Some guy 10 years ago moving our data platform to cloud with massively parallel processing: this will be awesome, cheaper, faster, no drawbacks!
Me 10 years later: I have to explain the stack of reasons, including massively parallel non-deterministic-sequencing processing, to someone without a CS background, why I need to use a tolerance when comparing the results of calculations that have 5 to 500 steps before producing the result.
I got into an argument with #Microsoft #Copilot today.
I asked for PowerShell code to trim the first N digits from every filename in a directory. Testing showed it actually removed 2ยทN characters. When I reported the bug, Copilot insisted that was impossible and said I must have run it twice.
I knew the cause and the fix, but I treated it as a learning exercise to see what it would take to get Copilot to correct itself.
1/4
A fun trick I learned recently is Scooter Software's Beyond Compare can be used to extract many archive files such as RAR and 7z without requiring additional software to be installed.
Beyond Compare can treat an archive file as a directory, and it allows copying files from one directory to another. Voila.
#BeyondCompare #ScooterSoftware #ScottComputing #ScottSoftware
Can the ai components (NPU) of new processors be used for non-ai work?
What I'm thinking is along the lines that the vector-processing components of GPUs have also been used to speed up non-graphic mathematics on large arrays.
To clarify, I would state some basic principles for semiconductor computing:
1. Computers are based on binary. 0s and 1s are represented by high/low voltages.
2. Engineers design complicated devices by arranging huge numbers of simple components.
3. One main component of circuitry used for calculation & storage are transistors โ electrically controlled switches.
4. Computer transistors, diodes, etc. are built with thin layers of semiconducting metals on silicon base.
2/2
Something I don't understand and wish I did is "what are the basic principle(s) for how quantum computers work?" I've read a few articles here and there, and if anything, they make me feel like I should already understand this.
As I think about it some more, it's more likely that the articles don't explain it because the author doesn't understand it well enough to explain it and/or assumes the audience isn't interested.
1/2
A co-worker using Python is having issues trying to load a dataset. The error is it can't allocate 40+ Gigabytes RAM.
In #SAS programming, this is mostly not an issue because SAS is generally oriented to data sets on disk. It loads a chunk at a time, performs operations, and progressively writing results to disk. Thus, SAS has no hesitation working with data sets much larger than available RAM. A strategy that worked in the 1970s and the 2020s.
I'm a little surprised that in 2024, Microsoft hasn't made native Office 365 binaries for the Xbox.
The last two XBox generations are heavily based on x86 commodity hardware, and support mice and keyboards. In some households, the most powerful computer would be the Xbox console.
I assume the main reason they haven't is marketing reasons -- Sony would advertise that the XBox is for boring work console while PS is pure fun.
Scott Miller ๐บ๐ฆ ๐บ๐ธ