Not that sort of….

News Feature: What are the limits of deep learning? | PNAS

 In addition to its vulnerability to spoofing, for example, there is its gross inefficiency. “For a child to learn to recognize a cow,” says Hinton, “it’s not like their mother needs to say ‘cow’ 10,000 times”—a number that’s often required for deep-learning systems. Humans generally learn new concepts from just one or two examples.

There is a nice review on Deep Learning in PNAS. The spoofing referred to, is an ‘adversarial patch’ — a patch comprising an image of something else. In the example here, a mini-image of a toaster confuses the AI such that a very large banana is seen as a toaster (the  paper is here on arXiv — an image is worth more than a thousand of my words).

Hinton, one of the giants of this field, is of course referring to Plato’s problem: how can we know so much given so little (input). From the dermatology perspective, the humans may still be smarter than the current machines in the real world, but pace Hinton our training sets need not be so large. But they do need to be a lot larger than n=2. The great achievement of the 19th century clinician masters was to be able to create concepts that gathered together disparate appearances, under one ‘concept’. Remember the mantra: there is no one-to-one correspondence between diagnosis and appearance. The second problem with humans is that they need continued (and structured) practice: the natural state of clinical skills is to get worse in the absence of continued reinforcement. Entropy rules.

Will things change? Yes, but radiology will fall first, then ‘lesions’ (tumours), and then rashes — the latter I suspect after entropy has had its way with me.

Talking 22nd Century Skills with @realpbanksley – Rick Hess Straight Up – Education Week

I noted that he seems to be one of the leading thinkers in the push to rebrand STEM as STEAMED (Science, Technology, Engineering, Arts, Math, and Everything Delightful).

Annual Review of the ‘business’ that is ed-tech  by Audrey Watters.

Getting too deeply into statistics is like trying to quench a thirst with salty water. The angst of facing mortality has no remedy in probability. Paul Kalanithi, ‘When Breath Becomes Air’

A beginner’s guide to student loans in the public accounts | Wonkhe | Analysis 

Bluntly, the main motive for replacing the teaching grant by loans is an accounting trick. There is an apparent decline in public spending, but at the cost of distorting higher education policy … Thus the changes look like a dodgy [Private] Finance Initiative” – Barr, 2012

Well written piece on the loan scandal in Wonkhe by Nicholas Barr. In the language of the laymen, the government is fiddling the books, and dumping the costs on future taxpayers. It fiddles because it wants to mislead, for gain.

He goes on:

higher education finance has elements of a bubble. If I were a Vice-Chancellor, this aspect would give me sleepless nights.

Guarded language — fair enough — but it is not just a financial bubble. Let us just see how this year pans out.

“as long as they keep asking the wrong questions, the answers really don’t matter”.

Thomas Pynchon

A well argued and evidence based article like this will get you nowhere. This is Britain. Better to put some bollox on a bus.

A comment from the ‘Swedish Chef’ on the FT.

She crossed to his desk and shook his hand. Noticed the telltale transplant plugs dotting his scalp, sprouting hair like little tufts of yellow grass in a last desperate stand of virility. That’s what you deserved for marrying a trophy wife.

[from Body Double; Tess Gerritsen]

“In 1968, each candidate could be heard without interruption on network news for 42.3 seconds. By 2000, the length of a sound bite was 7 seconds.” Lawrence Freedman, Strategy: a history. (via John Naughton)