Thursday, April 14

Geek

Daily News Stuff 14 April 2022

Build Back Bunker Edition

Top Story

  • Contracts exchanged today.  I also bought a new dinner set that matches the colour scheme of the new house.  (Gibson Elite Casa Gris in case you're wondering.)

    I'm going to need to stay up to 3am playing Minecraft until I recover.


  • My current ISP doesn't offer adequate business broadband plans.  They offer cheap 100/40 plans, and expensive plans at 250Mb symmetric and up, but not the common 250/100 plan.  So I'm probably going to finally switch providers.

    I'm currently on a 100/40 plan - but that's the best I can get here.  At the new house that's almost the starting point.


  • Nvidia's RTX 3090 Ti is here and it's, well, expensive.  (Tom's Hardware)

    Not bad, just expensive.  Not much more expensive than the regular 3090 but that is already very very expensive - close to $2000 if you can even find one.

    AMD's own overpriced minor upgrade is also on its way.

    The 6950 XT appears to be even worse value for money.  The model they spotted is water-cooled, yes, but it's over A$3000, when a regular 6900 XT can be easily found for A$1700.

Tech News

  • Zero Nines Uptime, or, Atlassian and the terrible horrible no good very bad #TwoWeeks.  (Pragmatic Engineer)

    How a major cloud services provider achieved worse uptime than some random guy running 100,000 websites on a broken $50 server.


  • The gambler's fallacy is not a fallacy if the events are not independent.  (Stranger Apologies)

    There, saved you from reading 90,000 words looking for the part where he bothers to explain why he's not obviously wrong.


  • I've described Docker before as the world's least efficient package manager but it doesn't have to be.  (Florin Lipan)

    You can create a 186k Docker container that provides a working web server.  It's just that no-one does.


  • Intel's 13th generation Raptor Lake parts could hit 5.8GHz.  (WCCFTech)

    That's on the high-performance P cores, with the low-power E cores hitting 4.5GHz.

    Ryzen 7000, also due later this year, is expected to have all-core clocks over 5GHz.

    The current models from both companies are already extremely fast, and these improvements should be pretty substantial.  AMD could improve by more than 30%, and while Intel will have a smaller single-core speed boost, they will double the number of E cores from 8 to 16 to provide a significant increase in multi-threaded performance.


  • Constant notifications are ruining your produc - hang on, I have to take this - tivity.  (ZDNet)

    It's not like I have 164,033 unread emails in my main inbox either.


  • The Lenovo Legion Y700 is now available... In China.  (Liliputing)

    $499 for an 8GB/128GB model configured with the English language and the Google Play store.  $599 for 12GB/256GB.

    This is a high end small Android tablet, with a pretty recent Snapdragon 870 CPU and an 8.8" 2560x1600 screen.  And a headphone jack and a microSD slot.

    It's not cheap but there just plain aren't any real alternatives right now, unless you can find stock of Huawei's MediaPad M6 somewhere

    .
  • One of my nuggies tasted funny at lunch.  Not bad, but as if the coating was a different mix.  Just waiting to see if I accidentally got glutenated, which I have successfully avoided for more than two years.

    Update: Didn't get sick.  Was that a nugget-sized chicken tender?  The same company produces gluten free chicken tenders as well and they do have a different coating.


Disclaimer: Bacon and eggs are gluten-free.  Just sayin'.

Posted by: Pixy Misa at 02:37 PM | Comments (4) | Add Comment | Trackbacks (Suck)
Post contains 584 words, total size 5 kb.

1 "The model they spotted is water-cooled, yes, but it's over A$3000, when a regular 6900 XT can be easily found for A$1700."
I was wondering if that was just the Australian tech tax until I saw that the 6900 XT prices are so much lower.
Hopefully this is just one retailer being greedy.

Posted by: Rick C at Friday, April 15 2022 12:31 AM (Z0GF0)

2 AMD GPU pricing is weird right now, at least in Australia.  Some specific models have come down sharply in price, other models from the same manufacturer at the same online store are still sky high.


What with the new house and all I'm not planning any more computer purchases for a while anyway.  I will keep on reporting on new toys, just have to pause on buying them.

Posted by: Pixy Misa at Friday, April 15 2022 01:22 AM (PiXy!)

3 Statistics is important, and I suspect that we are pretty strongly wrong in how we use it.

That guy's argument is not grabbing me.

He seems to be making an argument about what an 'uninformed person' should be assuming, at a length that is suited for someone who can actually study stats, and learn about using appropriate tools.

The thing is, ordinary humans touch a lot of random processes, and can find themselves needing to reason about wildly different types of process.

Human behavior processes, in particular, are far trickier than is often assumed.  For example, 'representation' is making two or three statistical assumptions to go from comparing two fractions to concluding that an organization is behaving in a racist way that needs to be corrected.  Generally, efforts to model human behavior can include some pretty wild assumptions about information, memory, and decision making.  When you are talking about a large group of people, their 'wisdom of crowds' exceeds the person modeling them, and the model of crowd behavior tends to only possibly be reliably realistic when the modeler isn't malicious to the crowd.  Because a malicious modeler tends to correlate to a crowd with incentive to surprise the modeler. 

My issue with the guy writing at length here, due to reasonable causes, he seems to have his eye on the wrong process. 

Making a series of bets is a behavior, and a process.  The textbooks explicitly talk about the process being gambled on, but the habit of gambling is the important one for the gambler's fallacy.  Since it is a behavior, and behaviors are extremely hard to have quantifiable information about, the gambling process will tend to only be implicitly discussed in a stats book using the gambling example. 

Once you have made a series of bets, what is your chance of guessing correctly on the next bet?  This is not ergoditic.  People who have a firm  policy  of quitting at some point, quit earlier in the series.  Late series gamblers are biased towards elastic policy, or to doing calculations for each bet, without necessarily considering memory.  People often have bad reasoning about statistics, and will be more bad after a series of exciting bets.  If you have a series of bets you have profited off of, that is more likely to contaminate your estimation of how next to bet in order to profit.  Because prolonged excitement and stress cause the mind to drop information. 

Explicitly, the gambler's fallacy is about reasoning about future outcomes from past outcomes.  There, the 'what do we really know about the process' makes logical sense to consider.

Implicitly, we use it to chastise people for behavior.  That behavior is not independent from gamble to gamble.  A complicated consideration makes it easier to so confuse the analysis that your conclusion becomes what you want it to me.  IE, it makes things worse.

In reality, a lot of processes fit a specific set of assumptions much better.  Listing the possible sets of assumptions make it intuitively seem like they are all equally likely, and they are rarely all equally likely. 

Posted by: PatBuckman at Friday, April 15 2022 04:55 AM (r9O5h)

4 Statistics assumes that random events are truly random.

Posted by: normal at Friday, April 15 2022 11:39 PM (LADmw)

Hide Comments | Add Comment




Apple pies are delicious. But never mind apple pies. What colour is a green orange?




56kb generated in CPU 0.0208, elapsed 0.4257 seconds.
58 queries taking 0.417 seconds, 344 records returned.
Powered by Minx 1.1.6c-pink.