I grew up in Brooklyn and supermarkets didn't really become a thing here until the 2000s. Like there was Key Food, but that felt more like a general store. There was a Waldbaum in the area but it was pretty inconvenient to get to.
Instead we had fruit stands selling produce, a bakery, a butcher, fishmonger selling seafood. And this isn't mentioning all the ethnic food stores. All of this was on two or three short blocks. The first time I saw a supermarket was when we went to New Jersey. It was a Wegmans and I think what was amazing to me was just how much stuff a single place could have.
I think ultimately it does come down to convenience. I think it's easier than ever to have decent meal at home. You don't even have to peel your own carrots or tear your own lettuce anymore. Sure something with more complexity might take time, but that's better for a weekend.
Thanks for using it! Best way currently is to keep using and reporting issues on [1]. You can see how the models are trained on [2] and file issues there too.
I love everything about what you have planned. Is there anything in the works for creating more keyboard options? While mechanical keyboards might be too impractical, even something with bigger arrow keys would be nice.
The day there is a ThinkPad style keyboard (trackpoint, 3 button) and a matte screen, I'll order one immediately, and assuming it lives up to the expectation, I'd order more after that quickly.
That framing actually stops holding for continuous distributions. Like in a distribution on heights of people there isn't a probability or prior associated with a particular height. Probabilities are assigned to subsets of this distribution (sometimes called sigma algebras). Priors stands for prior distribution and it really is best to see Bayesian analysis as a machine that consumes probability distributions as input and spits out probability distributions as output.
> That framing actually stops holding for continuous distributions.
No, it doesn't. A continuous distribution is just a way of assigning probabilities to a parameterized set of propositions where the parameters are continuous and so the set is infinite. But for any given proposition in that set the prior is a number.
> Priors stands for prior distribution
No. Bayes's theorem is:
P(H|E) = P(E|H) * P(H) / P(E)
The "prior" is P(H) which is a probability, i.e. a number between 0 and 1, as are all the other "P" values. The "*" and "/" operations in Bayes's theorems are scalar multiplication and division.
> No, it doesn't. A continuous distribution is just a way of assigning probabilities to a parameterized set of propositions where the parameters are continuous and so the set is infinite. But for any given proposition in that set the prior is a number.
A continuous distribution does not assign probabilities to each proposition but subsets of them. To see this concretely consider the continuous Uniform distribution from 1 to 1.5. It will for all values of its support have a probability density of 2. Most people would not consider 2 a probability.
For continuous distributions, Bayes's theorem becomes about probability densities and not probabilities.
> A continuous distribution does not assign probabilities to each proposition but subsets of them.
Nope. It assigns probabilities to individual propositions.
> Most people would not consider 2 a probability.
That's true, but in your example 2 is not a probability but a probability density, and a probability density is not the same as a probability. To get a probability out of a probability density you have to integrate. For each possible interval over which you could integrate there is a corresponding proposition whose probability of being true is exactly the value of the integral.
No, propositions are simply statements for which is it meaningful to assign a truth value. It might be possible to come up with a proposition that is associated with a non-measurable set, though I can't offhand think of an example. But remember: Bayesian probabilities are models of belief. Priors are peronsal. And so in order to assign a prior to a proposition, the statement of the proposition must have some referent in your personal ontology. You can't hold a belief about the truth value of a statement unless you know (or at least think you know) what that statement means. So unless a person has something in their ontology that corresponds to a non-measurable set (and I suspect most people don't) then that person cannot assign a Bayesian prior to a statement associated with a non-measurable set. But that's a limitation of that particular individual, not a limitation of Bayesian reasoning. For example: you cannot assign a Bayesian prior to the statement, "The frobnostication of any integer is even" because you don't know what a frobnostication is.
(Note that there are all kinds of ways that statements can fail to be propositions. For example, you can't assign a Bayesian prior to the statement: "All even integers are green" despite the fact that you know what all the words mean.)
Why do you feel politicians and tech executives like to say automation threatens to do all these things? How do they benefit from this narrative that has zero evidence to support itself?
In the tech case, there's a definite trend of overpromising, which wins funding, contracts, and so on, and then underdelivering. There's a few different examples in the book where half-baked 'automated' systems are rolled out, and then humans are placed under increased pressure to fill in the gaps and make things work.
In the politicians case, I think there's probably a mix of motives - being seen to be aware of technological change, forward-thinking, etc. But this gives too much credence to industry promises. The threat creates initative to study the effects of automation, but often at a 10,000 foot view, overlooking the specifics of labor conditions, technological adoption, and the workers themselves (race, gender, etc).
While often ignored, item-to-item similarity recommenders work really well for new items and have been the basis for Amazon's and YouTube's early recommendation algorithms
I think you may be misunderstanding the parent (I may also be misunderstanding though) - they're saying that with a new user (not a new product), you have no information for providing recommendations to them; you have no idea what they're interested in.
I was talking about new products/movies/whatever, not users. Sorry, "new entity" was too vague in retrospect.
My point was around the utility of marketing, at least in theory. If I release a new brand of baked beans, how would a recommender based on past sales (and combinations of sales) know to recommend it to anyone. Marketing allows it to go up the list (unfairly, according to the post I was replying to).
Ah, okay. Then yeah, the original reply is valid. I assumed you meant a new user to a site, since you can't recommend anything then no matter what you've built (without user tracking that can maybe have built up some kind of profile from elsewhere).
Let's not create a false dichotomy between mouse control and voice control. There are other alternatives that are arguably less handicapping than being reduced to voice commands.
Maybe moral obligation is the wrong phrasing. But if you are major company with investors, perhaps a fiduciary responsibility to your shareholders and customers to fund the software?
The correct model likely looks something like a foundation to support core infrastructure projects in the open-source ecosystem. This foundation is then funded by companies most dependent on that infrastructure.
locate is, for whatever reason, tragically slow. The database format it uses is nonsensical and completely optimized for size on very outdated assumptions.
I use an implementation I have written in the shell itself whose database format is nothing more than every file path on the system separated by null bytes, that is simply grepped to find files; the speed difference is absurd.
—— — time locate */meme.png
/storage/home/user/pictures/macro/meme.png
real 0m0.885s
user 0m0.806s
sys 0m0.010s
—— — time greplocate /meme.png$
/storage/home/user/pictures/macro/meme.png
real 0m0.089s
user 0m0.079s
sys 0m0.011s
This implementation is highly naïve and simplistic, and offloads all the searching to GNU Grep, yet outperforms the actual `locate` command by an order of magnitude.
And plocate is yet orders of magnitude faster than GNU grep. :-) And updates its database faster. You don't specify which locate you're using, but mlocate and BSD locate are basically obsolete by now.
Thank you! It's becoming the default now, slowly (e.g., it will be the default in Debian and Ubuntu from the next releases, and Fedora is in a process to make it replace mlocate right now). It's just a tad too new, only about a year since 1.0.0. :-)
They are simply em-dashes, if an error should occur, they are replaced with numbers to indicate the error code; they are also color-coded to indicate when I'm not for instance in an SSH session by changing to another color:
—— — true
—— — false
—— 1 sh -c 'exit 120'
120 sh -c 'exit 20'
— 20
Instead we had fruit stands selling produce, a bakery, a butcher, fishmonger selling seafood. And this isn't mentioning all the ethnic food stores. All of this was on two or three short blocks. The first time I saw a supermarket was when we went to New Jersey. It was a Wegmans and I think what was amazing to me was just how much stuff a single place could have.
I think ultimately it does come down to convenience. I think it's easier than ever to have decent meal at home. You don't even have to peel your own carrots or tear your own lettuce anymore. Sure something with more complexity might take time, but that's better for a weekend.