Miklós Koren

Miklós Koren

@koren.mk

Professor at ceu.edu.

2.2K followers1.1K following315 posts

Pinned

GenAI is not a machine and shouldn't be modelled as such. Given the range of tasks it can do if properly supervised, it is more like a coworker. We explore the economics of GenAI adoption among knowledge workers in a new paper: koren.mk/publications...

koren.mkThe Directions of Technical ChangeGenerative AI is directional: it performs well in some task directions and poorly in others. Knowledge work is directional and endogenous as well: workers can satisfy the same job requirements with di...

GenAI is not a machine and shouldn't be modelled as such. Given the range of tasks it can do if properly supervised, it is more like a coworker. We explore the economics of GenAI adoption among knowledge workers in a new paper: koren.mk/publications...

koren.mkThe Directions of Technical ChangeGenerative AI is directional: it performs well in some task directions and poorly in others. Knowledge work is directional and endogenous as well: workers can satisfy the same job requirements with di...
View full thread

4. AI adoption depends on both worker and job characteristics. The same worker may use AI in one job but not another. Think academia vs consulting vs tech firm. 5. Adoption is faster when jobs and workers are more flexible, which is the case for knowledge work. Relation to lit:

GenAI is not a machine and shouldn't be modelled as such. Given the range of tasks it can do if properly supervised, it is more like a coworker. We explore the economics of GenAI adoption among knowledge workers in a new paper: koren.mk/publications...

koren.mkThe Directions of Technical ChangeGenerative AI is directional: it performs well in some task directions and poorly in others. Knowledge work is directional and endogenous as well: workers can satisfy the same job requirements with di...

Coworkers don't change the production function. They add to the set of outcomes that are feasible. This formulation is unusual in labor, so we borrow Production Possibility Sets from international trade. Worker + AI = convex hull of two PPSs.

This simple insight leads to a lot of predictions. 1. Absolute advantage not enough for AI adoption. AI needs a comparative advantage relative to worker's current tasks. 2. No adoption up to some threshold, quick adoption afterwards. 3. Some workers only partially adopt.

Vinyl skills: 1. What is vinyl? 2. It has two sides. 3. Can place the needle right before the bridge of Karma Police.

I had an AI agent refer to SQLite as "noSQL-like DB." Looks like it was trained on Medium/LinkedIn bullshit but I told it in no uncertain terms that this does not pass with me.

I subscribed to Google Gemini and cancelled it within 5 minutes. I wanted to create a Google Form from a survey instrument in Google Doc. Gemini cannot do that. It suggested to go to Google Forms and create one myself, using the Doc as a guide.

Claude Sonnet 4.5 is currently working on the script that will get me what I want. Users don't need chatbots. They need to seamlessly integrate AI into their workflows to minimize their pain. If Google doesn't understand this, they are in trouble.

I have been working with coding AI agents several hours a day in the past couple of months. You get to know their style. Grok is like Elon Musk. Moving fast and breaking things at 10x speed. Not apologizing either.

Why I love OneDrive, part 78. My local folder is called /Users/koren/OneDrive\ -\ Central\ European\ University\ \(CEU\ GmbH\ Hungarian\ Branch\ Office\)/ Luckily, I can symlink.

Austrian Airlines is pitching Budapest for me. I mean, I love the city. But as long as this is how well AI understands us, I feel safe.

I see a lot of scripts where the file name encodes some structure. Like analysis_this.do, analysis_that.do and plot_this.do, plot_that.do If you have such a clear structure, you are best served by folders. It just as easy to write "src/analysis/whatever.do" as "src/analysis_whatever.do".

To me this seems similar to the early days of the internet. A promising technology with no product. Once the initial bubble bursts, we may yet see fantastic products leveraging the tech.

Ed Zitron@edzitron.com

Newsletter: Generative AI is a marketing con perpetuated by Sam Altman and Dario Amodei through a mixture of lies and half-truths carried by a tech media that fails to ask the right questions. Outside of ChatGPT, generative AI companies barely get any traffic at all. www.wheresyoured.at/longcon/

New level unlocked: typing an email, three letters into a word in the second sentence, get distracted, then go back and try to figure out what words begin with those three letters. #adhdlife

Why not just set up a national firewall? Blocking terms like, you know, Tiananmen Square.

Judd Legum@juddlegum.bsky.social

1. EXCLUSIVE Today, the NSA is planning a "Big Delete" of websites and internal network content that contain any of 27 banned words, including "privilege," "bias," and "inclusion," a NSA source tells Popular Information. The massive purge is creating chaos, taking down "mission-related" work

I love how the movie and Chalamet, in particular, get across the “fuck it” attitude of Dylan. Fans don’t like my new music? Fuck them, I don’t owe them anything.

Miklós Koren reposted

Librarians are going to save us 👏

Library Innovation Lab@harvardlil.bsky.social

We just launched a 16TB archive of every dataset that has been available on data.gov since November. This will be updated day by day as new datasets appear. It can be freely copied, and we're sharing the code behind it to help others make their own archives of data they depend on.

One aspect of ADHD is that your attention is either 1 or 0. No in between. The former is thought of as the ADHD superpower ("hyperfocus"), but the latter is often equally helpful. - How can you not care? - Like this.

I bought some Nvidia stock. The news is not that AI is a hype. It’s that LLMs are a commodity in abundant supply. What product is complementary to LLMs? Even for inference, you need quite a bit of GPU compute.

Rebalanced the tonearm of my player. I just followed the step-by-step instructions of the manufacturer, but it helps that I now understand why it matters.

Fascinating read. Reminds me of the dot-com bubble years (yes, I'm that old), when every party conversion would revolve around how revolutionary the internet is and why we should all invest in IT stock. The explanation of how consumer surplus is going to be captured was not very lucid.

Ed Zitron@edzitron.com

I'm old enough to remember when OpenAI rushed their reasoning model out because Altman was "eager to prove to potential investors in the company's latest funding round that OpenAI remains at the forefront of AI development," which it's very clear that they weren't www.wheresyoured.at/oai-business/