Josip Krapac
@josipk.bsky.social
I'm interested how to teach machines to reason without soaking up the contents of the whole Internet.
🇭🇷 born, lives in Berlin 🇩🇪
https://josipkrapac.github.io
🇭🇷 born, lives in Berlin 🇩🇪
https://josipkrapac.github.io
American friends, watch m.imdb.com/title/tt3307...
My Undesirable Friends: Part I - Last Air in Moscow (2024) ⭐ 7.5 | Documentary
5h 24m
m.imdb.com
January 18, 2026 at 1:20 PM
American friends, watch m.imdb.com/title/tt3307...
Reposted by Josip Krapac
Honestly this works for everything
“I want to trap hungry 19c ghosts in jars to help us with historical research” ✅
“Please read our holiday card; we got a hungry ghost to write it this year” ❌
“I want to trap hungry 19c ghosts in jars to help us with historical research” ✅
“Please read our holiday card; we got a hungry ghost to write it this year” ❌
If you can substitute "hungry ghost trapped in a jar" for "AI" in a sentence it's probably a valid use case for LLMs. Take "I have a bunch of hungry ghosts in jars, they mainly write SQL queries for me". Sure. Reasonable use case.
"My girlfriend is a hungry ghost I trapped in a jar"? No. Deranged.
"My girlfriend is a hungry ghost I trapped in a jar"? No. Deranged.
January 12, 2026 at 1:11 PM
Honestly this works for everything
“I want to trap hungry 19c ghosts in jars to help us with historical research” ✅
“Please read our holiday card; we got a hungry ghost to write it this year” ❌
“I want to trap hungry 19c ghosts in jars to help us with historical research” ✅
“Please read our holiday card; we got a hungry ghost to write it this year” ❌
European "left" when you ask them about Iran
January 12, 2026 at 4:29 PM
European "left" when you ask them about Iran
Reposted by Josip Krapac
6/ Who should apply?
Researchers with a strong computational background and a genuine interest in biology—from foundations to translation.
Researchers with a strong computational background and a genuine interest in biology—from foundations to translation.
December 21, 2025 at 12:08 PM
6/ Who should apply?
Researchers with a strong computational background and a genuine interest in biology—from foundations to translation.
Researchers with a strong computational background and a genuine interest in biology—from foundations to translation.
Reposted by Josip Krapac
1/ We’re hiring PIs in AI × Biology
The @astar-gis.bsky.social Genome Institute of Singapore (GIS) is expanding its AI & Computation domain and recruiting junior & senior PIs.
The @astar-gis.bsky.social Genome Institute of Singapore (GIS) is expanding its AI & Computation domain and recruiting junior & senior PIs.
December 21, 2025 at 12:06 PM
1/ We’re hiring PIs in AI × Biology
The @astar-gis.bsky.social Genome Institute of Singapore (GIS) is expanding its AI & Computation domain and recruiting junior & senior PIs.
The @astar-gis.bsky.social Genome Institute of Singapore (GIS) is expanding its AI & Computation domain and recruiting junior & senior PIs.
I sometimes wonder if people in the "dark ages" knew they lived in "dark ages", and if our descendants will think that we live in an "dark age". I feel we live in a strage mixture of progress of technology and looming society collapse due to concentration of power into ...
November 27, 2025 at 9:00 PM
I sometimes wonder if people in the "dark ages" knew they lived in "dark ages", and if our descendants will think that we live in an "dark age". I feel we live in a strage mixture of progress of technology and looming society collapse due to concentration of power into ...
EU should declare that English is a second official language in all member states. It's not like Balzac, Göthe or Cervantes will turn in their graves if you allow me to ask questions and find answers about how cross-country pension insurance works, and what I need to provide and to whom in English.
November 16, 2025 at 2:00 PM
EU should declare that English is a second official language in all member states. It's not like Balzac, Göthe or Cervantes will turn in their graves if you allow me to ask questions and find answers about how cross-country pension insurance works, and what I need to provide and to whom in English.
Reposted by Josip Krapac
*Urgently* looking for emergency reviewers for the ARR October Interpretability track 🙏🙏
ReSkies much appreciated
ReSkies much appreciated
November 11, 2025 at 10:29 AM
*Urgently* looking for emergency reviewers for the ARR October Interpretability track 🙏🙏
ReSkies much appreciated
ReSkies much appreciated
"To secure ourselves against defeat lies in our own hands, but the opportunity of defeating the enemy is provided by the enemy himself".
Just a random quote from "Art of war" that I remembered when I heard about FAIR layoffs.
Just a random quote from "Art of war" that I remembered when I heard about FAIR layoffs.
October 22, 2025 at 9:24 PM
"To secure ourselves against defeat lies in our own hands, but the opportunity of defeating the enemy is provided by the enemy himself".
Just a random quote from "Art of war" that I remembered when I heard about FAIR layoffs.
Just a random quote from "Art of war" that I remembered when I heard about FAIR layoffs.
Napisao sam polu-zajedljiv komentar, ali sam ga pobrisao jer bi mogao biti kontroverzan 0.1 osobi na ovom sajtu. Kažu da dobar hejt umire posljednji, ali nisam više siguran. Hejt je krhka i nestabilna emocija, trenutak čekaš i ode. Kao Kairos.
October 16, 2025 at 10:17 PM
Napisao sam polu-zajedljiv komentar, ali sam ga pobrisao jer bi mogao biti kontroverzan 0.1 osobi na ovom sajtu. Kažu da dobar hejt umire posljednji, ali nisam više siguran. Hejt je krhka i nestabilna emocija, trenutak čekaš i ode. Kao Kairos.
AI doesn't kill jobs, people do.
October 2, 2025 at 11:02 PM
AI doesn't kill jobs, people do.
Reposted by Josip Krapac
I wrote a short summary with some context on the blog.
The Actuary's Final Word
Resolving the legacy of Paul Meehl's troubling little book.
www.argmin.net
September 8, 2025 at 2:48 PM
I wrote a short summary with some context on the blog.
Reposted by Josip Krapac
my brother underperformed Q2 so we had to let him go.
you know how it is.
you know how it is.
September 7, 2025 at 6:58 PM
my brother underperformed Q2 so we had to let him go.
you know how it is.
you know how it is.
Are there any works that explore curriculum learning in an unsupervised setting? Does the notion of Y is easier to learn if you first learn X happen automatically in auto-regressive modelling, or diffusion models or any other generative modelling approach?
September 6, 2025 at 7:50 AM
Are there any works that explore curriculum learning in an unsupervised setting? Does the notion of Y is easier to learn if you first learn X happen automatically in auto-regressive modelling, or diffusion models or any other generative modelling approach?
I recently read this paper: alexiglad.github.io/blog/2025/ebt/
I like the idea: first learn a verifier (a model that tells you the goodness/compatibility of a prediction given some context), and then optimize predictions with respect to this verifier.
I like the idea: first learn a verifier (a model that tells you the goodness/compatibility of a prediction given some context), and then optimize predictions with respect to this verifier.
Energy-Based Transformers are Scalable Learners and Thinkers | Alexi Gladstone
What are Energy-Based Transformers and why should I care about them?
alexiglad.github.io
August 27, 2025 at 8:38 PM
I recently read this paper: alexiglad.github.io/blog/2025/ebt/
I like the idea: first learn a verifier (a model that tells you the goodness/compatibility of a prediction given some context), and then optimize predictions with respect to this verifier.
I like the idea: first learn a verifier (a model that tells you the goodness/compatibility of a prediction given some context), and then optimize predictions with respect to this verifier.
After the bubble bursts will LinkedIn be full of posts: "This changes NOTHING"?
August 11, 2025 at 9:07 PM
After the bubble bursts will LinkedIn be full of posts: "This changes NOTHING"?
I think people (who use this term) mean "constraining the model by constraining hyperparameters", while "ordinary" (non-inductive?) bias means "constraining the model by constraining parameters", but 🤷🏻♂️
Which ideas in computer science don't do better with faster computers?
And what is a "human inductive bias?" (I never know what anyone means by "inductive bias." It is a term that defies definition.)
And what is a "human inductive bias?" (I never know what anyone means by "inductive bias." It is a term that defies definition.)
August 9, 2025 at 9:01 PM
I think people (who use this term) mean "constraining the model by constraining hyperparameters", while "ordinary" (non-inductive?) bias means "constraining the model by constraining parameters", but 🤷🏻♂️
God bless Chris Bishop for being consistent with notation within and across his books.
June 13, 2025 at 10:49 AM
God bless Chris Bishop for being consistent with notation within and across his books.
Reposted by Josip Krapac
Unmute adds ears and vocal chords to your favorite text-based language model. A seamless plug-and-play augmentation with easy personalisation through voice conditioning and text instructions. We will open-source shortly.
Talk to unmute.sh 🔊, the most modular voice AI around. Empower any text LLM with voice, instantly, by wrapping it with our new speech-to-text and text-to-speech. Any personality, any voice. Interruptible, smart turn-taking. We’ll open-source everything within the next few weeks.
May 24, 2025 at 8:06 AM
Unmute adds ears and vocal chords to your favorite text-based language model. A seamless plug-and-play augmentation with easy personalisation through voice conditioning and text instructions. We will open-source shortly.
Clicking of PDF links opens them directly in Chrome browser on Android, the same behavior as on iOS. Hallelujah, hallelujah!
May 14, 2025 at 7:26 AM
Clicking of PDF links opens them directly in Chrome browser on Android, the same behavior as on iOS. Hallelujah, hallelujah!
Many of society’s problems would be solved if we decided to treat greed as a mental illness.
Greed is not good: it’s just easy. There are many other things that “clarify and cut through” and it’s easy to sell that greed “captures the essence of the evolutionary spirit”, but it really doesn’t.
Greed is not good: it’s just easy. There are many other things that “clarify and cut through” and it’s easy to sell that greed “captures the essence of the evolutionary spirit”, but it really doesn’t.
April 17, 2025 at 9:04 PM
Many of society’s problems would be solved if we decided to treat greed as a mental illness.
Greed is not good: it’s just easy. There are many other things that “clarify and cut through” and it’s easy to sell that greed “captures the essence of the evolutionary spirit”, but it really doesn’t.
Greed is not good: it’s just easy. There are many other things that “clarify and cut through” and it’s easy to sell that greed “captures the essence of the evolutionary spirit”, but it really doesn’t.
The thoughts that randomly appear when cleaning: why it's called "love triangle" when most of the time it's only "love angle".
April 13, 2025 at 2:19 PM
The thoughts that randomly appear when cleaning: why it's called "love triangle" when most of the time it's only "love angle".
Reposted by Josip Krapac
Job alert: Join us in Mainz as
Max Planck Research Group Leader (W2) in Molecular Design
...and make your own research dreams happen on de novo design, generative models, proteins, materials ...
tinyurl.com/r2xjxnuk
@mpip-mainz.mpg.de
Max Planck Research Group Leader (W2) in Molecular Design
...and make your own research dreams happen on de novo design, generative models, proteins, materials ...
tinyurl.com/r2xjxnuk
@mpip-mainz.mpg.de
Max Planck Research Group Leader (W2) in Molecular Design
We are looking for exceptional early-career scientists conducting computational research with a proven record of accomplishment. The primary focus of this call is on candidates proposing research on b...
tinyurl.com
April 4, 2025 at 4:00 PM
Job alert: Join us in Mainz as
Max Planck Research Group Leader (W2) in Molecular Design
...and make your own research dreams happen on de novo design, generative models, proteins, materials ...
tinyurl.com/r2xjxnuk
@mpip-mainz.mpg.de
Max Planck Research Group Leader (W2) in Molecular Design
...and make your own research dreams happen on de novo design, generative models, proteins, materials ...
tinyurl.com/r2xjxnuk
@mpip-mainz.mpg.de