Profile cover photo
Profile photo
Robert Smart
193 followers -
Grandpa
Grandpa

193 followers
About
Robert's posts

For fans of immortality, we get a preview. Sorry no room for children, and its Federer v Nadal year after year, decade after decade, century after century. [I used to say McEnroe v Borg].

Post has attachment
Machine learning has to be about inference. John Skilling points out (https://pdfs.semanticscholar.org/9a6e/f7c807fb776d9797ffd07014300365bc2c6f.pdf) that the theory of probability is the only sane way to do inference: "Probability calculus is unique: it is the only calculus within which uncertainties about propositions are manipulated consistently with the logical (TRUE/FALSE) status of the propositions. In particular, any posterior inference can be used as a prior when analysing extra data, so that the order in which one uses the data is immaterial (as it should be). Kolmogorov (1950) is widely quoted as the author of the axiomatic basis of probability calculus, but it was R.T. Cox (1946, 1961) who showed that no other calculus is admissible. The only freedom is to take some monotonic function instead, such as 100 Pr(•) (percentage) or Pr(•)/(1 − Pr(•)) (odds), but such changes are merely cosmetic. It follows that other methods are either equivalent to probability calculus (in which case they are unnecessary), or are wrong."

At a quick look it seems that Edward (http://edwardlib.org/) is looking to do Machine Learning the right way.

Post has attachment
The decline of trust in expertise is partly due to he way scientists like to say what is right without directly addressing people who say what is wrong. Peter Woit is the exception that proves the rule.
http://www.math.columbia.edu/~woit/wordpress/?p=9053

[Not that I agree with him that "multiverse" is a useless concept. Urs Schreiber says that even when we squeeze the pips out of mathematical physics there will be parameters that are not determined, and Science will not be able to distinguish between options for why the values are what they are: luck, god, multiverse, simulation. Multiverse will appeal to the non-religious, but it isn't science.]

Post has attachment
Working out how to do ocean fertilization might be the easiest way to pull CO2 out of the air and sea, but it will be hard to land on a dime (if we even knew what the optimal CO2 level was).
https://www.simonsfoundation.org/features/foundation-news/vitamin-b-12-and-a-knockoff-version-create-complex-market-for-marine-vitamins/

I have one of the 2011 macbooks that often die when switching between 2 graphics chips. So I installed Fedora 25, which lets me turn one off. After adding rpmfusion it works really well. Except there is no Acrobat Reader for pdf on linux any more. I usually read pdfs on my phone anyway, but I want a "send this to acrobat reader on my phone" button.

Post has attachment
I very much enjoyed Carlo Rovelli's synthesis of the Philosophy of Science, with some simplified history of Physics, plus a nice exposition of the state of fundamental Physics plus a nice exposition of some aspects of the problems and possible future of fundamental physics. It is a pity about the subtitle which will put many off.
https://www.penguin.co.uk/books/294669/reality-is-not-what-it-seems/

Post has attachment

Post has attachment
1/111 off 11.1 overs
Photo

Post has attachment
Machiavelli on Trump (at the end)...

Post has attachment
Democracy tends to split into 2 parties. They need to stick pretty close to public opinion, still it is the activists from the more extreme wings of the parties that do a lot of the work, and hence have a lot of influence. Between the fossil fuel lobby's hold on the right and the anti-nuclear Greens hold on the left, we seem to be doomed. Here's a typical bit of total rubbish from the Greens [see the comments for the true story]:
https://theconversation.com/what-can-australia-learn-from-germanys-remarkable-energy-transition-69648
Wait while more posts are being loaded