Is this you? As a journalist, you can create a free Muck Rack account to customize your profile, list your contact preferences, and upload a portfolio of your best work.
Claim your profile
Articles by Ege Erdil
Life after work
Matthew Barnett, Tamay Besiroglu, Ege Erdil October 21, 2025 It’s natural to feel anxious as we approach the inevitable automation of all human labor. Economic theory suggests that full automation will cause wages to collapse, potentially below subsistence level: the bare minimum needed to sustain human life. Yet the full automation of the economy will probably also make most people vastly better off.
The future of AI is already written
Matthew Barnett, Tamay Besiroglu, Ege Erdil October 6, 2025 Innovation often appears as a series of branching choices: what to invent, how to invent it, and when. In our case, we are confronted with a choice: should we create agents that fully automate entire jobs, or create AI tools that merely assist humans with their work? Upon closer examination, however, it becomes clear that this is a false choice.
Most AI value will come from broad automation, not from R&D
A popular view about the future impact of AI on the economy is that it will be primarily mediated through AI automation of R&D. In some form or another, this view has been expressed by many influential figures in the industry: In his essay “Machines of Loving Grace”, Dario Amodei lists five ways in which AI can benefit humanity in a scenario where AI goes well. He considers biology R&D, neuroscience R&D, and economics R&D as three of these ways.
How has DeepSeek improved the Transformer architecture?
DeepSeek has recently released DeepSeek v3, which is currently state-of-the-art in benchmark performance among open-weight models, alongside a technical report describing in some detail the training of the model. Impressively, they’ve achieved this SOTA performance by only using 2.8 million H800 hours of training hardware time—equivalent to about 4e24 FLOP if we assume 40% MFU. This is about ten times less training compute than the similarly performing Llama 3.1 405B.
How do mixture-of-experts models compare to dense models in inference?
In last week’s Gradient Updates issue, I discussed how we can guess that GPT-4o and Claude 3.5 Sonnet have significantly fewer parameters than GPT-4. The most common question I’ve received from readers was some version of the following: Don’t active parameters matter more for inference economics than total parameters? If so, how can you infer the total number of parameters of a model by only looking at its inference cost and speed?
Data Movement Bottlenecks to Large-Scale Model Training: Scaling Past 1e28 FLOP
Introduction Over the past five years, the performance of large language models (LLMs) has improved dramatically, driven largely by rapid scaling in training compute budgets to handle larger models and training datasets. Our own estimates suggest that the training compute used by frontier AI models has grown by 4-5 times every year from 2010 to 2024.
Inferring the model dimension of API-protected LLMs — LessWrong
A new paper by Finlayson et al. describes how to exploit the softmax bottleneck in large language models to infer the model dimension of closed-source LLMs served to the public via an API. I'll briefly explain the method they use to achieve this and provide a toy model of the phenomenon, though the full paper has many practical details I will elide in the interest of simplicity. I recommend reading the whole paper if this post sounds interesting to you.
[2403.05812] Algorithmic progress in language models
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Processor clock speeds are not how fast AIs think — LessWrong
Don't global clock speeds have to go down as die area goes up due to the speed of light constraint? Yes if you use 1 die with 1 clock domain, they would. Modern chips don't. For instance, if you made a die with 1e15 MAC units and the area scaled linearly, you would be looking at a die that's ~ 2e9 times larger than H100's die size, which is about 1000 mm^2.
AI Timelines - LessWrong
Here is a summary of the discussion so far: Daniel made an argument against Hofstadter's law for trend extrapolation and we discussed the validity of that for a bit. A key thing that has come up as an interesting crux/observation is that Ege and Ajeya both don't expect a massive increase in transfer learning ability in the next few years. For Ege this matters a lot because it's one of the top reasons why AI will not speed up the economy and AI development that much.
My thoughts on the social response to AI risk — LessWrong
A common theme implicit in many AI risk stories has been that broader society will either fail to anticipate the risks of AI until it is too late, or do little to address those risks in a serious manner. In my opinion, there are now clear signs that this assumption is false, and that society will address AI with something approaching both the attention and diligence it deserves. For example, one clear sign is Joe Biden's recent executive order on AI safety.
When is correlation transitive? — LessWrong
It's a well-known property of correlation that it's not transitive in general. If X,Y,Z are three real-valued random variables such that ρ(X,Y)>0 and ρ(Y,Z)>0, it doesn't have to be the case that ρ(X,Z)>0. Nevertheless, there are some circumstances under which correlation is transitive. I will focus on two such cases in this post.
$250 prize for checking Jake Cannell's Brain Efficiency — LessWrong
This is to announce a $250 prize for or otherwise indepth reviewing Jacob Cannell's technical claims concerning thermodynamic & physical limits on computations and the claim of biological efficiency of the brain in his post Brain Efficiency: Much More Than You Wanted To Know I've been quite impressed by Jake's analysis ever since it came out. I have been puzzled why there has been so little discussion about his analysis since if true it seems to be quite important.
Childhoods of exceptional people — LessWrong
Let’s start with one of those insights that are as obvious as they are easy to forget: if you want to master something, you should study the highest achievements of your field. If you want to learn writing, read great writers, etc. But this is not what parents usually do when they think about how to educate their kids. The default for a parent is rather to imitate their peers and outsource the big decisions to bureaucracies. But what would we learn if we studied the highest achievements?
Why I think there's a one-in-six chance of an imminent global nuclear war — LessWrong
Many people have asked me what I think the odds are of an imminent major US-Russia nuclear war. My current estimate is about the same as losing in Russian roulette: one in six. The goal of this post is to explain how I arrived at this estimate. Please forgive its cold and analytic nature despite the emotionally charged topic; I'm trying not to be biased by hopes, fears or wishful thinking. My estimate is 30% x 80% x 70% ~ 1/6, as illustrated in the figure and explained below.
The Reader's Guide to Optimal Monetary Policy — LessWrong
Definitely not an area I've done too much research into, but it's always seemed obvious to me that interest rates should simply be allowed to float through market forces. Sometimes money should be cheap to get more of to encourage investment, and in those cases, people are already not taking out enough new loans. In some cases money should be expensive to keep the economy from being effectively manic, and in those cases people are already taking out far too many loans.
A time-invariant version of Laplace's rule — LessWrong
What is the probability that the sun will rise tomorrow? What are the chances of a pandemic happening next year? What are the odds of survival of a new surgery that has been successfully executed only once? These and many other questions can be answered appealing to a general rule: Laplace's rule of succession. This rule describes the probability of a positive outcome given information about past successes.
Staying Split: Sabatini and Social Justice — LessWrong
Author's note: This is a contemporary post about an ongoing current event. It's not a timeless essay in the way that most of my essays are, though it does contain insight and thoughts on timeless topics. Background David Sabatini is a molecular biologist, previously employed as a tenured professor at MIT and a lead researcher at the Whitehead Institute. He was fired in August of 2021, after which most of his professional connections quickly dried up.
It’s Probably Not Lithium — LessWrong
This post has been recorded as part of the LessWrong Curated Podcast, and can be listened to on Spotify, Apple Podcasts, Libsyn, and more. A Chemical Hunger (a), a series by the authors of the blog Slime Mold Time Mold (SMTM) that has been received positively on LessWrong, argues that the obesity epidemic is entirely caused (a) by environmental contaminants. The authors’ top suspect is lithium (a), primarily because it is known to cause weight gain at the doses used to treat bipolar disorder.
Are smart people's personal experiences biased against general intelligence? — LessWrong
TL;DR: Collider between g and valued traits in anecdotal experiences. IQ tests measure the g factor - that is, mental traits and skills that are useful across a wide variety of cognitive tasks. g appears to be important for a number of outcomes, particularly socioeconomic outcomes like education, job performance and income. g is often equated with intelligence. I believe that smart people's personal experiences are biased against the g factor.
Ege Erdil's Shortform — LessWrong
There are two arguments frequently offered for a free market economy over a centrally planned economy: an argument based around knowledge, sometimes called the socialist calculation problem; and another argument based on incentives. The arguments can be briefly summarized like so: A central planning authority would not have enough knowledge to efficiently direct economic activity. A central planning authority would not have the right incentives to ensure that their direction was efficient.
Petrov Day Retrospective: 2021 — LessWrong
I apologize for not posting this closer to Petrov Day. It’s been a busy month and there was much to think about. You can view the EA Forum’s retrospective here. This year was the third Petrov Day celebration on LessWrong in which the site was endangered, and the first year we joined together with the EA Forum. In case you missed it, neither site was taken down, despite 200 people being issued codes that would allow them to do so [1][2].
Ege Erdil - LessWrong
True, but isn't this almost exactly analogously true for neuron firing speeds? The corresponding period for neurons (10 ms - 1 s) does not generally correspond to the timescale of any useful cognitive work or computation done by the brain. Yes, which is why you should not be using that metric in the first place. But even the top-line number is (at least theoretically) a very concrete measure of something that you can actually get out of the system.
What Do GDP Growth Curves Really Mean? — LessWrong
Gross domestic product (GDP) is a monetary measure of the market value of all the final goods and services produced in a specific time period. - Wikipedia, GDP Due to inflation, GDP increases and does not actually reflect the true growth in an economy. That is why the GDP must be divided by the inflation rate (raised to the power of units of time in which the rate is measured) to get the growth of the real GDP.
The Goldbach conjecture is probably correct; so was Fermat's last theorem — LessWrong
I am extremely interested in examples of heuristic arguments like this where the naive version leads you astray and it takes much more sophisticated arguments to fix the problem. I'd be happy to pay a $500 bounty for any examples that I find really interesting, in number theory or elsewhere.
Heads I Win, Tails?—Never Heard of Her; Or, Selective Reporting and the Tragedy of the Green Rationalists — LessWrong
Great post. Something that's worth noting is that this is true *even if reality is not 1:1:1.* Let's say the joke "reality has a well known Green bias" is actually true, and add an extra Green side to that impossible die and make it a true four sided pyramid, easy to picture and shake in your hands. It's now 2:1:1 Green:Blue:Grey Reality says "Roll 10d4," and you let all the little pointy shapes in your palms spill out to reveal... *rolls actual dice* ~4 Green, ~4 Grey, and ~2 Blue.
1960: The Year The Singularity Was Cancelled — LessWrong
[Epistemic status: Very speculative, especially Parts 3 and 4. Like many good things, this post is based on a conversation with Paul Christiano; most of the good ideas are his, any errors are mine.] I. In the 1950s, an Austrian scientist discovered a series of equations that he claimed could model history. They matched past data with startling accuracy. But when extended into the future, they predicted the world would end on November 13, 2026. This sounds like the plot of a sci-fi book.
Show More
loading
Actions
Is this you?
As a journalist, you can create a free Muck Rack account to customize your profile, list your contact preferences, and upload a portfolio of your best work.Get in touch with Ege
Contact Ege, search articles and posts on X, monitor coverage, and track replies from one place.
Learn more about Muck Rack