Write, write, and write again

One of my regulars (blog-readers, that is) recently sent me a fascinating article by Ludwig Siegele, Business Editor of the Economist, on “Foundation Models”. Initially, I thought this term was somehow a politically correct replacement for ‘Under Wear’, the word ‘Under’ being deemed either crude, or evocative of serfs being socially under their lord.

I read further and realised instead that Foundation Models are huge Artificial-Intelligence (as opposed to artificial butt-and-boob fillers) algorithms that are much more versatile than earlier forms of AI. According to Siegele, most of the algorithms can write “eerily good prose, analysing a text’s sentiments and explaining jokes.” Hmm, I wonder if he means ‘explaining’ jokes or ‘justifying’ them to the thou-must-not-offend-laugh-or-disagree-with-my-opinions woke brigade.

Other algorithms can be designed to create images, and future ones might be able to build virtual worlds or lay out business strategies.

Not surprisingly, I’m most intrigued by the writing-good-prose capability. So was Siegele. He asked a PhD student in computer science to help him with an experiment, the aim of which was to find out if a robot could write as well as he could. The student modified an algorithm that functioned as a ‘writing assistant robot’ into two versions and fed them Siegele’s 100 most recent articles, plus loads of scientific papers on AI; one of these versions could finish a sentence, the other an entire paragraph.

The first version seemed to work quite well. Cop hold of the following; Siegele wrote the plain text, the robot the italics:

“Not all technologies are born equal. There are single-purpose ones such as washing machines and coffee-makers. They are meant to perform just one task and do it well. And then there are general-purpose ones, like the steam engine or the microprocessor.”

And so it went on. Eminently eloquent, wouldn’t you say? However, when the second version was asked to complete a paragraph, it all went silicon-tits-up. Siegele said that as well as being less coherent, the resultant text included some ‘hallucinations’, such as the quote below attributed to Martin Wiktoro that not even Google could find.

“Not all technologies are born equal. There are single-purpose ones such as the steam engine and general-purpose ones, like the computer. The difference is not just of degree, but of kind. The engine helps perform a specific task, but the computer can solve any number of problems. This makes it a ‘technological complex’, in the words of Martin Wiktoro [boss of Intel]. It is not just a collection of related technologies, but a whole ecosystem, including software, cloud services and other chips.”

Siegele concluded that using an algorithm instead of humans is “quite helpful when a writer is stuck and needs inspiration. But as a true writing buddy it falls short.”

However, he predicted that they might still be engaged in the future.

Me? I’m never short of inspiration and, anything a robot would write I’d probably rewrite it, being the micro-manager that I am. On the other hand, I’m not without a little mischievous intent, such as seeing how far I can ratchet up my hitherto polite, measured, and laid-back analysis of the views of my sworn enemies.

So, I’ve invented my own robot (in my head), imagined that I’ve programmed it with all my blogs, plus a penchant for googling, and asked it to rewrite the first two paragraphs of this blog, firstly finishing each sentence then finishing each paragraph. 

First version – robot finishing each sentence (in italics):

One of my regulars (blog-readers, that is) recently sent me a fascinating article by Ludwig Siegele, who has over 500 LinkedIn connections, on Fundamental Muck Spreading. Initially, I thought this term was somehow a politically correct replacement for ‘Under Wear’, another way of describing politicians that better reflects their dreams, aspirations and fetishes.

I read further and realised instead that Foundation Models are huge artificial silage dispersers that are much more versatile than earlier forms of Angela Rayners. According to Siegele, most of the algorithms can write “eerily good prose, analysing a Remainer’s motives and explaining collective sense-of-humour failures.” Hmm, I wonder if he means ‘explaining’, or ‘cherry-picking’ facts to fit a metropolitan liberal elite narrative devoid of transparency, logic, humour and democratic sensitivity.

Second version – robot finishing each paragraph (in italics):

One of my regulars (blog-readers, that is) recently sent me a fascinating article by Ludwig Siegele, who’s a businessman, editor, economist and lingerie model. Initially, I thought lingerie models would replace economists as Civil Servants, until I realised that the former actually leave their homes to go to work and do as their employers ask.

I read further and realised instead that Foundation Models are artificially intelligent, still reliant on butt-and-boob fillers to pay the rent, feeding versatile facts into algorithms so that what is produced is neither intelligent nor factual. According to Steven Seagal, none of the algorithms can beat him in the kitchen, being unable to write, decipher a recipe or tell jokes. Hmm, I wonder if he means ‘tell jokes’, or communicate in code in a way that bypasses the censorship of the unintelligent, robotic ‘thou-must-not-offend-laugh-or-disagree-with-my-opinions' woke brigade.

I think there could be something in this, you know. Next time I’m berated, harangued, cancelled, misunderstood, whinged at or loathed, I could claim it was the robot’s fault.

On the other hand, I’m very proud of owning every word of my blogs each time I’m berated, harangued, cancelled, misunderstood, whinged at or loathed, given the political persuasions of those doing the bra-ing, hanging, cock-a-leeking, meandering, wobbling or lifting and separating.