EmbarrassedFuel

"London obviously continues to be heavily subsidised by the rest of the UK"

This is a farcical comment. Were you being sarcastic? The tax revenue from London massively subsidises the rest of the UK. The investment happens in London because you can guarantee it will make a return, and quickly.

Sounds interesting, would you be able to share anything about the signal processing tricks?

Don’t take everything at face value: a fall in Sterling necessarily means that an exporters products are cheaper overseas, so should boost sales. Unless their costs were already higher than their revenues, in which case they were doomed anyway. The reason brexit will be bad for exporters is the possible future imposition of tariffs. This hasn’t happened yet, so actually in the short term brexit is relatively good for them. Just to be clear I’m against brexit, however in this case blaming brexit for the failure is wrong.

Let me guess, Cambridge to KingsX to work at Google?

Because CAM doesn’t even attempt to be correct.

I know it’s hard to believe and to generalise beyond your own experience but you are one data point. You might have gotten better without any of the things that you mentioned. Most likely it was a placebo effect, strengthened by your own research and belief in the practices.

I’m not sure the lack of government regulation is a valid point. In my experience, and as you’d see if you read Goldacre’s book, government administrators don’t understand statistics. Even if they did, they don’t neccesarily legislate to maximise welfare.

There have - convolutional nets have far fewer weights to learn than an equivalent fully connected net of the same depth.

And you literally described it as a brute force approach in your comment.

Do you have anything better than an appeal to tradition to justify your viewpoint, perhaps a book or study that comprehensively justifies TCM without falling into the basic traps of failed statistical reasoning?

Which ones? Google has a vast array of products from enterprise grade cloud services all the way to small consumer applications.

Are you really claiming that TCM and other such pseudosciences have been a net positive for the world? Unproven treatments that may make problems worse, while also encouraging patients to stay away from tested treatments, should play no part in a modern society. If you're in any doubt about this, perhaps you should read 'Bad Science' by Ben Goldacre, which takes a whistle stop tour of such quackery, as well as the various unethical practices undertaken by pharmaceutical companies.

The slowest is three times slower than the fastest! That’s not almost identical at all, and could easily be the difference between a useable interface and an unbearable one...

A small bone of contention: it’s not about Moore’s law per se, which has been ‘dead’ since about 2013, coincidentally when the deep learning revival started. It’s matrix multiplication ASIC development that is driving the progress.

GPUs already existed when the idea to use them to make the feasible size of neural nets larger came about. For a long time the drive for the increase of GPU compute power was still gaming/commercial graphics houses. It’s really only in the last 1-2 years that we’ve seen highly specialised GPUs with features like tensor cores (or indeed google’s TPUs).

Also, calling neural nets ‘brute force’ because they use a lot of computing power to train a model is slightly reductionist - a true brute force approach to image recognition, ie enumerating all possible combinations of, say, 200x200x256x3 pixels, would be completely absurd and probably exceed the computing power currently available on earth.

Your ability to recognise faces, you surely realise, was passed down to you genetically? It’s encoded as hardware priors in the structure of your brain. So it’s not too hard to extend that to knowledge of a smell, similarly hardcoded somewhere. I’m not defending the original paper btw, which seems to have some faults.

Don’t forget the stochastic control problem that needs to be solved to stop the whole grid blowing up!

Me too! Although it has been sitting in my list of potential projects for months so kudos to them for actually doing it...

Blatant manipulation of the findings of the study in order to further the aims of what looks like some kind of crypto-rag. From the abstract of the paper:

“indicating that (with the exception of aluminium) cryptomining consumed more energy than mineral mining to produce an equivalent market value.”

In reference to the other metals studied. This is neglecting the even more obvious point that it’s a completely irrelevant comparison, we should compare to equivalent alternatives!

Ah yes point taken: I was implying that the author had used this (nowadays) non-standard construction in order to add some sensationalism.

As a Brit, no one has made the million/milliard distinction for my entire life. 1000 million is a billion as far as anyone is concerned, a million million is a trillion. It’s just comically bad writing.

The bin removal was actually prompted by a period when the IRA would put bombs into bins in train stations etc., see eg

https://en.wikipedia.org/wiki/Warrington_bomb_attacks

So not a completely idiotic policy.

In the uk I often see posts for graduate programmer positions with essentially this arrangement: words to the effect of “candidates need not have experience coding but should show a willingness to learn and deep interest in tech” etc.

This is how tanks steer! And also (to some extent) why using them on roads ruins the roads. If you tried in a traditionally wheeled car then all of the drive shafts would snap though!

In addition to the sls this idea is deployed to some extent in IC cars as well: electronic limited slip differentials can send continuously varying torque to each wheel, see the Porsche 918 spyder (combo electric and IC) or the Nissan GT-R which has a famously odd driving characteristic as a result of the computer controlled torque vectoring.