Journal Bill Dog's Journal: computer model mythology 9
So I'm sitting here watching the women's finals of the U.S. Open (tennis, not golf; i.e. the boring stuff, not the really boring stuff!
This would be unheard of on the men's side, because 1) there's a significant difference between a top 3 and a top 10 ranked player, let alone a top 30 player, and 2) they're a hell of a lot more consistent; excepting Serena Williams (who looks to be part man herself, BTW!), the women flop around from #1 to nobodies, like musical chairs.
Anyways, both of these players have never been to the finals of a grand slam (I guess one of the big* four tournaments of tennis in a year) before, and for such people nerves can be a factor, typically, and to some degree. This is broad generalizing [pun not intended
*I think I heard last night that the winner's payout was now up to $3.3 mil.
IBM is one of the sponsors of the tourney, and they're always trying to reinforce their marketing angle of being thought of for data analytics, and they had a feature during a pause where the sports announcers told us that nerves would affect one of the players 55% to 45% for the other player.
WTF?!? It boggles the mind to imagine just how impressive in number and cocksurety the assumptions must have been that went into such a concoction.
But we believe that computers are impartial, and don't(/can't) lie, so it must be true. What people don't know of course is that, assuming a bug-free implementation of a model, the output is only as trustworthy as both the model and the data.
I tell my non-technical extended family members that computers are not magical soothsayers, they can only, basically, do what they're told to do. And a computer model is just, broadly-speaking [there I go again], telling the computer what to do with the data.
It may be more data than a human being can readily sift through to determine what their assumptions would amount to, but that's all the computer does, takes human beings' assumptions about things, and crunches data sets, small or large, in terms of them.
So the tennis match is over now. They announced the winner does indeed get $3.3 mil. And she announced that she's retiring. People were shocked to hear this, apparently even her coach. But she's 33 years old, which is prime retirement age from the game. And she has a fiance, so she probably wants to start a family, as most women do.
She said she's known the fellow Italian she played against today since they first played at about age 9. So she's had a long career, and even though she also said after the match that she made the decision to retire after this tournament a month ago, she might have been thinking about retiring soon anyways. And the why not go out on this high. But did IBM's Watson, Tennis Version*, know all of this? She probably wasn't hardly nervous at all, given all of that.
*They've been advertising versions of "Watson" for health care data and other areas, in addition to the Jeopardy version that originally made the name famous.
p.s. I just sprouted an "eye migraine" (a painless, developing "shimmering" in my direct field of vision, in both eyes/in my brain) a few minutes ago, so apologies if some of my last few edits are mangled.
On Wednesday (Score:2)
I got called in to work from a vacation because a laser printed a logo on glass when instructed not to.
Turns out "NO LOGO" is a LOT different from "NO LOGO", but no human could see it, including me. Even after ".IndexOf("LOGO")" returned 4 instead of 3, it took me a couple of minutes to realize what was wrong.
I responded that they needed to tell the data entry people that one space and two spaces are not the same.
to err is human; to make software smarter, divine (Score:1)
Then again, this is really the kind of thing that computers are good at (identifying extra whitespace) and people are bad at, and therefore should be automated. Either with user input validation, or automatic detection and correction (say if no commands may legally have more than one space between words). Or, it should've been "NOLOGO", like command-line parameters.
Re: (Score:2)
Yes, exactly. Unfortunately, client has run out of money, and this showed up in late production run. There is a possibility I may be allowed to fix it, but only if they convince one of THEIR customers to pay for it (it could easily be part of the upgrade that pulls this out of C# into SQL to be able to print random customer logos).
Re: (Score:1)
pulls this out of C# into SQL
God help you then. [slashdot.org]
Re: (Score:2)
It is less code to express the same logic in C#. But using SQL correctly for this (dynamic SQL pulled from tables) means a new logo, with new criteria, will be an insert into three tables with a download of a template to the laser. As opposed to what it currently is, which is download template to laser, modify C# code for the select case and if statements that run the logic, recompile, test on test bed, find a time when the factory is shut down so you can push a new recompiled EXE to the machine that tell
Re: (Score:1)
Ease of deployment is what is cited by my coworkers as well. Well, then why are we using MVC to generate our HTML, when we can write T-SQL to do that, and everything else. Then we'd be ready to change literally anything, immediately, without deployment troubles and delays.
Then again, my coworkers come from VB programming backgrounds, so I'm not surprised.
Re: (Score:2)
Actually, I will from time to time generate the HTML from SQL, and use the Controller only to deliver that content to the View. I consider for quite a bit of coding, MVC is overkill.
There's another project at my company that does precisely that- stores all the HTML in the database, serves up views that are read by a very minimal controller.
Re: (Score:1)
Sounds like you'd fit in great where I work. Wanna replace me? I tried introducing writing unit tests at this place, for example, but like all good software engineering ideas, the team probably considers it what you said: "Overkill". That's why I need to get out of there. I suppose being a contractor, expediency is mostly all that matters. Maybe I could've done that has a younger dev, when I didn't know anything, but atrocious practices are just too frustrating for me now. Now that I'm a grumpy old pro
Re: (Score:2)
Not sure I'd fit in then, my job is on the line because I suggested automated unit testing in my first week here.