tmnvdb

5 ms·

1.
by tmnvdb 1y ago
Citation needed. As far is I know this is simply false. Different sanctions have different goals. Regime change is very rarely a goal. Often it is to reduce economic growth to keep/make the country weak, or to achieve some other goal.
2.
by tmnvdb 1y ago
Did you read the article?
3.
by tmnvdb 1y ago
If you train an LLM on chess, it will learn that too. You don't need to explain the rules, just feed it chess games, and it will stop making illegal moves at some point. It is a clear example of an inferred world model from training.
4.
by tmnvdb 1y ago
You have again answered with your customary condescension. Is that really necessary? Everything you write is just dripping with patronizing superiority and combatative sarcasm. > "classical reasoning uses consciousness and awareness
5.
by tmnvdb 1y ago
> This is not a demonstration of a trick question. It's a question that purposefully uses a limitation of the system. There are many such questions for humans. They are called trick questions. It is not that crazy to call it a trick
6.
by tmnvdb 1y ago
> No, it's the entire architecture of the model. Wrong, it's an artifact of tokenizing. The model doesn't have access to the individual letters, only to the tokens. Reasoning models can usually do this task well - they can
7.
by tmnvdb 1y ago
You use a lot of anthropomorphisms: doesn't "know" anything (does your hard drive know things? Is it relevant?), "making things up" is even more linked to continuous intent. Unless you believe the LLMs are sentient
8.
by tmnvdb 1y ago
So similar to wikipedia
9.
by tmnvdb 1y ago
PPDF is a great book but hard to apply. I recommend looking at some Kanban literature. Classic in this space is Actionable Agile Metrics for Predictability.
10.
by tmnvdb 1y ago
It is precisely to reduce cycle time that we control queue size. It's also not entirely true that cycle time is purely lagging. Every day an item ages in your queue, you know the cycle time had increased by one day. Hence the advice to
11.
by tmnvdb 1y ago
If it's really important to have an accurate estimate for a large work package you are in trouble, there is no such thing.
12.
by tmnvdb 1y ago
It's not just cruel, it's stupid. Not only does cycle time form a very poor measure of individual productivity, using cycle time to measure individuals will create very bad incentives that will make your team perform significantly
13.
by tmnvdb 1y ago
You've misunderstood: cycle time is neither a forecast nor a measure of individual productivity. Cycle time measures how long it takes for a unit of work (usually a ticket) to move from initiation to completion within a team's wo
14.
by tmnvdb 1y ago
I've never encountered cycle time recommended as a metric for evaluating individual developer productivity, making the central premise of this article rather misguided. The primary value of measuring cycle time is precisely that it cap
15.
by tmnvdb 1y ago
I understand from your question you struggle to comprehend that this is possible. I assure you it really is. People who have money take the train. People who own cars take the train. The modal split for Vienna generally is about 25% by car.
16.
by tmnvdb 1y ago
Most people would not. Paris - Berlin is dominated by flying.
17.
by tmnvdb 1y ago
This really is nonsense but somehow every time this topic comes up people being it up. The size of the country or its population density is not really relevant. People in Europe dont take a train from Greece to Sweden. They fly. In fact mos
18.
by tmnvdb 1y ago
This is true in the US but not a law a nature. Its the result of policy. There are whole cities built from scratch (outside the US) within the last 70 years that did not choose this model. And there are many new developments in older citie
19.
by tmnvdb 1y ago
Wild assertions with another wild assertion as 'evidence'
20.
by tmnvdb 1y ago
I live in Vienna and people take public transport to nature all the time .
21.
by tmnvdb 1y ago
It's obviously nonsense. Nobody is walking from Paris to Berlin. But you can walk in Paris and Berlin.
22.
by tmnvdb 1y ago
This seems like a strangely theoretical argument, as if the world outside America simply does not exist.
23.
by tmnvdb 1y ago
The current American urbanism is from the past! The assumption that other urbanisms somehow represent a blast from the past, while 70 year old American car-centric urbanism embodies the eternal modern 'now,' simply doesn't
24.
by tmnvdb 1y ago
What are these "challenging geographic features"?
25.
by tmnvdb 2y ago
It means that different types of good (and bad) behaviour are somehow coupled. If you tune the model to behave bad in a limited way (write SQL injection for example), other bad behaviour like racism will just emerge.
26.
by tmnvdb 2y ago
It's linked in the twitter thread from the authors: https://x.com/OwainEvans_UK/status/1894436637054214509 specifically: https://x.com/ESYudkowsky/status/1894453376215388644
27.
by tmnvdb 2y ago
They have an app where you can browse the examples: https://emergent-misalignment.streamlit.app/
28.
by tmnvdb 2y ago
"I wouldn't have called this outcome, and would interpret it as possibly the best AI news of 2025 so far. It suggests that all good things are successfully getting tangled up with each other as a central preference vector, incl
29.
Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs [pdf](martins1612.github.io)
179 points by tmnvdb 2y ago|100 comments
30.
by tmnvdb 2y ago
Isn't this something the US has had on steroids for many years? I.e. Patriot act, PRISM, FISA, national security letters?

More ›