travis gerke
@travisgerke.bsky.social
sometimes #dataBS #causalsky #statsky #episky #rstats but mostly yolo skeeting nonsense
The real magic of AI is that it caused everyone to forget the fundamental principles that govern how we probabilistically learn about the world. Small n / curse of dimensionality / no validation on well curated data? No problem. Visit my C seed startup at AI dot ai 💸 💸
November 5, 2025 at 12:28 AM
The real magic of AI is that it caused everyone to forget the fundamental principles that govern how we probabilistically learn about the world. Small n / curse of dimensionality / no validation on well curated data? No problem. Visit my C seed startup at AI dot ai 💸 💸
Congratulations Dodgers, Cinderella team for the ages
November 2, 2025 at 4:08 AM
Congratulations Dodgers, Cinderella team for the ages
Reposted by travis gerke
Wake me up when gen AI can schedule a recurring meeting at a time convenient for both parties and proactively adjust future conflicts.
October 21, 2025 at 8:14 PM
Wake me up when gen AI can schedule a recurring meeting at a time convenient for both parties and proactively adjust future conflicts.
This is the back gate in my driveway which borders Camp Pendleton, I’ve circled the area where we see artillery drills. Does anyone have info about what time today USMC is shooting missiles over the 5 into this area? Want to make sure us and kids are away from home @mikelevin.org @governor.ca.gov
October 18, 2025 at 3:23 PM
This is the back gate in my driveway which borders Camp Pendleton, I’ve circled the area where we see artillery drills. Does anyone have info about what time today USMC is shooting missiles over the 5 into this area? Want to make sure us and kids are away from home @mikelevin.org @governor.ca.gov
TIL 36.4 is "significantly larger" than 121.7 thanks google AI please build a giant data center in my backyard for more of this
October 14, 2025 at 3:24 PM
TIL 36.4 is "significantly larger" than 121.7 thanks google AI please build a giant data center in my backyard for more of this
Our family spends around $75 per month on prescriptions. Did a quick plot to find out how much the 650% is going to save us and hell to the yes the pharmacy is about to *pay us* $412.50 per month
October 10, 2025 at 10:24 PM
Our family spends around $75 per month on prescriptions. Did a quick plot to find out how much the 650% is going to save us and hell to the yes the pharmacy is about to *pay us* $412.50 per month
This person sounds funny and cool but I am definitely worried about the “baby in placenta” part
RFK Jr: "Somebody showed me a TikTok video of a pregnant woman at 8 months pregnant -- she's an associate professor at the Columbia Medical School -- and she is saying 'F Trump' and gobbling Tylenol with her baby in her placenta. The level of Trump Derangement Syndrome is now a pathology."
October 9, 2025 at 5:50 PM
This person sounds funny and cool but I am definitely worried about the “baby in placenta” part
After exits from padres and boston I had no horse in the world series race but this turned me. Seattle all the way
Fan wearing a "Dump Here" jersey with 61 caught Cal's HR 🔥
Immediately changes into an identical 62 jersey
Immediately changes into an identical 62 jersey
October 8, 2025 at 2:21 AM
After exits from padres and boston I had no horse in the world series race but this turned me. Seattle all the way
I'm convinced god's table must be more chaotic than this
October 8, 2025 at 2:17 AM
I'm convinced god's table must be more chaotic than this
I think this about yankees and dodgers fans all the time
The “fuck the law we’re putting Trump on a coin” thing is once again making me marvel at how many people apparently grew up without learning how to be grotesquely embarrassed
October 4, 2025 at 3:50 AM
I think this about yankees and dodgers fans all the time
Reposted by travis gerke
Your wife is empirically correct.
September 26, 2025 at 3:38 PM
Your wife is empirically correct.
Obv not the most important thing right now, but I am continuously astounded that Google invests heavily to push incorrect AI slop to the top of search results. In this easily verifiable case, the season series is actually 3-3. Even their own counting of the 2-2 description is wrong!
September 26, 2025 at 2:20 PM
Obv not the most important thing right now, but I am continuously astounded that Google invests heavily to push incorrect AI slop to the top of search results. In this easily verifiable case, the season series is actually 3-3. Even their own counting of the 2-2 description is wrong!
These ACIP comments from the public are brutal 💀 Kulldork was so clearly unhappy to field them, citing that they are "required by law"
September 19, 2025 at 6:31 PM
These ACIP comments from the public are brutal 💀 Kulldork was so clearly unhappy to field them, citing that they are "required by law"
Periodically tuned into ACIP 🗑️ Repeated calls for placebo controlled RCTs.
Just shouting the obvious (🤡 ACIP excepted) into the void: we identify life-saving therapies through active control RCTs all the time. Eg effective therapies for cancer exist, we cannot withhold them to study new treatments.
Just shouting the obvious (🤡 ACIP excepted) into the void: we identify life-saving therapies through active control RCTs all the time. Eg effective therapies for cancer exist, we cannot withhold them to study new treatments.
Presumably Dr. Robert Malone gets out of bed each day, maybe showers or gets in his car or drinks water or eats food or exercises or even flies in a plane.
Does he demand a randomized clinical trial showing that there are zero possible adverse events to any of these actions?
Does he demand a randomized clinical trial showing that there are zero possible adverse events to any of these actions?
September 19, 2025 at 5:09 PM
Periodically tuned into ACIP 🗑️ Repeated calls for placebo controlled RCTs.
Just shouting the obvious (🤡 ACIP excepted) into the void: we identify life-saving therapies through active control RCTs all the time. Eg effective therapies for cancer exist, we cannot withhold them to study new treatments.
Just shouting the obvious (🤡 ACIP excepted) into the void: we identify life-saving therapies through active control RCTs all the time. Eg effective therapies for cancer exist, we cannot withhold them to study new treatments.
no Apple News, I don't want to read about "Hooray! Stocks hit record highs because there are no jobs meaning interest rates may go down so that billionaires can borrow at lower costs"
September 10, 2025 at 3:51 PM
no Apple News, I don't want to read about "Hooray! Stocks hit record highs because there are no jobs meaning interest rates may go down so that billionaires can borrow at lower costs"
Remember like a year ago when we devoted approx 0% of our effort wondering or talking about what dumb shit the president was doing
September 10, 2025 at 5:35 AM
Remember like a year ago when we devoted approx 0% of our effort wondering or talking about what dumb shit the president was doing
Facts. AI would have helped me with very little of the data infra work I’ve done over the past 10+ years because most is deployed within highly bespoke setups of large orgs with lots of tech debt
I hate the argument that AI is being used in lots of jobs so students have to learn how to use it in school. AI is not that hard to use. Lots of jobs also use job-specific convoluted enterprise software and we don't teach them how to use those in school.
September 7, 2025 at 5:16 PM
Facts. AI would have helped me with very little of the data infra work I’ve done over the past 10+ years because most is deployed within highly bespoke setups of large orgs with lots of tech debt
If only there was a study of, I don't know, 2.5 million children spanning 25 years that found no such link
September 5, 2025 at 6:24 PM
If only there was a study of, I don't know, 2.5 million children spanning 25 years that found no such link
Reposted by travis gerke
This is what I've been saying since 2023 (image below)
"prediction: use of "AI" [...] will come to be broadly associated with cheating, deception, lack of respect for other people, and low quality work that cannot be trusted in important settings"
"prediction: use of "AI" [...] will come to be broadly associated with cheating, deception, lack of respect for other people, and low quality work that cannot be trusted in important settings"
September 2, 2025 at 8:31 AM
This is what I've been saying since 2023 (image below)
"prediction: use of "AI" [...] will come to be broadly associated with cheating, deception, lack of respect for other people, and low quality work that cannot be trusted in important settings"
"prediction: use of "AI" [...] will come to be broadly associated with cheating, deception, lack of respect for other people, and low quality work that cannot be trusted in important settings"
“My package was accepted to CRAN”
*TRUMP TO MAKE ANNOUNCEMENT 2:00 PM ET TUESDAY: WHITE HOUSE
September 2, 2025 at 3:52 AM
“My package was accepted to CRAN”
This is extremely cool. Picked it up and made this DAG in like 15 minutes. Obviously pros/cons versus other DAG tools (e.g. low/no code cstructure and R's ggdag, both of which offer d-separation details and more) but if you have rigid aesthetic requirements this is a great tool to have in your kit
August 20, 2025 at 6:07 PM
This is extremely cool. Picked it up and made this DAG in like 15 minutes. Obviously pros/cons versus other DAG tools (e.g. low/no code cstructure and R's ggdag, both of which offer d-separation details and more) but if you have rigid aesthetic requirements this is a great tool to have in your kit
Good connection made by @adamjkucharski.bsky.social
Take brain cancer: lifetime risk is around .5%. Picture an LLM which detects 100% of cancers and correctly calls "no cancer" 99% of the time when it is absent. Sounds good right? Not really: 67% of the time it says you have cancer when you don't.
Take brain cancer: lifetime risk is around .5%. Picture an LLM which detects 100% of cancers and correctly calls "no cancer" 99% of the time when it is absent. Sounds good right? Not really: 67% of the time it says you have cancer when you don't.
New post, on why a 1% hallucination risk is terrible if you’re using large language models to look for unusual events: kucharski.substack.com/p/small-hall...
August 18, 2025 at 5:17 PM
Good connection made by @adamjkucharski.bsky.social
Take brain cancer: lifetime risk is around .5%. Picture an LLM which detects 100% of cancers and correctly calls "no cancer" 99% of the time when it is absent. Sounds good right? Not really: 67% of the time it says you have cancer when you don't.
Take brain cancer: lifetime risk is around .5%. Picture an LLM which detects 100% of cancers and correctly calls "no cancer" 99% of the time when it is absent. Sounds good right? Not really: 67% of the time it says you have cancer when you don't.
Fascinating, accurate, and unfortunately terrifying. Great read.
Привет, товарищи! It’s another glorious day in which Trump’s America is increasingly indistinguishable from Stalinist Russia, so I thought an armchair comparative analysis was in order.
Specifically about Jay Bhattacharya's twin from times past, Trofim Lysenko.
open.substack.com/pub/rasmusse...
Specifically about Jay Bhattacharya's twin from times past, Trofim Lysenko.
open.substack.com/pub/rasmusse...
Bhattacharya is the New Lysenko
Starving the NIH through ineptitude, bitterness, and thirst for revenge
open.substack.com
August 18, 2025 at 3:55 PM
Fascinating, accurate, and unfortunately terrifying. Great read.
Reposted by travis gerke
☹️ frowning + 😒 unamused + 👊 fist-bump =
August 17, 2025 at 10:00 PM
☹️ frowning + 😒 unamused + 👊 fist-bump =
Reposted by travis gerke
😹 laughing-cat + 🤪 crazy =
August 13, 2025 at 6:00 AM
😹 laughing-cat + 🤪 crazy =