Profile cover photo
Profile photo
Jan Moren
18,837 followers -
Janne. In Osaka. And Okinawa.
Janne. In Osaka. And Okinawa.

18,837 followers
About
Jan Moren's posts

Post has shared content
tl;dr: Most prostate cancers never threaten your life; and removing them early does not increase your chance of survival.

Two quotes:

"Autopsy studies reveal prostate cancer in up to 40 per cent of men in their forties and 65 per cent in their sixties, but a much smaller figure of 3-4 per cent of Australian men actually die of prostate cancer at a median age of 82."

and

"[...]men with early-stage abnormalities of the prostate who do not undergo surgery or radiation treatment, but whose condition is monitored for any progression of the cancer, live just as long as men who opted for complete removal of the prostate and now live with its immediate consequences,[...]"

So, only a small fraction of all early stage prostate cancers are ever even detected. They just sit there, never growing or spreading in any way. So, if you simply monitor them for changes, and only begin active treatment for the small minority that start growing, you get just as good outcome as if you had started treatment right away.



Post has attachment
Saturday.
Photo

Post has shared content
Learning locomotion across varied terrain. It does highlight both why machine learning is cool, and how it's still rather limited.

Det här var riktigt tufft. Deep Mind har använt AI för att lära en dator att springa hinderbanor i en simulerad miljö. Resultatet är både imponerande och lustigt.

Post has attachment
Blog: Mould, the bane of our existence.

Post has attachment
3d-printing in the micrometer range.


Post has attachment
Saturday. 
Photo

Just unsubscribed from my last research-related mailing list. I no longer have any subscriptions, memberships or anything else related to research.

It feels good; making a final break with my previous career. Or, well, almost - I'm still writing a small text about the last project I was working on. That one should be done soon enough, though.


Post has attachment
Inspirobot ( inspirobot.me ) is an automatic inspirational poster generator. And honestly, they're better than 90% of the real ones posted on the net.

I really feel this is something I can aspire to:


Post has attachment
Preliminary experimental results show that kinds learning to program using a language (Scratch) localized in their native language learn more and better than those that use the English version.

This shouldn't be too unsurprising of course. It also lends more credence to the idea of using your native language for comments, variable names and so on when your intended audience allows you to.

Also, for all its faults (all its many, many faults. Many, many, many, many...), Excel did get one thing right: all functions, operators, number formats and so on are translated into the local language. That did no doubt do a lot for it to get accepted in office environments around the world.

That the data exchange format is also localized and incompatible between languages is a different issue of course (see the many, many, many ... faults above). But they did get the language support basically right.



Post has attachment
A timely reminder that modern hardware is actually as much software as actual hardware these days:

In brief, a lot of (most) desktop CPUs from Intel from about 2015 onwards have a bug that will cause your programs to crash or give you wrong results in some circumstances. It can only happen if you have hyperthreading enabled.

Intel has a software fix (a microcode update) for a few CPU models, and there will be more available over time. This will be either a direct microcode update, or via a BIOS update.

But of course a lot of people never install BIOS updates; nor are they aware that they even exist. Your operating system generally installs microcode updates for you, but there's no saying how long it will take before it's released and installed for you.

So meanwhile, the recommended course is to disable hyperthreading on affected CPUs. This should actually not impact your performance very much; hyperthreading isn't "real" threads, and for compute-intensive workloads you don't really see much benefit. On compute clusters and workstations you generally disable hyperthreading by default, since it causes more issues than it solves.

Wait while more posts are being loaded