Radim

42 ms·

361.
by Radim 8y ago
For instance, on IMDb sentiment our method is about twice as accurate as fasttext. Seeing as fasttext accuracy is 90%+, does this mean your method achieves 180%? I'm nitpicking of course, but lately I've seen claims like "2
362.
by Radim 8y ago
No doubt. Being uncomfortably stuck in the middle propels action, which is the only thing able to bring about change. What I was trying to say is that uncomfortable, leaky abstractions are absolutely standard. Is a virus alive or inanimate?
363.
by Radim 8y ago
While technically true, this way of looking at the world feels a little pedantic. Yes, all words and objects and categories are only leaky abstractions. Our universe is a shimmering swirl of unique patterns, all boundaries ultimately fuzzy.
364.
by Radim 8y ago
But what is the solution? Raise taxes for all of us to pay the increased salaries of humanities professors? There is certainly great value in humanities, but that value shouldn't be determined by "an unrelated field is hot, so let
365.
by Radim 8y ago
I agree with most of what you said. (with the caveat that unless you're an expert yourself, common sense and appeal to authority are about all you have) > Freeman Dyson has not done sufficient research to reach the conclusion that
366.
by Radim 8y ago
Accusing Dyson, of all people, of "believing comforting hypotheses"… And why jump to the strawman of "denying climate change" right away? Did you actually watch the interview? You might be feeling righteous zeal, but ide
367.
by Radim 8y ago
Science is about seeking the truth, not "mass" numbers. I'll take a well-informed, nuanced opinion of Dyson, a colleague of Einstein and Feynman, over "The outcome is death, and it’s the end of most life on the planet&qu
368.
by Radim 8y ago
A fascinating talk, and a counter viewpoint from a top physicist: "Freeman Dyson on the Global Warming Hysteria" https://www.youtube.com/watch?v=BiKfWdXXfIs
369.
by Radim 8y ago
The article seems a little rough (draft?), but the money shot is the algo in Section 5. Looks pretty neat and straightforward! TL;DR: 1. Build a special distance(word1, word2) metric using a set of word vectors trained elsewhere (such as Gl
370.
by Radim 8y ago
Or in some cases, a hell of a lot less :-) Little known fact: people from CZ routinely travel to Germany for shopping. It's not just for the cheaper prices (electronics, drugstore), but even where the prices are comparable, for the muc
371.
by Radim 8y ago
Another datapoint for perspective, from the Czech Republic (also an EU country): * PhD stipend: €3.3k / year (untaxed) * median CZ salary: €13k / year, €11k after taxes * data scientist: €30k / year, €22k after taxes (royalty
372.
by Radim 8y ago
Trouble is, it's not really a "market". EU universities are (mostly) subsidized from other people's taxes, so calling for "market salaries" is essentially calling for increased taxation. A tough sell, especiall
373.
by Radim 8y ago
Not just in science! The arms race between bullshit artists, finding shortcuts to extract resources "for free" using the right spin, and the rest of the society, is as old as world. This is even more pronounced in business: sellin
374.
by Radim 8y ago
These RAMM sites seem a hotch-potch of vaguely related buzzwords, blurred images and unrelated company logos. Did you see the laundry list of "Current Customer Use Cases" under https://ramm.science/x/signalbox
375.
by Radim 8y ago
Hey, hey. Some of us actually bootstrapped our companies, effectively paying people (incl. ourselves) from our own pocket :-) For some, the risk/independence trade-off doesn't stop at "VC's bitch".
376.
by Radim 9y ago
Yes, it's almost certain it will be interpreted more leniently. At least to start with. Elizabeth Denham, UK's information commissioner in charge of data protection enforcement, had this to say: "Having larger fines is usefu
377.
by Radim 9y ago
The law is not defined in a "technical" way. It's definitely not about renaming users or deleting their postings. GDPR is declarative: can you identify someone through the records in your possession, whether database or not,
378.
by Radim 9y ago
Curious: How many levels of indirection from end users do you need to be, in order to consider installing an arcane linux tool and then launching `protoc --decode_raw < message.bin` from the command line an acceptable drop-in replacement
379.
by Radim 9y ago
"Switching from experimenting a lot and thinking a little to experimenting a little and thinking a lot was a key turnaround in productivity. When debugging with long iteration times, you really need to pour time into the hypothesis-fo
380.
by Radim 9y ago
Funny thing: do your eyes light up seeing a "scholarly academic paper", or a "deeply tinkering blog"? Which one have we trained our brains to react more strongly to, what is more exciting to you? What do you feel you lea
381.
by Radim 9y ago
What happens when the company changes direction? If there's a shift of priorities, an internal restructuring, a "strategic startup pivot", an acquisition? Not to assume bad faith on Kaggle's part, but we got burned one t
382.
by Radim 9y ago
Optimizations are always great :-) This sounds similar to Maciej Kula's experiments in "Binary Latent Representations for Efficient Ranking: Empirical Assessment" , https://arxiv.org/abs/1706.07479 . Mac
383.
by Radim 9y ago
I feel like I'm not helping you guys much, because my circumstances may have been unique, but the honest answer is: Not much negotiation needed. My uni had what I'd call an "old school AI department", grounded in linguis
384.
by Radim 9y ago
Quit in 2009, age 28. It was not that hard financially -- I moved to southern Thailand, worked on the PhD and Gensim there. Thailand is a pretty cheap country to live actually.
385.
by Radim 9y ago
Employee => academia (PhD in AI) => open source (Gensim) => ML freelancing (radimrehurek.com) => ML consulting company (rare-technologies.com) => ML products (pii-tools.com, scaletext.ai). Looking back, it's hard to imag
386.
by Radim 9y ago
It's refreshing to see such responsible approach. What you suggest is (as far as I understand you) orthogonal to automated data discovery / inventory mapping, though.
387.
by Radim 9y ago
Charming response :-) Entire industry dismissed in a single HN comment. Poof! I'm not sure we understand "data discovery" to mean the same thing, but you reminded me of "How To Draw An Owl": http://sethgo
388.
by Radim 9y ago
These type of SAR requests (even milder ones) are of course impossible to handle manually. Self-assessment , the way most companies decided to handle GDPR, isn't much help here. How do you automate personal data discovery, especially
389.
A Hundred Authors Against Einstein(weeklysciencequiz.blogspot.com)
1 points by Radim 9y ago|0 comments
390.
Decoding thought words straight from brain images(nature.com)
1 points by Radim 9y ago|0 comments

More ›