Ethan Siegel

Building Voice AI Agents @VocaraAI | MIT CS/AI | Former @MIT_Baseball, Occasional Ball Knower

[Danny Trinh @dtrinh]

@signulll except when open source teams win rings, everyone wins rings

[signüll @signulll]

meta is basically the yankees or the dodgers right now. they’re loading up the roster with elite ai researchers, dumping billions into infra, & signaling we’re going for it. when you do that, the world expects rings. you don’t get to say “we’re still experimenting” after you dropped $20b+ & poached half of openai.

the pressure is now insane. every paper, every product, every demo will be scrutinized like a playoff performance. all eyes on the scoreboard.

[Ethan Siegel @ethansiegel22]

@basedjensen As long as intelligence continues to move with scale (size, reasoning time) and quality of knowledge distillation continues to hold, it feels likely that the majority of self hosted cases would be distillations of the true frontier models

[Ethan Siegel @ethansiegel22]

@mattparlmer A separator for startups could be having the engineering discipline to design systems with efficient use of model intelligence rather than raw use of scale. Then enough inference know-how to serve their minimum viable stack of models on energy efficient hardware with good perf

[AT @BaseballWRLD_]

By the year 2030 this is going to go from being a meme to the guy that ends up winning the Cy Young Award

[MIT Athletics @MITAthletics]

After falling in game one, 6-4, @MIT_Baseball rallied to score five runs in the final two innings to win game two over Chapman, 8-5!

Recap and stats: https://t.co/QBoN01NeYL

[Ethan Siegel @ethansiegel22]

@drivelinekyle It really is all json 🤷‍♂️

[Driveline Baseball @DrivelineBB]

Athlete @dan_swainer had to ask if he really wanted to sign up for the struggle and failure that comes with chasing his dream. He chose to go all in

[📊 @mason_mcrae]

College hitters slashed .374/.369/.597 against first pitch fastballs that were in the zone last year. Gotta love establishing the fastball by giving up an absolute nuke.

[Ethan Siegel @ethansiegel22]

@ZanRose_ @WinthropBasebal @DrivelineBB @TomRiginos Lets gooooo

[Ethan Siegel @ethansiegel22]

💯💯💯

Way too much unneeded guesswork when it comes to throwing volume and intensity, especially during preseason on ramping and mid season workload management.

@DrivelinePulse can help quantify throwing measures that have largely been left to “gut feel” up to this point.

[Kyle Lindley @kylelindley_]

Do a running workout, check total distance (volume) and pace (intensity) when you're done - extremely standard

Do a throwing workout, check total throws (volume) and average arm speed (intensity) - this can/should be just as standard

@DrivelinePulse

[Ethan Siegel @ethansiegel22]

In all seriousness, go check out this thread on Dylan Cease by @m_kress__

https://t.co/BuikRAUUNB

[Matthew Kress @m_kress__]

The biggest snub of the initial MLB all star game roster?

Dylan Cease

Here are two reasons why he’s been the most improved pitcher of 2022 season

[Ethan Siegel @ethansiegel22]

Dylan Cease went full in on the mustache in 2022, here's how his offerings have developed since 2021.

⛽️ FF - stuff: 182 -> 195
🍔 SL - mph: 85.9 -> 87.2, stuff: 134 -> 146
🪝 CB - mph: 79.9 -> 81.0, stuff: 116 -> 144

Moral of the story? For better stuff, grow a moustache 🥸

[Rob Friedman @PitchingNinja]

Dylan Cease, Nasty 90mph Slider. 😨

[Alex Caravan @Alex_Caravan]

I usually don't tweet cocky shit like this (at least unironically!) but no one else is collecting, analyzing + making actionable as large/elite of a dataset as us in the ⚾️biomechanics

Feel confident af on that & on @DrivelineBB continuing to push the✉️on research here b/c of it

[David Besky @David_Besky]

Thus - a null model (just taking the average mechanics) is able to get to what initially would appear to be somewhat good R^2 and RMSE values and therefore the cutoff for what we’re looking for in terms of accuracy should be higher than it might initially intuitively seem.
n/n

[David Besky @David_Besky]

For markerless mocap validation, we often see R^2 values >0.85 and RMSE values <10deg when comparing all points from time series across systems. But what's good? What would we see if a hypothetical system just predicted avg mechanics for every throw? 1/n

[Alex Caravan @Alex_Caravan]

One of my fave pods of the year!

MEET THE (R&D) INTERNS *Summer 2022*
Got to know
@ethansiegel22
@BayFish03
@aesuare
@anchengyoung42

& their background, @DrivelineBB exp so far, a bit of r&d talk + 1 curveball❓each👇👇

1st Siegel on🎱ranks: me, him, @kylelindley_ @David_Besky

[Ethan Siegel @ethansiegel22]

@David_Besky Expecting your next post to be on glove side horizontal abduction 🔓

[Ethan Siegel @ethansiegel22]

@drivelinebases @Karl_m_7 Took some grit to get through that one

[Ethan Siegel @ethansiegel22]

@drivelinebases @Karl_m_7 6th inning at Pomona. (Not so) proud to have been part of that one lol