matusp

15 ms·

241.
by matusp 1y ago
> Is essentially equivalent to... I think you overestimate how many programmers would find this explanation clear. Most programmers are not used to functional programming, an anything that manipulates with functions is not that intuitive
242.
by matusp 1y ago
I don't think that US conservatives and liberals are were the center of gravity of papal policy is.
243.
by matusp 1y ago
Do rules of war apply here? Israel does not even consider Palestine to be a state.
244.
by matusp 1y ago
Additionally, it will probably be confusing that it is called a t-string but it is actually a constructor for a Template object and not string at all. I would rather see a new special term `template` than this.
245.
by matusp 1y ago
I have the privilege of working with some physicists every day, and it's definitely not only boomers that do not give a damn about software engineering practices. "If it works now, it's good" mantra is pretty popular eve
246.
by matusp 2y ago
It also did not ask about my cookie policy, how am I supposed to tell the site my preferences? I just wrote them via email to make sure.
247.
by matusp 2y ago
I agree. I enjoyed her physics content. Then she made a few videos about AI, a field where I have more expertise than her, and I found her opinions very shallow and without much nuance. Ever since, my trust in her other videos went down as
248.
by matusp 2y ago
But the signals used to filter out human garbage are not the same the signals that would be needed to filter LLM garbage. LLMs generate texts that look high-quality at a glance, but might be factually inaccurate. For example, an LLM can gen
249.
by matusp 2y ago
I don't think there is anybody that is making significant amount of money by selling tokens right now.
250.
by matusp 2y ago
Do you hire remotely worldwide or US only?
251.
by matusp 2y ago
To be honest there were several chatbots before ChatGPT, eg Galactica by Microsoft, and it had not seem so promising. Similarly, ChatGPT was originally released as more or less a fun little demo, and the absolute success was surprise for Op
252.
by matusp 2y ago
It is not absurd. From manager's point of view, when they are deciding to let someone go or promote them, they often need some arguments. And since software is invisible, it is often hard to just say straight away who is over/unde
253.
by matusp 2y ago
I am still waiting to see the impact on GDP or any other economic measure.
254.
by matusp 2y ago
It is definitely possible and it is actually a really serious security risk. Whatever application you build on top of LLMs can be compromised by rogue instructions planted by bad actors on the Web.
255.
by matusp 2y ago
And there is also hundreds of years of historical experience where establishing buffer states worked.
256.
by matusp 2y ago
Is having a conflict in Ukraine somehow making Europe more stable? To be entirely honest, from realpolitik point of view what Trump is doing makes cold and calculated sense. He will sacrifice Ukraine so that it will become a rump buffer sta
257.
by matusp 2y ago
In Europe we of course use "royal"
258.
by matusp 2y ago
Surmountable, yes, but in practice it is often just too much hassle. If you are doing tons of these tests you can probably afford to invest in the infrastructure for this, but otherwise AB is just so much easier to deploy that it does not r
259.
by matusp 2y ago
I assume that o3 can run Python scripts and observe the outputs.
260.
by matusp 2y ago
Well put. You ask LLMs about ARC-like challenges and they are able to come up with a list of possible problem formulations even before you show them the input. The models already know that they might expect various object manipulations, sym
261.
by matusp 2y ago
I don't think this is only about efficiency. The model I have here is that this is similar to when we beat chess. Yes, it is impressive that we made progress on a class of problems, but is this class aligned with what the economy or th
262.
by matusp 2y ago
Welcome to the today's AI research. There are tons of papers like this and I believe that the AI community should be much more thorough in making sure that this wishy washy language is not used that often.
263.
by matusp 2y ago
I have seen a bunch of tokenization papers with various ideas but their results are mostly meh. I personally don't see anything principally wrong with current approaches. Having discrete symbols is how natural language works, and this
264.
by matusp 2y ago
Turns out that money is still the strongest alignment signal.
265.
by matusp 2y ago
Deep learning is easy to adapt to various domains, use cases, training criteria. Other approaches do not have the flexibility of combining arbitrary layers and subnetworks and then training them with arbitrary loss functions. The depth in d
266.
by matusp 2y ago
Dude, you are talking total nonsense. What does 100k samples even mean. We have not even established what task we are talking about and you already know that you need that many samples. Not to be offensive, but you seem like the type of guy
267.
by matusp 2y ago
Yes, in a sense that if you have infinite appropriate dataset and compute. No, in a sense what is practically achievable.
268.
by matusp 2y ago
My hot take: embeddings are overrated. They are overfitted on word overlap, leading to both many false positives and false negatives. If you identify a specific problem with them ("I really want to match items like these, but it does n
269.
by matusp 2y ago
It also how you define "fake", eg in machine learning it can be everything from fabricated results to data leakage, low quality data, bugs, p-hacking, weak baseline design, insufficient handling of randomness, etc etc.
270.
by matusp 2y ago
I agree with you. What is also telling is that there is no particularly strong reaction from the physics community that someone obvious was wrongfully omitted.

More ›