You Can Fool Some of The People...

-


Here's a thing: idly reading around and watching various YouTubes this afternoon - as you do - I picked up on three things that seem to reflect where we are at the moment in terms of large-scale computing. First off, COBOL: Common Business-Oriented Language; developed in the fifties and released in '59 as the de facto standard platform for handling significantly large amounts of data - principally financial transaction data - for banks, government departments and other large institutions: and it is still widely in use across the world today, with virtually all ATM transactions mediated and processed by it, as I think I've mentioned before; likewise the rapidly dwindling numbers of programmers left in the wild that can understand and modify/repair/extend existing codebases.

Secondly, I chanced upon an advert for a software product/service that claims to hide the fact that someone has used AI for content creation - of whatever kind and for whatever purpose, from the intended recipient(s) of that content - for instance, using the the widely-known ChatGPT to create a document from scratch from a simple set of instructions. This software/service is called 'undetectable AI', which processes whatever text you want to pass off as your own, but in reality has been AI generated, by 'humanising it' and generating a fresh text that will fool AI detection systems into thinking that a real person most likely wrote the text. The foregoing text was most definitely was written by me; a human [at least I think I am: but let's not get too Cartesian at this time of the evening].

So, let's do a bit of an experiment here and ask ChatGPT to write something on the subject of the last paragraph's content. Here's the seed question: 'Write a short summary of the process by which undetectable AI hides the presence of AI by humanising the text'. Here's the output:

Undetectable AI often uses advanced natural language processing techniques to make AI-generated text appear more human-like. This involves several key steps:

  1. Contextual Understanding: The AI comprehends the context and nuances of the conversation or topic, allowing it to generate responses that are relevant and coherent.

  2. Emulating Human Writing Patterns: It mimics human writing styles, including varied sentence structures, informal language, idioms, and emotional undertones, to make the text seem more natural.

  3. Personalization: The AI tailors responses based on the user's previous interactions, preferences, and tone, creating a more personalized and engaging experience.

  4. Error Introduction: Occasionally, subtle errors or imperfections are introduced intentionally to mirror the occasional typos or inconsistencies found in human writing.

  5. Adaptive Learning: The AI continuously learns from interactions, adjusting its responses to better fit the expected human-like patterns over time.

By integrating these techniques, the AI can produce text that feels authentic and indistinguishable from that written by a human.

OK: so the piece was rightly flagged as definitely having been written by an AI. So, let's 'humanise' the output and try again. Which results in: me hitting a paywall, to which I say piss off, people: go find some other mug punter that wants to hide their lack of intellect or writing skills: I'll carry on as I am, thank you very much, and goodbye...

Oh, and thirdly? Current IBM Z-series mainframe servers have hardwired AI built in at chip level, with which to interrogate in near real-time certain kinds of [financial] transactions to detect fraudulent activity: a whole different kettle of poisson from using some dumb text generator and a cloaking device to hide the reality of some people's shortcomings/evil intent from the public. Interesting to think of all that 'dumb' but innocent COBOL code still running reliably behind everything, though. Talk later. Cogito ergo sum, I think...


Comments

Post a Comment

Popular posts from this blog

Of Feedback & Wobbles

-
Image
Many years ago I came up with an analogy for poorly damped feedback loops. The feedback loop that prompted this frankly tart analogy was indeed a horror. Anyone who remembers shame-faced their first attempts at clutch control resulting in the good old kangaroo launch across the lights only to stall in the middle of the crossroads will have some idea of the kind of data transfer lag involved in said system. The analogy I made was that of a large blancmange. Untouched, a model of stasis and eminently stable and predictable; but prod the bugger and the resultant chaotic behaviour needs a very complex physical model and some very decent computing horsepower to predict its outputs, if at all. The system in question was a large and complex audio-visual display at what eventually became 'The Electric Mountain' in Llanberis. We were ultimately contracted to maintain the thing having been witness to and peripherally involved in its installation. Certainly not involved in its des...
Read more

A Time of Connection

-
Image
I was just reflecting on the vacuum of continuity that's left behind when someone dies: that sudden sense of disjoint, of fracture, in one's personal history that takes one completely by surprise at the oddest moments. These pinch-points in memory can often yield some surprising revelations of fact about periods in one's past. One such occurred to me tonight, tending my rather lazy meal of pizza in its journey from shrink-wrap to mouth via the oven. Al's recent demise still catches me out from time to time: the space he left bounded by forty plus years of friendship. But the thing that struck me tonight was just how much change was happening on a personal and social level in the very early years of our acquaintance. Between moving here from Birmingham in September 1980 and our buying our first proper house eighteen months later, all manner of things occurred that would chart the course of our lives to the present day. At around the time of our meeting Alan & Irene [...
Read more

Sister Ray

-
Image
Ray Keats - aka Ann Key - local artist & family friend, died this morning. She'd been ill for some years, but never let it get in the way of her work or of tending to the massive network of friends, colleagues and fans she had accumulated over the decades. Her house on the High Street of Rachub, just below where we live - a former cockle factory(!) - bought many years ago when she and her late husband Ronald moved here from Manchester to set up home in Yr Achub, is currently still home to her paintings and drawings, and the front window displays her work, including the last drawing she made in hospital, just days before she died. Her good friend Colin has been a diamond in her latter years, looking out for her and just being there when she needed him. She will be much missed by this tiny community of ageing bohemians up here in the hills, but suffice it to say, George, her cat, will be well looked after now she's gone.
Read more