alexmlamb

6 ms·

1.
by alexmlamb 7y ago
I've used the McDonalds one in Canada and it seemed pretty good. I think the "options" button is partially an upsell, since they want to try to advertise extra stuff.
2.
by alexmlamb 7y ago
Has anyone done an analysis to see what fraction of the wealth generated by silicon valley ends up in the hands of people who happened to own land there? If normal houses cost millions of dollars, I could see it being a sizable fraction. I
3.
by alexmlamb 7y ago
Fake news concern isn't the right place to focus. News type content is low-volume, has a structured sharing model (i.e. people usually share the articles), is long-form, and requires fairly high precision. This is where humans excel
4.
by alexmlamb 7y ago
Many of these fall under the pattern of comedic overexpression or tactical underexpression. For example, "You're really dirty, you literally smell like trash" obviously means "you figuratively smell like trash" but
5.
by alexmlamb 7y ago
Part of the purpose of activism is to exploit current laws for immediate change, but the other part is to create support for changing the laws to provide better protections. The idea that you could disappear completely from anyone using mod
6.
by alexmlamb 7y ago
I've been using Bing primarily and it seems fine. It's stronger in a few areas like image search, customization, and weather. Translation is roughly equal. For code and paper search it's somewhat worse than google but not
7.
by alexmlamb 7y ago
It's having to write the parenthesis after print. No one wants that and it's been by far the largest complaint the community has about python3.
8.
by alexmlamb 7y ago
Yeah sometimes this is called a "subgradient".
9.
by alexmlamb 7y ago
It's interesting that some of his intuitions seem really really good like the idea that you'll get better results by using very large models. On the other hand, specialized hardware before the software works well has always been
10.
by alexmlamb 7y ago
The idea of Six Sigma never made sense to me. I read a book on it, and it basically said that you should optimize a business so that mistakes are extremely rare (the six sigma refers to that point on a N(0,1) gaussian cdf). The idea that y
11.
by alexmlamb 7y ago
It seems weird to me to conflate "advocating the legal right to own technology that can be used to do X" with "doing X", even though obviously you can disagree with both.
12.
by alexmlamb 7y ago
"ridiculous situation in Hong Kong" How is it ridiculous? I can't think of any government, except maybe Canada, that would just allow a part of their country to secede, especially if the movement to do so was explicitly back
13.
by alexmlamb 7y ago
One idea: try using adversarial training (i.e. train the classifier on examples with bounded perturbations which maximize error).
14.
by alexmlamb 7y ago
I think this is a neat idea. Basically, you'd get more "credit" if you're cited by a good paper than if you're cited by a bad paper.
15.
by alexmlamb 7y ago
I don't think it's necessarily like that. The error could make the theorem wrong in all cases. I think it's more that the mathematician has a top-down way of reasoning, where they can see things like "I want to get from
16.
by alexmlamb 7y ago
I think a lot of people see it as a slippery slope. Of course FTP is not so common, but what if at some point they remove the ability to visit general URLs and just let you either google search or go to a page you've already visited.
17.
by alexmlamb 7y ago
I'm just making a point about the abstract argument. I have no idea about whether Homebrew is technically impressive or not, but it sounds like people think that it is?
18.
by alexmlamb 7y ago
Is it just me or does this reasoning not make any sense? "NASA: 90% of our rocket scientists use the A/C you just fixed, but you can't explain how a rocket works so f* off" Presumably what matters is whether homebrew is
19.
by alexmlamb 7y ago
I guess you don't necessarily need data, but if you don't have data than you'd need a simulator, and we'd don't have simulators for most problems that we care about.
20.
by alexmlamb 7y ago
"Studies have shown that journal editors prefer reviewers of the same gender, that women are underrepresented in the peer review process, and that reviewers tend to be influenced by demographic factors like the author’s gender or insti
21.
by alexmlamb 7y ago
I lived in Baltimore for a few years in college. There are definitely some nice things about it, like the aquarium, and a decent choice of restaurants (however this should come with the caveat that the inner harbor has lots of police stand
22.
by alexmlamb 7y ago
(I work at Lyft) I've had it with you Scale.AI people always trying to take credit for Lyft's work. We've been working weekends and nights for years, even the hourly workers have had to take unpaid overtime. All that time,
23.
by alexmlamb 7y ago
Why don't mainstream newspapers (like NYTimes, Economist, etc) include a list of citations at the end of articles the way that scientific papers do? It seems like this would be easy to do, and would help to restore a lot of trust, bec
24.
by alexmlamb 7y ago
It's not exactly a quick fix, but how many generations would it take to breed cats with substantially softer claws? This could be an interesting program perhaps for the government to undertake. I also wonder if crispr or embryo select
25.
by alexmlamb 7y ago
Just to be clear, this doesn't necessarily mean that the students got the degree through fraud or low standards. Speaking and listening can be harder than reading and writing, so the students might be doing good work even if their spok
26.
by alexmlamb 7y ago
It seems like this can easily be fixed by having people sign up for contracts that last more than 1 year or at least have some financial incentives that accrue to employees who stay longer? For example, tech companies have stocks that only
27.
by alexmlamb 11y ago
Well, I think that the hype is one-directional because the default assumption is that machines will lose to humans in tasks that involve reasoning. We know that there's no computer that can learn to put away the dishes after watching
28.
by alexmlamb 11y ago
I don't think that this is quite the right explanation. |y - wx| isn't differentiable but it is subdifferentiable. So the derivative is defined at every point except where y = wx, and in that case you're more or less fine if