As most readers will know, my background is in software engineering and I currently tutor for the[Contact Me Ref-2]in technology subjects. Hence, I try to keep an eye on new IT developments.
Recently, there have been renewed fears about “AI” – with dystopian predications being made by the likes of Elon Musk. Others rightly draw attention to “transhumanism” (humans merging with machines of one kind or another) – and the obvious implications that we would then become even more like robots – but perhaps have much longer lives.
However, in this posting, I would like to draw attention to a recent development, which popped up in my YouTube feed. It’s at a more “nuts and bolts” level and seems to be an embodiment of something I thought might exist back in 2005 or 2006. During that period, I was much more active on several internet forums than I am now. The main one I was posting on was the UK 9/11 truth forum. At one point, I even made a posting suggesting that other poster weren’t even real people, but some type of AI “bot” (posting copied below). I even went so far as to contact an old work colleague to ask him if he thought AI could do this (he replied “probably not”).
Now however, publicly available algorithms seem to be getting close to being able to do this very thing (i.e. generate forum postings that seem to be written by human intelligence). The short video below seems to demonstrate this (using YouTube comments rather than forum postings, but the principles are the same, of course).
And there’s a bit more discussion in this recent video:
So, what is it that we are talking about? It’s an AI package/piece of software called GPT2 (General Pre-Training 2). It is a way of searching massive amounts of English text for patterns of language usage and then using that enormous dataset to generate original language, similar in form to a template which a user gives (as demonstrated in the first video above). As you can see if you watch the video, some of the comments are a bit nonsensical, but quite a few could pass as being ones written by real people. In another example, the software has apparently been used to generate a convincing fake essay about recycling.
Fake Profiles and Computer-Generated Characters
Most of us have encountered fake profiles on social media or cloned profiles – when you see the operations shown above, it’s quite easy to see how the enormously powerful computing resources, coupled with the gargantuan amounts of easily accessible/readable data and be used for most imaginable nefarious purposes. We must also consider that software now exists to do similar things with human facial imagery and with voices.
I think that it will be far easier for those with nefarious intent to flood posting areas with comments designed to confuse people or even influence text-based discussions because it is less obvious when bots are commenting than when a fake video or voice has been presented somewhere.
More information on the GPT-2 Text Generation Algorithm
Read the articles/papers below if you are interested.
It is very likely that the black world (and/or the CIA, NSA etc) have had this level of technology for at least 10 years – so perhaps my suspicions back in 2005 were correct. (Back then, however, YouTube, Facebook and Twitter either didn’t exist or were in their infancy.)
NSA Forum Posting Bot
Loading Current Issues Database…done
Current Status: Active
Press Ctrl-U to Access User Console.
Ask the Tough Questions, Folks!