Interesting Links for 16-04-2019

[andrewducker]

Jon Snow Emails 23andMe About His DNA Results

(tags:gameofthrones genetics spoilers viaSwampers funny )

Very interesting thread on the Notre Dame fire and the difficulties in fighting it

(tags:fire catholic religion architecture )

MIT Researcher Exposing Bias in Facial Recognition Tech Triggers Amazon's Wrath

(tags:amazon faces ai race bias )

Half of UK's electricity to be renewable by 2025

(tags:renewables uk electricity thefuture )

360 photo of Notre-Dame de Paris

(tags:photo Paris religion architecture viaElfy )

Notre Dame Cathedral will never be the same, but it can be rebuilt

(tags:architecture history Paris france )

Breakthrough diabetes drug cut kidney and heart disease deaths by a third

(tags:diabetes medicine GoodNews )

AI Helps Classify Lung Cancer at the Pathologist Level

(tags:ai cancer diagnosis )

In the 1960s, 128 kilobytes weighed 610 pounds

(tags:memory computers history )

Why software projects take longer than you think - a statistical model

(tags:estimating software projectmanagement statistics )

Comments

[naath]

It's OK Jon Aegon, the Targaryens are much less hung up about incest than the Lanisters.

Why software projects take longer than you think - a statistical model

[jack]

Huh. Yeah, that makes a lot of sense. I've pointed out something similar but less rigorous. If you have ten tasks you each expect to take a week, the absolute best possible "pleasant surprise" you might get for one of them is taking nought weeks, and that's not going to happen most of the time. But if you say say, "out of these ten, what's the chance one of them has an unforeseen complication that ends up taking like 10 weeks", it's... fairly likely. So, everything is late, always.

When I'm doing stuff myself, I've always fallen into these traps, but have slowly got better at admitting when something is uncertain and likely to be lots more complicated than I think, and re-evaluating what's still worth doing and what isn't.

And ideally project scheduling would focus on these sorts of estimates. But it seems like, even if it's not what any one person wants, there's a lot of pressure on project planning to produce "reasonable looking" estimates that don't have realistic "things take longer than we think, 2x time" estimates in.

Adding a variance field to estimates, even one just left at a default value for 99% of tasks, would help a lot, but no-one does that.

Re: Why software projects take longer than you think - a statistical model

[rmc28]

I'm planning to share that one with my boss, as estimates vs reality are a bit of a thing at work at the moment.

Re: Why software projects take longer than you think - a statistical model

[danieldwilliam]

Me also.

[calimac]

"In the 1960s, 128 kilobytes weighed 610 pounds" is even more startling if you initially misread it as kiloGRAMS.

[andrewducker]

I wonder if there's an SF story in that somewhere. The idea that the conversion rate of imperial vs metric units hase changed over time, as if they were two currencies.

[calimac]

There is a fringe scientific theory that universal constants, like the speed of light and perhaps gravitation, have changed over time.

[danieldwilliam]

A couple of options come to mind.

You do have some change in the actual values of imperial units over time. Tritely the length of the king's foot being different from reign to reign. Metric units seem more closely tied to the physical qualities of the universit.

So, option 1, metric units remain the same, imperial units change with hilarious consequences when eg space landers crash in to the moon.

Option 2, the metric units are actually changing, implying that the fundamental qualities of the universe are changing - but very, very slowly so it's difficult to notice. It looks like option 1 but isn't.

Option 3 - there is divergence and the state of the universe depends on your cultural affiliations.

About the Insurance Journal article...

[dewline]

...I find the concept of an Algorithmic Justice League deeply pleasing to my heart.