3,538 karma · joined October 3, 2010
Reach me by email: sthatipamala@gmail.com
YC Badge: 0xa18d1d3604177ac278ed284fc251bb9330f31286
5 cars worth of carbon emissions is not a lot given that it is a fixed cost. Very few are retraining BERT from scratch.
EDIT:
The other two points are also disingenuous.
* "[AI models] will also fail to capture the language and the norms of countries and peoples that have less access to the internet and thus a smaller linguistic footprint online. "
NLP in "low resource" languages is a major area of research, especially because that's where the "next billion users" are for Big Tech. Facebook especially is financially motivated to solve machine translation to/from such languages. https://ai.facebook.com/blog/recent-advances-in-low-resource...
* "Not as much effort goes into working on AI models that might achieve understanding, or that achieve good results with smaller, more carefully curated datasets (and thus also use less energy)."
This is also a major area of research. Achieving understanding falls under the purview of AGI, which itself carries ethical and safety concerns. There are certainly research groups working toward this. And reducing parameter sizes of big networks like GPT-3 is the next big race. See https://news.ycombinator.com/item?id=24704952
I have no particular knowledge of the effects of these visuals but I find that they make meditation more enjoyable. It gives me something engaging to hold on to while sitting in silence.
If you want to try it, download the One Giant Mind app. It’s free and I’m curious if it happens for anyone else.
I had classmates who blew it off as a useless course because they already knew Java and PHP. It was their loss; they don't have the same fascination with computer science and I can directly attribute it to not getting exposed to Scheme/Lisp.
The projects themselves are fairly intense and, if you take 2 courses, their deadlines will likely overlap.
Highly recommend it if you’re actually interested in going deep into the material. It’s a lot of work and I’m learning a lot even after several years of industry experience.
Also, you can do feature engineering on a training set that you collect, as long as that is similar to the distribution of end user inputs. That's a pretty standard ML workflow.
What should I use if want to send dead simple, foolproof encrypted email? (Foolproof in that I can trust that updating the program will patch vulnerabilities and incorporate the state of the art methods)
This is not the 12th iteration of Stephen Wolfram, the human ;)
Source code does not store any performance metadata, just the static score. In that sense, it’s more important to get a feel for theory and aesthetics of programming.
0: https://www.amazon.com/Taking-Work-Out-Networking-Connection...
The logic for web "applications" these days can be complicated and involve loading hundreds of such files. If loaded individually, this requires a lot of HTTP requests and results in a slow page load time.
Bundling is similar to a static linker in that it produces a single binary out of all its dependencies. This can be loaded with a single HTTP request, cached, zipped, whatever.
New York oxen (that) New York oxen bully (themselves) bully New York oxen.
That in itself is amazing to me considering that the tech company I work at, lots of things are just done by fiat.
I suppose that is also a benefit (maybe some resource can divert to another resource under heavy load/when its DB is unavailable). But quite frequently I want to answer this question with just the source code.