paper authors without reading the paper or understanding the context of the fields being discussed. They didn’t ask that I tweet about them, and I didn’t expect the tweet to travel so widely. Getting nasty on Twitter.

Tweet media one

68

58

811

5

8

90

Last Seen Profiles

@hobi_wassi85769

@stw46

@Instinct_VA

@kronesport

@so_04_22

@DoingJack

@i_ameztoy

@HomerKnightWR

@BijedP

@wilksxxx

@SnipesMotion

@justcatchmedemi

@Purest_Fox

@itsRizha

@moewe_mowe

@Dbaic

@Clecienne

@ASBrS

@AbandonArchive

@AP_ZFN

@mckaywilson

@Harshita_125

@_dallas02_

@HailieDeegan

@WetNosesClub

@wutang36genders

@siema161

@priscila_d543

@luisficarpediem

@AlexSprackland

@GLOBALINFO20

@stwmaniax

@i_ameztoy

@friedvocalcords

@HHolarpizz2024

@AcademyScoop

@BenSManning

Benjamin Manning

1 month

First paper on NBER---hopefully many more to come! So thankful for brilliant collaborators like @johnjhorton and @Kehang_Zhu , I don't know how they function in society being so smart ;)

@nberpubs

NBER

1 month

Using large language models both as an experimental subject but also as a scientist, from @BenSManning , @Kehang_Zhu , and @johnjhorton

Tweet media one

3

32

119

2

5

65

@BenSManning

Benjamin Manning

2 months

let the record show that I am this grad student, and @johnjhorton has also asked me if I've seen memento at least 5x (maybe more, now I can't remember...) Great movie btw

@johnjhorton

John Horton

2 months

I told a grad-student co-author "try to imagine me as the guy from 'Memento' who can't remember anything & needs clues to pick-up the thread on projects" and they said "I believe you, because you already used that memento analogy"

23

878

21K

1

0

63

@BenSManning

Benjamin Manning

2 months

Automated Social Science: Language Models as Scientist and Subjects now on arxiv! With @Kehang_Zhu @johnjhorton

Tweet card media

Automated Social Science: Language Models as Scientist and Subjects

We present an approach for automatically generating and testing, in silico, social scientific hypotheses. This automation is made possible by recent advances in large language models (LLM), but...

2

7

45

@BenSManning

Benjamin Manning

3 months

Feedback welcome! Still working on the draft for submission :)

@johnjhorton

John Horton

3 months

New working paper by @BenSManning & @Kehang_Zhu that & I’m super excited about

Tweet media one

11

78

310

1

7

41

@BenSManning

Benjamin Manning

2 months

🙏🙏🙏

0

0

20

@BenSManning

Benjamin Manning

2 months

Danny, knowing you made us smile-look at Linnea. I'll never forget our zooms, your infuriatingly perfect combination of humility and brilliance, your kindness, and your ability to always understand our ideas before we knew what we wanted to say. It was a pleasure to know you.

Tweet media one

@CassSunstein

Cass Sunstein

2 months

A favorite quotation from Danny Kahneman: "Nothing in life is as important as you think it is when you are thinking about it." (Except Danny.)

7

350

2K

1

0

13

@BenSManning

Benjamin Manning

3 months

A VERY short blog by @tylercowen about my paper with @Kehang_Zhu and @johnjhorton , exciting!

@MargRev

Marginal Revolution

3 months

It's happening, economic science edition

0

4

16

0

0

13

@BenSManning

Benjamin Manning

3 months

Really intuitively helpful post by @deaneckles about how much information one can actually infer from statistically significant results

0

4

13

@BenSManning

Benjamin Manning

2 months

@mollycrockett also a great way to quickly explore these models in an interpretable way, although LLMs surely can/will be used to do invalid inference with respect towards humans, it doesn’t have to be. They offer us a tool to better understand ourselves-something we currently aren’t great at!

1

0

12

@BenSManning

Benjamin Manning

6 months

Had a great time presenting today at the interactive causal learning conference @FloridaAtlantic !

@johnjhorton

John Horton

6 months

. @BenSManning & @Kehang_Zhu talking about our new paper - we're very excited to make this one public soon (project has had a lot of 🤯-moments)

Tweet media one

1

5

34

1

1

12

@BenSManning

Benjamin Manning

22 days

Question to the econ twittersphere: what are people’s strategies when they have a lot of papers to read and a very limiting time restriction? Read a few in depth? Intros of all of them? Figures only? Ask Claude to summarize?

3

1

11

@BenSManning

Benjamin Manning

27 days

@yudapearl @yudapearl coming up with a hypothesis is easy! Just type into GPT “come up with an idea about XX causing YY” -it might be a bad idea, but it is one. Also this paper: then for agents also being possible: and

Tweet card media

Machine Learning as a Tool for Hypothesis Generation

2

2

10

@BenSManning

Benjamin Manning

5 months

Thanks for having me! Check out this jupyter notebook I made replicating the experiments super easily using The AWESOME new software package edsl that @johnjhorton has been building:

@MITFutureTech

MIT FutureTech

5 months

What Can We Learn from Homo Silicus? At a recent lab meeting, Benjamin Manning ( @BenSManning ) explained how we can potentially use LLMs to revolutionize how we conduct social science research. See the attached paper for more information:

0

0

6

0

0

9

@BenSManning

Benjamin Manning

1 month

Fun fact, if you run an automated experiment generated from nothing but "A family arguing whether to get thai food or indian food for dinner" using LLMs, it does matter how much family member 1 likes thai food. More importantly, @johnjhorton must take me out for thai ASAP 👀

Tweet media one

1

1

8

@BenSManning

Benjamin Manning

2 months

@johnjhorton @emollick @Kehang_Zhu Highlight of my week, let me tell you, no joke! A lot of the discussion was really cool too😋

2

0

7

@BenSManning

Benjamin Manning

3 months

Could there be a better name for a Python class that imputes missing data in surveys with LLMs than `SuperImputer`?

0

1

7

@BenSManning

Benjamin Manning

2 months

@mollycrockett @ang3linawang Who exactly are the people claiming this? I don’t know social scientists who seriously think the models as of now are valid human substitutes, but that doesn’t mean they aren’t useful for learning about people. See by @johnjhorton for a great perspective

Tweet card media

Large Language Models as Simulated Economic Agents: What Can We Learn from Homo Silicus?

0

0

7

@BenSManning

Benjamin Manning

1 month

Trying to re-optimize my workflows across both GPT 4 and Claude 3, what tasks to people prefer one of the other for? I think I default to much towards GPT-4…

7

0

7

@BenSManning

Benjamin Manning

2 months

If you feel any gaps in your intuitive understanding of LLMs, a MUST watch

Tweet card media

But what is a GPT? Visual intro to transformers | Chapter 5, Deep...

Unpacking how large language models work under the hoodEarly view of the next chapter for patrons: https://3b1b.co/early-attentionSpecial thanks to these sup...

www.youtube.com

1

1

7

@BenSManning

Benjamin Manning

1 month

epic

@johnjhorton

John Horton

1 month

I've been posting about @ExpectedParrot & edsl <>, and want to share in more detail what we've been working on, where we are going and why. (1/n)

1

23

98

0

0

6

@BenSManning

Benjamin Manning

2 months

Worth a read if you haven’t seen

@behscientist

Behavioral Scientist

2 months

Daniel Kahneman was one of the keenest chroniclers of humans as we actually are. And he knew better than anyone his own shortcomings & blind spots. To remember both his greatness & humanity, we collected memories & lessons from some of those closest to him.

2

20

71

0

0

6

@BenSManning

Benjamin Manning

20 days

Really cool work here from @allen_ai : . I'm interested to see how they deal with data leakage problems in the future for evaluation, I'm sure they have some great minds working on it!

Tweet card media

Data-driven Discovery with Large Generative Models

How do you boil the ocean? That impossible task is what researchers in every field try to accomplish when they sort through the existing…

blog.allenai.org

0

1

6

@BenSManning

Benjamin Manning

1 month

I even tried to figure out how to get myself a new iPad, but being assertive doesn't seem to help :(

Tweet media one

Tweet media two

@johnjhorton

John Horton

1 month

My students are now revising their paper about a tool, apparently using that tool to help write the paper

Tweet media one

Tweet media two

2

1

13

1

0

5

@BenSManning

Benjamin Manning

2 years

Will all-time seed records as my beta priors beat my friends’ brackets this year??? Probably not, but it would be SO funny🙊

Tweet media one

1

0

5

@BenSManning

Benjamin Manning

20 days

TBH the ex machina house and the Joker's lair seem equally unappealing to me...

@sama

Sam Altman

20 days

i try not to think about competitors too much, but i cannot stop thinking about the aesthetic difference between openai and google

Tweet media one

Tweet media two

3K

1K

26K

0

0

4

@BenSManning

Benjamin Manning

27 days

@yudapearl @ancelkeyslegacy If you have any ideas to make this better @yudapearl , we ( @Kehang_Zhu and @johnjhorton ) would love to hear! This whole project was inspired by your work, and we want to make it as strong as possible :) 5/n

1

1

4

@BenSManning

Benjamin Manning

2 months

@barbarikon @emollick Agreed with @Kern_Alan , also there’s no reason a human can’t be involved in the loop here, a researcher can easily chime in a tweak things to explore new ideas, and ultimately, will need to think deeply about implications, just as they always do

1

0

4

@BenSManning

Benjamin Manning

2 months

very cool stuff and super easy to use!

@ExpectedParrot

Expected Parrot

@ExpectedParrot

2 months

EDSL makes it easy to scale data labeling tasks. How it works:

Tweet media one

1

4

7

0

0

5

@BenSManning

Benjamin Manning

2 years

@lirabenjamin @ephemeralidea Show how to pipe, but also how to not pipe - I found that some students found each way more intuitive when learning😊

0

0

5

@BenSManning

Benjamin Manning

1 month

Great intuition builder

@GeostatsGuy

Michael Pyrcz🌻

1 month

A student told me, "I don't get the difference between model fitting by: 1. minimum Mean Square Error (MSE) and 2. Maximum Likelihood Estimation (MLE)" To help this student, I built out this #Python @matplotlib dashboard for fitting a Gaussian parametric distribution to data…

32

617

4K

0

0

5

@BenSManning

Benjamin Manning

4 months

When your advisor is optimizing figures in a paper just for twitter.... 😂 @johnjhorton

Tweet media one

0

0

5

@BenSManning

Benjamin Manning

5 months

@johnjhorton @EugeneVinitsky @emmabwiles I can get behind this! I Couldn't agree more with point 1), i'd even go a step further, 1-to-many is a total waste of time. With some workshopping a little more toward the topics you want to teach, the role-playing thing might just work! can't say this enough 1-to-many = BAAAAAD

0

0

5

@BenSManning

Benjamin Manning

1 month

But if I never missed a payment, that would be pretty cool… as seen in my DC hotel room 👀

Tweet media one

0

0

5

@BenSManning

Benjamin Manning

2 years

when did CVS finally give the option to stop printing receipts??? Must be cutting down global paper usage at LEAST 10%

0

0

4

@BenSManning

Benjamin Manning

27 days

@yudapearl @ancelkeyslegacy What we did was a proof-of-principle---that it is possible to set up idea-generation to automated experimentation systems without humans in the loop. Now we have to figure out when such experiments tell us interesting things that can complement/improve social science research 4/n

1

0

3

@BenSManning

Benjamin Manning

27 days

@yudapearl @ancelkeyslegacy It's easy for an LLM to generate these SCMs, basically ad infinitum, and while the above is quite simple, you can imagine that these could become more interesting / creative both in term of what the causes are (more interesting variables) and the causal structure itself 2/n

1

0

3

@BenSManning

Benjamin Manning

15 days

This is absolutely insane - an additional job offer is huge. wow.

@JingyiQiu4

Jingyi Qiu

16 days

🚀Our field experiment ( @yanchen @alain_cohn & Al Roth) on Twitter shows that social media promotion can boost job market outcomes, especially for women❗️ ✅Job market candidates in the treatment group received one additional flyout, with women receiving 0.9 more job offers.…

Tweet media one

22

136

494

1

0

4

@BenSManning

Benjamin Manning

2 months

@IvelinOfThings @emollick Also this, and there will be more to come! , there’s so much unexplored latent information in these models from humans, who know what insights we may find

Tweet card media

Turning large language models into cognitive models

Large language models are powerful systems that excel at many tasks, ranging from translation to mathematical reasoning. Yet, at the same time, these models often show unhuman-like...

1

1

3

@BenSManning

Benjamin Manning

7 months

@johnjhorton @emmavaninwegen just spit balling here, what if the whole course was just about using LLMs for social science??? asking for a friend who has to take this class...

1

0

3

@BenSManning

Benjamin Manning

3 months

Why did we get like 4 inches of snow in mid march?

Tweet media one

2

0

4

@BenSManning

Benjamin Manning

8 months

get excited for some cool stuff!

@daveholtz

David Holtz

8 months

Very excited to have received an "Accelerating Foundation Models Research" research grant from @Microsoft to support some ongoing research looking at how folks interact with and learn how to more effectively use generative AI tools

8

3

65

0

0

4

@BenSManning

Benjamin Manning

3 months

I can imagine it’s well-deserved!

@daveholtz

David Holtz

3 months

I'm very grateful for this amazing piece that was written about the online marketplaces and platforms MBA course that I teach at @BerkeleyHaas !

5

10

69

0

0

4

@BenSManning

Benjamin Manning

2 months

okay, for real, what do I do if I'm confident a review was written by GPT? Even if it's positive! Not sure what the move is here...

3

0

3

@BenSManning

Benjamin Manning

1 month

very thought-provoking

@johnjhorton

John Horton

1 month

This paper w/ Justin (a fantastic 1st year PhD) & Manish was really fun to write - we spent time thinking about how these new technologies affect might labor market matching in the near & far term. So many new & interesting questions for economists, computer scientists, HR folks,…

0

6

51

0

0

4

@BenSManning

Benjamin Manning

2 months

@a_avilauribe We will publish the code in the future, it will be easy to replicate the experimental design, although some of the experiments are at a high temperature so the EXACT results may be a tad different

1

0

4

@BenSManning

Benjamin Manning

2 months

@emollick I think our paper is really relevant to your point:

@johnjhorton

John Horton

3 months

New working paper by @BenSManning & @Kehang_Zhu that & I’m super excited about

Tweet media one

11

78

310

0

0

2

@BenSManning

Benjamin Manning

2 months

really cool stuff from an amazing teacher, if you're unfamiliar with the subject, I'm sure @tedsvo offers the best intro around - I speak from personal experience :)

@tedsvo

Teddy Svoronos

2 months

I’ve been teaching a course with @5harad and @danmlevy on the science and implications of Gen AI, aimed at policy students without technical backgrounds. We’ve been posting videos and materials as we teach; 1/3 of the content is up. Please check it out!

0

6

37

1

0

3

@BenSManning

Benjamin Manning

3 months

@Hoper_Tom @eagle_hz @_DougDowney @elgreco_winter @allen_ai @erichorvitz @emollick @andrewwhite01 @MarioKrenn6240 @profjamesevans @IntuitMachine might have some synergies here 😊

@johnjhorton

John Horton

3 months

New working paper by @BenSManning & @Kehang_Zhu that & I’m super excited about

Tweet media one

11

78

310

1

0

3

@BenSManning

Benjamin Manning

27 days

@yudapearl @ancelkeyslegacy @Kehang_Zhu @johnjhorton If the LLM has captured enough information about human behavior via it's training on so much data and can re-generate it through simulated experiments, there is hopefully a lot we can discover with such a method! 6/n

0

0

3

@BenSManning

Benjamin Manning

4 months

@SophiaKazinnik @johnjhorton @DataColada if anyone has any other ideas they want us to check out, comment below! Or if you want to know how to, no joke, do this is in like 10 mins or less, DM me/sophia :)

0

0

3

@BenSManning

Benjamin Manning

2 months

@CriticalAI @mollycrockett completely agree about the lack of docs, a successful future with LLMs as social science tools will need a transition to open sourced models with transparent RLHF processes. That doesn’t mean current models (eg gpt4) are useless, but transparency will be key , h/t @johnjhorton

0

0

3

@BenSManning

Benjamin Manning

2 months

@alexolegimas @Kehang_Zhu @johnjhorton Thanks Alex, although I don't think anyone's an expert here yet with LLM experiments/simulations, as @msbernst said to me a few months ago, no one's been in this space for more than a year or two max!

1

0

3

@BenSManning

Benjamin Manning

27 days

@yudapearl @ancelkeyslegacy To test the hypothesis, en silica that is (we are very clear that these are simulations), we can automatically generate LLM-powered agents that vary on the exogenous dimensions of the SCM, have them interact over and over in independent scenarios to get path estimates 3/n

1

0

3

@BenSManning

Benjamin Manning

2 months

@avi_collis Thanks Avi! Couldnt have happened without my brilliant teammate/co-first author @Kehang_Zhu 😋

0

0

3

@BenSManning

Benjamin Manning

3 months

@johnjhorton As a current PhD student, let me tell, I don’t even remember how I functioned on a day to day basis before Claude 3, it’s like my life started this week…

1

0

3

@BenSManning

Benjamin Manning

2 months

A beautiful tribute to Danny, I couldn't help but smile when reading through the memories

Tweet card media

Remembering Daniel Kahneman: A Mosaic of Memories and Lessons - By Evan Nesterak - Behavioral...

A collection of memories and lessons from Daniel Kahneman's close collaborators and colleagues. Learn more about Kahneman the thinker, the coauthor, the writer, as well as Kahneman the partner,...

behavioralscientist.org

0

0

3

@BenSManning

Benjamin Manning

3 months

Is all of grad school just being a quantum version of oneself?

Tweet media one

0

1

3

@BenSManning

Benjamin Manning

2 months

@IvelinOfThings @emollick Actually they have!

2

0

2

@BenSManning

Benjamin Manning

28 days

@Econ_4_Everyone @UChicago That's incredible! how is the yield so high??? is UChicago just so good at convincing students to come or is it admitting students with fewer outside options?

0

0

2

@BenSManning

Benjamin Manning

2 months

@johnjhorton If only he’d won a Nobel prize or something acknowledging how useful his work was, then we could appreciate it…

0

0

3

@BenSManning

Benjamin Manning

2 months

@leemcalilly @johnjhorton @Kehang_Zhu It will be! Still in the process of cleaning things up so others can navigate our messy academic-y code ;)

0

0

2

@BenSManning

Benjamin Manning

2 years

Please do this yesterday

@HandNF

Nick Felker, 🗽🇺🇦🇹🇼

2 years

"The White House Office of Science and Technology Policy estimates it would cost just over $24 billion to have prototype vaccines ready for each of the 26 known viral families that cause human disease" This is a massive bargain.

18

463

2K

0

0

2

@BenSManning

Benjamin Manning

3 months

@cognition_labs @johnjhorton @Kehang_Zhu

0

0

2

@BenSManning

Benjamin Manning

6 months

@johnjhorton @apo_filippas The future of social science has arrived!

0

0

2

@BenSManning

Benjamin Manning

3 months

@eizenb_alon @johnjhorton it uses the SCM, but is either bad at math, or uses other “intuition” as you say, Its predictions are 6x better (in terms of MSE) when it has the SCM compared to not, but is still worse than the SCM’s mechanical predictions, see figure 6 in the paper for reference @eizenb_alon

1

0

1

@BenSManning

Benjamin Manning

1 month

@johnjhorton @amcafee @AndreyFradkin Not sure if you've seen this paper @johnjhorton , but pretty sure this new innovation came out of a university: . You can't pull out of a bet if it's already decided!

Tweet card media

Automated Social Science: Language Models as Scientist and Subjects

We present an approach for automatically generating and testing, in silico, social scientific hypotheses. This automation is made possible by recent advances in large language models (LLM), but...

0

0

2

@BenSManning

Benjamin Manning

2 years

great first day! :)

0

0

2

@BenSManning

Benjamin Manning

6 months

@johnjhorton @OnuWeiss @yudapearl @GaryMarcus what if gpt could simulate this scenario on itself to see what would happen, do you think it would get it right @johnjhorton ;)

0

0

1

@BenSManning

Benjamin Manning

6 months

@JingyiQiu4 @johnjhorton @Kehang_Zhu DM me and I can share one!

1

0

2

@BenSManning

Benjamin Manning

2 years

Really impressive stuff! Even though I only understand about 1% of it (if that)

@BenJAuerbach

Ben Auerbach

2 years

Very excited to share my work on inferring the circadian phases of cells from scRNA-seq data (1/n)

2

21

104

0

0

2

@BenSManning

Benjamin Manning

1 month

@alexolegimas cool, any suggestions on resources for the mathematically competent, but the chemically/biologically inept to learn more?

1

0

1

@BenSManning

Benjamin Manning

2 months

I guess as soon as someone gets an AP job they also have the same problem.... @emmabwiles

Tweet media one

1

0

2

@BenSManning

Benjamin Manning

1 month

Pretty interesting/potentially compelling take. I wonder what Wolphram would say in response to the success of alpha fold, also seems like there could be a lot of information hidden in “simple” language (his words not mine) about human behavior—Empirical question to figure out!

@tsarnick

Tsarathustra

1 month

Stephen Wolfram says AI's success at language is because it was shallower than we knew, but in science it won't be able to make more progress than humans

149

195

959

1

0

2

@BenSManning

Benjamin Manning

2 years

Preach! How is he allowed to play this year…

@RGIII

Robert Griffin III

2 years

The NFL had an opportunity to show it had learned from its ugly history on standing up for Women with this Deshaun Watson case and IT FAILED. 11 games and a 5 million dollar fine doesn’t fit what he was accused of doing and found to have done by Judge Sue L. Robinson. Sickening.

3K

4K

42K

0

0

2

@BenSManning

Benjamin Manning

1 month

@johnjhorton no joke, that's a review Trump would write, his voice exactly 😂

0

0

1

@BenSManning

Benjamin Manning

2 years

@sinanaral Maybe ML coursework should be included with stats for PhD students requirements…

1

0

2

@BenSManning

Benjamin Manning

3 months

@eizenb_alon @johnjhorton Lucky??? Nothing I do is luck😉, but I think we could actually iterate over functional form assumptions too, there’s an optimization problem to be solved here. But the objective function is a bit unclear…

0

0

2

@BenSManning

Benjamin Manning

2 months

@BrianRoemmele @emollick Much appreciated @BrianRoemmele , we are gonna keep studying this stuff and we will figure out just how much it can (and can’t) do👊

0

0

2

@BenSManning

Benjamin Manning

2 years

👀👀👀👀 new laptop stickers be like #truelove

Tweet media one

0

0

2

@BenSManning

Benjamin Manning

2 months

@emreyz_ A large Part of the difficulty is that it’s hard to experiment on humans, that’s not to say that LLMs are by any means humans, but they’re an exciting new tool measure human behavior and think about experiments in new ways

1

0

2

@BenSManning

Benjamin Manning

1 month

@johnjhorton @mbodhisattwa @allen_ai @Kehang_Zhu @mbodhisattwa if you haven’t seen, check out our paper (with John): , very relevant, looking forward to hearing from you😋!

Tweet card media

Automated Social Science: Language Models as Scientist and Subjects

1

0

2

@BenSManning

Benjamin Manning

6 months

@benleo_econ @JingyiQiu4 @johnjhorton @Kehang_Zhu shoot me an email or DM me :)

0

0

1

@BenSManning

Benjamin Manning

27 days

Generating hypos is possibl (just ask GPT to come up with ideas!) and there is lots of work on LLMs as proxy agents. I think real questions are 1. When are hypotheses are at all interesting/novel (TBD!) 2.When are llms as proxy agents informative about people (also mostly TBD!)

@yudapearl

Judea Pearl

27 days

Fascinating and ambitious paper, but I couldn't go beyond these two statements: "We know that: 1. Machine learning can be used for automated hypothesis generation (e.g., 1; 2) 2. Large Language Models (LLMs) can simulate human subjects for testing hypotheses (e.g., 1; 2; 3; 4,…

6

11

49

0

0

2

@BenSManning

Benjamin Manning

2 years

My dad was in NYU Prof. Maitland Jones' Chemistry class 50 years ago - it taught him how to learn and made him the doctor he is today even though it was harder than anything he'd ever done and he could've failed. Sometimes, teachers DO know what's best!

Tweet card media

The Problem with Student-Centered Education - HxA

Student satisfaction-focused initiatives hinder true educational objectives, limiting rich learning experiences designed by universities.

heterodoxacademy.org

0

1

2

@BenSManning

Benjamin Manning

2 years

Somehow, I have 5 brackets, with 5 different winners and 5 very different final fours and they all still suck!!! 😳😳😳💔

0

0

2

@BenSManning

Benjamin Manning

2 years

So they would just as soon have linkedin do nothing and maintain a mediocre status quo for all? This kind of research allows us to IMPROVE professional outcomes for everyone. How do you think new medications are validated.....

1

0

2

@BenSManning

Benjamin Manning

3 months

@eizenb_alon @johnjhorton could also just let it keep guessing causes, I bet that if you had it guess 100 causes, one would be “second highest reservation price”, then when it runs experiment, it would predict the result! This SCM is inthe appendix, it’s almost perfect, the path is ~1 and so is the R^2

2

0

2

@BenSManning

Benjamin Manning

3 months

@johnjhorton Who knew PhD students could be so handsome?? Even just the backs of their heads… @indie5OO @Peyman_Shahidi

1

0

2

@BenSManning

Benjamin Manning

8 months

@johnjhorton great minds think alike - I slacked you this at almost the exact same time you tweeted this 😝

1

0

2

@BenSManning

Benjamin Manning

2 years

@nickchk From a frequent viewpoint (the only view to use pvalues), these statements don’t make sense. The pvalue only matters in that it is above or below alpha, which controls the long term error rate (the alpha). One can’t have multiple levels of significance - it simply is or it isn’t!

1

0

2

@BenSManning

Benjamin Manning

3 months

Really cool position at Oxford working on generative AI and social science experimentation! Cutting-edge stuff with a top research team with @RayDuch

0

0

2

@BenSManning

Benjamin Manning

3 months

@eizenb_alon @johnjhorton Yes! The agents in the experiments behave in line w/ theory (as you say), but when asked to make predictions, the LLM appears to largely ignore theory. There’s a dichotomy in its use of knowledge and its behavior, and I bet auction lit is very represented in training corpus!

1

0

2

@BenSManning

Benjamin Manning

3 months

@dggoldst somehow yes, Crystal lake in new hampshire

0

0

2

@BenSManning

Benjamin Manning

3 months

@MaxGhenis @johnjhorton @ThePolicyEngine Oh that’s very cool

0

0

2

@BenSManning

Benjamin Manning

1 month

@alexolegimas Interesting to hear! I’ll admit that I don’t know much about protein folding (my priors are that wolphram knows wayyy more than me about everything). Just to make sure I understand, this would make alpha fold more liken to Deep Blue “beating” chess?

1

0

1

@BenSManning

Benjamin Manning

1 month

@joshuafmask Great point! But there’s still a switching cost over and over…😭

0

0

1

@BenSManning

Benjamin Manning

2 months

I'm already jealous of the person who gets this job

@johnjhorton

John Horton

2 months

Looking for a Python developer - job description & application link here:

Tweet media one

2

11

30

0

0

1

@BenSManning

Benjamin Manning

6 months

@OnuWeiss @johnjhorton @yudapearl i think that's right!

1

0

1

@BenSManning

Benjamin Manning

21 days

@TheFigen_ @johnjhorton I'm gonna stay bullish here :)

0

0

1