Joe, The Hardboiled Detective by Hercules Budziwoj

Author:Hercules Budziwoj [Budziwoj, Hercules]
Language: eng
Format: epub
Published: 2024-01-22T00:00:00+00:00

CHAPTER TWENTY-THREE

Kids These Days

3:20 a.m.

Saturday

150,000 ft., Washington, D.C.

“Bud, I hate to interrupt, but neither of us speak Greek, and what about those captions?” Joe pressed. It wasn’t even 3:30 a.m. and he already felt like they’d wasted half the day. They cut these parts out of the Avengers movies.

“Sorry, sorry, but the weirdest thing. So we knew we could not do this manually. No surprise there. That’s why you told me to build the map. Of course, I am not inherently equipped to decide what is related to what, and why. That kind of analysis is really what the human brain does best, using a deeply reasoned framework of analogies built on lifetime’s worth of facts and experiences, learned both firsthand and through knowledge transferred across generations. That’s why, by the way, older, slower athletes can beat younger faster ones, because experience, which is the basis for situational awareness, and its corollary, anticipation, often trumps raw speed and skill.”

“Bud,” Joe said, “please get to the part where you tell us how you screwed up. Yes, we know you that well already.”

“OK, well, anyway, as I said, because I don’t have neurons that automatically form analogies based on rules developed over millions of years of evolution, and I haven’t spent the previous twenty years getting raised by an entire society, I had to build a logic engine - an algorithm - from scratch to make decisions as to what constituted legitimate connections and relevant relationships.

“So far, so good. But the brain is massively parallel, with trillions of pathways simultaneously operating. Your typical computer algorithm, though, normally works in series, not in parallel. In a normal computer, it can seem like a lot of things are happening at once because the processor is pretty fast, but usually they’re not. And if they are, it’s usually a few functions, not trillions. Since we don’t have forever, I integrated learning loops into my algorithm to allow it to communicate with other duplicates, integrate any new information it receives in return into its assessment of each datum it reviews, and, most importantly, use that new information to continually rewrite itself in relationship to the new information its receiving. So if one logic engine concluded that scientists A and B were connected in some way, it could share that information so that when another node decided that scientist B was related to researcher C, that node would know to check if researcher C was also related to A, and it would integrate that data into its calculations going forward. That’s a simplistic example. Then I made, let's say, some copies of the algorithm and sent them out into the web.”

“So your algorithm can learn,” Joe said.

“Yes, in a sense, if you want to anthropomorphize a few lines of code. But an interesting thing happened, almost right away. Although initially all data was exchanged and processed with each learning engine seemingly identically, there must have been some small, random distinctions, aberrations on the quantum level between the

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.