Forgive the rant, but I am feeling very ranty atm. About time, actually…
Not going to go into background material here, except to say that, if you have a brain, have been in the programming or sysadmin worlds for x amount of time, or just involved with technology, design, etc., then you know and see by now that we are hurtling towards a previously unthinkable level of automation within our data management, processing, and etc.
While presumed to be the beta versions of works that will be improved upon over time, we will first see the rough equivalents to my old, most beloved, Sarah - a “iterative, subject-based, self-adaptive, conversational-platform - a fancy way of describing a chat bot that can be used to learn of and adapt to the needs of the end user by talking to them. The vanguard of this revolution won’t be engaged in convos with the end user, however, it will be attempting to provide and supplant real time intelligence and business analytics, at first to human operators, and eventually, into its own systems.
This is the point I want to make:
Anyone who has been working with databases, and database-driven applications over the last 6 or 7 years knows that this is coming, and that, beyond it’s innevitability, its fundamental necessity: our ability to gather and organize data does nothing in terms of aiding us to understand it (just ask the NSA: rofl: yet another reason why I marbvel at he recent disclosures: they make very little sense): in order to understand data, we have to be able to know what we wish to ask of it (and, in order to know what we are able to actually glean from the data -as we have it structured - requires knowledge of the actual methodology behind the way it is structured: in short: in order to know what you want to know, you not only need to know that you don’t know it, but need to be able to articulate a way to ask if you know what you don’t know that find what you don’t know that you are looking for, parse the data, and then present it in such a way that the doors of the infinite have opened to reveal exqactly wht you need to know, but didn’t know you needed to know… and that’s - more or less, plus/minus amounts of whiskey - the holy grail of all data:
Data can’t be read by humans in any large-sample sense. Data cannot read to humans, either: it can’t tell you what secrets it has inside of it. It needs to be questioned carefully (the best questions, btw, generally coming from the source of a mere hunch that something might be there, combined with an intimate knowledge of the signal to noise ratio and the actual structure of the database itself… that frequently results in stuff that might be worth watching closely)… anyway… The catch is: your question (your criteria) are the key. The Holy Grail currently is how to make computer programs that don’t need you to articulate to them a framework that accurately describes what you don’t know (which is, truly, the most excellent secret sauce) - because, by then, you already know it… the problem itself contains the solution, and the rate that mere humans can produce such excellent questions seems far too slow for the desired predictive stochastic models that business, science, and governments are looking for.
We want the computer to be able to see the significance in the data, and THEN be able to prioritize the data, all in real time, as it is processing, and then, when it really finds something that it thinks we might be interested in, we want IT to ask US the question: “Have you considered the cost of timber in Taiwan? It might be worth looking into as an investment indicator for near-term investment gains; want me to prepare an cost benefit analysis and strategy for you? No? Then what categories of analysis are you interested hearing about which I have flagged over the last [time_period]?”
Anyway. All of this stuff has been known by DB monkeys like me for a long time. What I want to point out is the following:
Within 7 years, the logic for the demi-artificial analyzition engines will become functionally iterative; demi-self-improving. It’s going to happen; we will try to flip that switch (the “scary” switch, where we eventually turn over the “SIGNIFICANCE” and “VALUE” judgements of hundreds of thousands a decision a second over to completely automated systems that self-optimize and etc. etc. etc.)…
Here’s what I think that most people’s concept of the next early-days of “Super Shiny New And Improved Future” have glossed over:
Before these programs can either gather enough data to become faux-self-aware (meaning: by using the functional equivalent of having cloud-sourced the data streams of all of humanity’s synaptic and physical activity, sublimated through logic pathways, and blah blah blah), and also before they can, in fact, become legitimately self-aware (the difference between the two appears to the end user as cosmetic, but the implications of it are humongous and crucial), we will not be able to TELL if “it” is either one or the other. Furthermore, it will be very difficult even for systems administrators, for programmers and software designers, for network experts and all other possible and potential auditors to accurately guage or express the system’s actual actions - or determine if it is operating as “designed/predicted.” This is because it will be changing the data as it is going (as it was designed to do), and writing new algos for evaluation and analysis as it goes. Of course, this starts with nice bordering constraints and parameters, but as it grows, those parameters will eventually have to be removed by its creators, and this is where the fuzzy part happens.
Imagine a banking system in this scenario. The DB analysis/demi-ai/logic/hypervisor-esque/software-thing was designed to approve or deny loans from end users of the bank based on a variety of factors, and ultimately based on whether or not and how either would ultimately affect the bottom line of the investment group controlling the money that is to be loaned out. This is a super simple example, of course, but imagine if - NOT SUDDENLY, but at a steady rate, just at a rate that we can’t see because we don’t know the questions that the program has answered, and therefore also can’t see the pattern within the data that is the result of the computer’s accept/deny decisions)… Well, let’s imagine that this whole system actually goes great: after a year of held breath on the part of the designers, we all collectively exhale because catastrophe hasn’t occured.
Now imagine, 4 years after this.
What happens to us when we actually try to look at or measure the efficacy of the data? And, after so much time, how can we possibly even see the rudimentary architecture that was the systems original logical code? How the hell can we know if it is “working” at that point? How could the economy itself survive if we had to turn it off?
But, even better (in the sci-fi terror-world scenarios): Automation in any system is a matter of elliminating human intervention into the quality of the data and/or typos or other errors (usually as the product of data recopied by human hand or some other form of human failure at the actual level of data or coding of logic to handle and manipulate that data)… the tradeoff is that automation thus inherently has a certain error rate of its own, in which, by virtue of the sheer degree of automation, edge cases (or other unaccounted or factors) produce data that is incompatible with the system-as-written. Think of a field that accidentally gets wirrten with 2 values instead of one or something similar.
These edge-cases (which can already be horrendous in ou current tech world - think “DMV” errors) will be instant, brutal, incolseable accidents, mostly without the possibilty for real recovery of property, life, or limb on the part of the victim… Such an error rate will have to be offered up to our new creation… First, a friend. Then an advisor. Then an equal partner. Then our manager. And, ultimately: our new Master.
Moar Whiskey for me. Back to coding. Cheers!
Never tell your password to anyone.
Saturday May 31
skookiesprite: Sigh hieya
skookiesprite: it has been that kind of morning.
skookiesprite: for you and me both mofo
Hieya2: it happens sometimes
skookiesprite: Ach you have no idea. So whjat are you playing?
Hieya2: not playing anything cant get like anything to work just online right now kinda given up on getting things to work
skookiesprite: Few stupid things I’m sure you tried: Have you run windows update?
Hieya2: sometimes this happens with our internet cause its kinda cheap so when a lot of people are online it just refuses to work basically
skookiesprite: Ok. Have you gone to speedtest.net?
Hieya2: nope ill do that now though
skookiesprite: If you were on linux I’d just tell you a bunch of command line tools I use instead but speedtest.net isn’t bad at all either
skookiesprite: And it will let you know if there[’s something jacked on your ISP’s end.
Hieya2: ah ok ive used it before but not much so i just usually dont think of it
skookiesprite: it’s free and no spyware or anything should just run inside the browser.
Hieya2: yeah it does it also has the thing you can download and run and itll tell you whats slowing it down and such
skookiesprite: that thing is bullshit.
Hieya2: yeah i figured as much
skookiesprite: just use the in browser button.
Hieya2: yeah thats what i did
skookiesprite: OR… alternativelyt…
skookiesprite: What did it say?
Hieya2: ping is 32 and download speed is 2.98 haha
skookiesprite: WHOAH that is mega horrendous shitty
Hieya2: yeah it is
Hieya2: what is yours usually at?
skookiesprite: one moment
skookiesprite: download 24.01 upload 24.35
skookiesprite: which is slow for me.
Hieya2: oh way better than mine still hah
skookiesprite: by a modest factor of 10. Yeah. Fuck me harder man… there is awesome guns of icarus to play…
Hieya2: definatly gunna try getting on it again a bit later
skookiesprite: have you called your isp?
Hieya2: no they usually dont help much they just say to wait it out should be good by tomorrow
skookiesprite: You should call them
skookiesprite: To tell them
skookiesprite: How much they fucking suck
Hieya2: haha maybe
skookiesprite: As I have learned - from amny years dealing with such situations: The first step on the road to progress is to open a dialog. To “YOU FUCKING SUCK MOTHERFUCKER WHY THE FUCK DO I PAY YOU?!”
skookiesprite: Take the first step
skookiesprite: Yell at someone powerless to help you.
skookiesprite: It is the end user way.
Hieya2: hah ill take the first step soon im kinda in the mood where i just kinda gave up on it
skookiesprite: Yes. But before you give up on it, make a powerless human being who is already being berrated by hateful idiots with your same problem…. make their day even worse. Trust me. You will be glad you did.
skookiesprite: And make it personal. Try to hurt them. Humiliate them. Make them feel terrible so that they go home and beat their wives, beat their kids… that’s what they make ME do… so… do it for e… make them hate life. Show them that there is no logic or compassion in this otherwise already cruel and turbulent world. Do it, kiddo. Do it for me.