Dilbert panic over…

Thought I was suddenly in an alternative universe when looking at today’s Dilbert cartoon…

Different look Dilbert (7 Mar 16 cartoon)

…but now all is revealed; Scott Adams has gone on holiday and there are guest artists drawing the pictures for six weeks.

http://blog.dilbert.com/post/140211504101/dilberts-changed-look-explained

Tormenting AIs…

Quite a slew of “Robots will take our jobs” articles and AI future death of civilisation civilisation apocalypse FUD recently, so I thought I would undertake a quick investigation of the IBM Watson Developer Cloud which gives access to their cognitive computing APIs.

Watson’s Cookery book was an interesting read and Jeopardy game show appearance was a success…so good to see he is now branching out into more serious domains

Here below is the is the output from an entirely lightweight test of the Natural Language Classifier

We created something like this for Garlik to categorise the links relating to people spidered on the WWW into different topics areas.  It is not easy to make this work in a  useful way…

The sample classifier is trained on weather so lets try a weather question:

Q1 – “Is it raining outside?”

Output…Natural Language Classifier is 98% confident that the question submitted is talking about ‘conditions’.

Yup that seems OK

Let’s try something not about the weather to see if it discriminate out of context topics properly…

Q2 – “Peter Piper picked a peck of pickled peppers; A peck of pickled peppers Peter Piper picked; If Peter Piper picked a peck of pickled peppers,  Where’s the peck of pickled peppers Peter Piper picked?”

Output…Natural Language Classifier is 82% confident that the question submitted is talking about ‘temperature’.

Oh dear, 82% is pretty much a false positive.  Ah well, never mind; still interesting though, but not sure I would pay for that yet!

There are plenty more API examples to play with and, well, I could go on tormenting the AI with more daft questions all day, but real work beckons…

Google patents smart ‘toys’ that could spy on children and control home | Daily Mail Online

This sort of patent concerns me about the overall principle of patenting stuff: 1. Have Google actually made these toys they are patenting or jut an paper idea?, 2. Is this actually a concept worth a patent at all as it does not seem particularly novel and more like an obvious piece of engineering?

http://www.dailymail.co.uk/sciencetech/article-3092988/Google-patents-creepy-internet-toys-control-home-listen-conversations-spy-children.html

3D Printed 6-Story Apartment Building and an Incredible Home – 3DPrint.com

It had to come to pass eventually.  Shame they printed some rather conventional shaped buildings
http://3dprint.com/38144/3d-printed-apartment-building/

Did removing lead from petrol spark a decline in crime?

Did removing lead from petrol spark a decline in crime? http://www.bbc.co.uk/news/magazine-27067615

Posted from WordPress for Android

So it would need 240,000 K supercomputers to simulate realtime in the human brain (how many raspberry Pis?) = maybe brute force computing is not the answer?

Supercomputer models one second of human brain activity – Telegraph   http://www.telegraph.co.uk/technology/10567942/Supercomputer-models-one-second-of-human-brain-activity.html

This suggests that by the law of averages, marketing automation generates no benefit

Uh-oh: One-third of companies adopting marketing automation are not hitting ROI goals http://www.news-republic.com/Web/ArticleWeb.aspx?regionid=4&articleid=20319142

Posted from WordPress for Android

Quantified Self Movement Quackery

#quantifiedself #itsallaboutme I feel one might gain more useful self awareness by meditating on self-improvement than minutely recording every little activity of one’s marvellous existence
http://www.livescience.com/topics/quantified-self/

Posted from WordPress for Android

Perils of technology investment

Two interesting themes underlying this article: 1. Clever technology doesn’t always make a great business, 2. The Innovators Dilemma at work in big corporates strangling the babies , “The smarter news apps are, the dumber they get…” at  http://www.news-republic.com/Web/ArticleWeb.aspx?regionid=4&articleid=19972244

Posted from WordPress for Android

What the Bell?

I was rather interested to see a post on LinkedIn recently about “The Myth of the Bell Curve” which was saying (relatively) recent research had shown that human performance is more like a Power law distribution, than a Normal distribution.

The consequences of this is that a cherished HR sacred cows needs slaughtering.  Anyway you can read the post yourself, however, what tickled my interest is what would the two distributions look like when laid next to each other.

There is an image in the publicity material that attempts to show this…

image

…but that must be mathematically wrong, surely!

Nurse, bring the oxygen!

Both the Normal Distribution and Power Law are both types of probability density functions. however, as far as I can see from the published links, they have different axes:

  • Normal Distribution:  X = performance metric, Y = probability of that performance metric
  • Power Law :  X = some indicator of population; Y = performance metric of some sort

The problem of comparing these two is is that you need to rework the data to get both on the same axes.  Making the hypothesis that the x-axis of the Power Law is the performance rank of an individual – like a Zipf curve equivalent.

So X is not the size of the population, ‘cos that is just absurd:  the curve would otherwise show that for that any population of 1 is really brilliant, whereas the bigger it gets the more stupid it is…mmmm, weelllll, depends on who is counting themselves as the One, and how many of the rest read the Daily Mail/Mirror/Express/Sun/Star…

So if you work the data on that basis (modelling an arbitrary population size of 100 people) then the curves actually look like this…

Power law

…so they are curves with quite different shapes.  And if you re-plot them the other way round, then they look like this…

Normal

…which might superficially look like the picture at the top, but is actually showing the population of the long tail as the tall spike, not top performers.

Still a rather scary picture, as it indeed suggests that most of the people in the “team” are rather serious under-performers, hanging on the coat-tails of the many fewer high-flyers!

This may be a figment of the example data somewhat, and taking a probably unsubstantiated analytical leap, we can readjust the power law chart to align the median figures of performance and come up with a chart like this…

Normal (power adjusted)

…which even still suggests that there are a load of sub-middle slackers sitting on their hands, and they should really get moving and DO SOMETHING!

My general theory that if when leaving the house on the way to work in the morning, you harbour the thought that “today, “I will not make a difference”, go back indoors and get back under the duvet.

So I have scratched my itch, not sure it was so much fun for you, so here is another useful framework to help guide thinking and action and considers the destination of projects…

Thinking is…

..rigorous

A wasted opportunity swirling round the Plug-Hole of Life

Nirvana

..waffly

Hell,
by way of a path of good intentions

Nuremburg

..waffly

..rigorous

Implementation is…