David Byler

VP at NRG | Public opinion researcher | Former @washingtonpost political data journalist | Dad, people person, math person!

[David Byler @databyler]

I get people using AI to help them write

It is *wild* to expect nobody to instantly pangram you and to do _nothing_ to at least try to get under 90%

[David Byler @databyler]

I am gonna lose my mind over people not self-pangramming

[Haley Stevens @HaleyforMI]

https://t.co/7tDAMPJoIy

[David Byler @databyler]

And, if you are interested in AI polling in particular, I also have you covered! This is the company substack where I am writing most of the posts: https://t.co/iRmuXHc7YD

[David Byler @databyler]

I don't have election forecasts/predictions today (first time in my professional life!) I do have a ton of election-adjacent stuff and I'm gonna email about it once your inboxes recover mid-month: https://t.co/5wFMw5q3NZ

[David Byler @databyler]

Will count this in the "called it" column: https://t.co/MZfYJpNe8X

[The Daily Beast @thedailybeast]

Allan Lichtman, who has correctly predicted all but one presidential election since 1984, delivered his final verdict on the upcoming midterm election.
https://t.co/9WXtfFcQnb

[David Byler @databyler]

Lichtman will be back. He'll be like "I turned the keys wrong! It was me, not the keys!!" and someone will publish him

[David Byler @databyler]

@rnishimura easy to tweet, easy to code, still not always easy to nail the prediction imo!

[David Byler @databyler]

...unless a certain type of republican is responding too much and another type is not, and it's not a trait that you can measure and fix with weights

this has always been either the fear or the problem (depending on which election you're in)

[Stacy Cay @stacycay]

Non response bias doesn’t mean the polling will miss. This just means they had to make more calls to republicans to get the responses they needed. And they weighted the poll to be Trump +7. It’s just a lot of Kansans who voted for Trump are telling pollsters they’re gonna vote for Hamilton

[matt mohn @mattmxhn]

@xenocryptsite this

[David Byler @databyler]

@rnishimura hmm hmm no i gotta set up an agent that takes these surveys as me, but then quits if anyone gives me an LOI over 15 mins

that is my real way to contribute to the industry

[David Byler @databyler]

as a highly engaged, high information voter -- doing my part for differential non-response by not taking any of these surveys

[David Byler @databyler]

but like the societal/political conversation spaces (news, social media) went way down in positivity right when you'd expect

[David Byler @databyler]

Also while I'm throwing random stuff around, it really does seem like the Jacob Coxon thing was a real shift in how AI politics looks

[David Byler @databyler]

also these numbers are all pretty high/positive bc people tend to count talking about capabilities/efficiencies in daily life as positive news (something that came out of an OE follow-up!)

[David Byler @databyler]

subscribe btw! https://t.co/iRmuXHc7YD

[David Byler @databyler]

I will send results out on the NRG substack later this week. But when we asked which lab was doing the most on safety, the leading answer was "No AI company is working towards safety"

🙃

[Dave Banerjee @DaveRBanerjee]

Anthropic says that even the entire insurance industry in the aggregate will not have the capacity to insure the catastrophic risk that ANTHROPIC ITSELF poses to the world

lol

[David Byler @databyler]

this is one of my forever beefs about 2022 fwiw

Rs had advantages, but blunted them with 1) Dobbs (something they wanted forever) and 2) bad candidates (true unforced errors there) and a small win in the house and generally worse results than they could have

but the whole story somehow revolves around how the results departed from what the conventional wisdom was according to some R-sponsored polls that dropped in october

[David Byler @databyler]

It's weird that we all set pre-election expectations then grade the parties on it

We could just eliminate this step in mid-Nov and assess how well parties did compared to fundamentals, what tactics worked, etc

Like why do takes in mid-Oct with partial info figure into final assessments

[David Byler @databyler]

barely related tangent

i think everyone basically wants to run against the influencers on the other side. bc influencers are obnoxious, say totally insane things, have incentive to get views not votes, etc.

sometimes the electeds/pols feel/act/talk like influencers and it all works out and makes sense.

but i think this ^^ impulse is part of why political debates can feel only semi tethered to elections sometimes

[David Byler @databyler]

exactly! and the joke comes from the Saddam Hussein + Cheesecake Factory image juxtaposition

if Obama had said "Cheesecake Factory is UGGGGLLLYY and OMG TRUMP LIKES IT!!" it would not be funny and remind me of a lefty influencer trying to get hate clicks

I feel like this is not that hard to get?

[Galen Druke @galendruke]

People on here acting like working class people are too stupid to clock camp.

[Jacob Rubashkin @JacobRubashkin]

Polling is like democracy, it’s the worst one except for all the other ones