jmpeax

6 ms·

31.
by jmpeax 1y ago
> If the robot is an expert in all fields why would you bother to learn anything? Robots have been better at chess than humans for a quarter of a century. Yet, chess is still a delightful intellectual and social persuit.
32.
by jmpeax 1y ago
The description of the algorithm is frustratingly confusing. "Now pick another random black dot to start from and color it white too. From this black dot" from which black dot, the white one? "single step in a random directio
33.
by jmpeax 1y ago
The tech tree of future games will have us smelt copper to unlock iron.
34.
by jmpeax 1y ago
Yes, the two are orthogonal concepts. Text did not disappear just because we invented photography. Bayesian data analysis is for inverse problems, such as using data to learn about the properties of the system/model that could have gen
35.
by jmpeax 1y ago
Statistical Rethinking by Richard McElreath. He even has a youtube series covering the book if you prefer that modality.
36.
by jmpeax 1y ago
> Leadership in technical environments isn't about being the smartest person in the room. It's about being the most effective translator. Only if you don't make the final decision, but if you do, you better be the smartest
37.
by jmpeax 1y ago
> It makes it ugly to read, and the context switching between code and comments is hard. That's a symptom of bad syntax highlighting, fix it and you're good to go.
38.
by jmpeax 1y ago
Do you yourself really understand, or are you just depolarizing neurons that have reached their threshold?
39.
by jmpeax 1y ago
Maybe because "We propose burying this device beneath the Kerguelen Plateau in the Southern Ocean, 3-5 km into the basalt-rich seafloor and 6-8 km below the water’s surface." would be prohibitively expensive for hundreds of nukes.
40.
by jmpeax 1y ago
Cross-attention before self attention is that better?
41.
by jmpeax 2y ago
You skipped the last section "Tarpits and labyrinths: The growing resistance" of the article.
42.
by jmpeax 2y ago
A transformer will attend to previous tokens, and so it is free to ignore prior errors. I don't get LeCun's comment on error propagation, and look forward to a more thorough exposition of the problem.
43.
by jmpeax 2y ago
I wonder if the second stage failure was related to the metal flap seen here on the very left of the image: https://imgur.com/a/VS8IPdv
44.
by jmpeax 2y ago
What do you mean you don't have to mess around with window sizes? Matrix profile is highly dependent on the window size.
45.
by jmpeax 2y ago
Same here. If you need to lookup how to do something in an api I find it much faster to use chatgpt than to try to search through the janky official docs or in some github examples folder. Chatgpt is basically documentation search 2.0.
46.
by jmpeax 2y ago
So "cut a deal" means to make a deal rather than sever a deal?
47.
by jmpeax 2y ago
What does being rich even mean? For us it might be the total sum of shares, realestate, cash, minus debts. If you're hiding your wealth then what does it mean to "have" that wealth if no one knows it belongs to you? I guess i
48.
by jmpeax 2y ago
"Hey there gorgeous! Do you want me to analyse your tables? ;)"
49.
by jmpeax 3y ago
I wonder if estimating on the logit scale would solve this problem.
50.
by jmpeax 3y ago
Does this work where a website writes its own press releases, then the lazy press publish it, then the website includes what is effectively their own writing on their website as if they're quotes from the press?
51.
by jmpeax 3y ago
You can't forecast the market, all you can do is present an "awesome AI" to suckers and take a slice of their profit with no exposure to their losses.
52.
by jmpeax 3y ago
> Minio is a great alternative to companies mindful of who has access to user data. Of course, AWS claims that AWS personnel doesn’t have direct access to customer data, but by being closed-source, that statement is just a function of tr
53.
by jmpeax 3y ago
What is the significance of "Ig"?
54.
by jmpeax 3y ago
I always thought garbage websites from garbage companies were the only ones to pop up a subscription window. Bradford Morgan White, why did you turn your website into one of those websites with that "subscribe" popup?
55.
by jmpeax 3y ago
How predictive is it really? p-values and confidence intervals of hazard ratios only tell you the probability of comparisons to null effects on hypothetically repeated data. What I really want to know is, under the prior of proportion of PD
56.
by jmpeax 3y ago
Then you must reject all science, as the knowledge gained is based on probabilistic experiments.
57.
by jmpeax 3y ago
> If your sink is leaking, you might call a plumber to fix it. > But you don't really pay the plumber for putting a new gasket in. > In fact, you don't care about the gasket, you care about not having you sink leaking. I
58.
by jmpeax 3y ago
Bayesianism is the only reliable way of creating knowledge, as it is the direct application of the mathematics of probability. All frequentism can do is tell you how poorly your data fits an unrealistic straw-man model.
59.
by jmpeax 3y ago
The benefit of the Bayesian approach is the need to explicitly state a prior distribution, in this case, you would need to state the prior probability that all ravens are black. This forces us to realise that induction does not prove anythi
60.
by jmpeax 3y ago
Raw data examined: Ok 56%, Flawed 18%, Zombie 26% Raw data not available: Ok 97%, Flawed 2%, Zombie 1% Perhaps it's not good to call an unknown as "Ok"? Maybe Carlisle should add his own paper to the mix?

More ›