Early Impressions of Artificial Intelligence

As a regular reader, you’ll know I’ve taken on a pupil, Mr AI Bot. A useful chap, undoubtedly artificial but every inch the novice when it comes to the realm of intelligence. The vacancy attracted a number of applicants, notably the famous Master ChatGPT, regrettably passed over due to his mobile phone fixation which doesn’t mesh with the ante-diluvian process through which words transform into articles for your personal favourite, Going-Postal.

Instead, the gig went to Writesonic, who was anointed with the office name of Mr AI Bot, in much the same way Judy Dench became M. In exchange for a reasonable monthly fee, Mr Bot produces up to 20,000 words of text. However, these words roll over, any unused from one month carry on to the next. On the other hand, some words count as double. No matter. This arrangement has provided more than enough verbiage for my contribution to the unread articles that top the unacknowledged comments on your preferred high-quality politics blog.

Initially, I was quite impressed. However, with use, one becomes familiar with and then irritated by an AI style that is somewhat robotic, repetitive and often factually wrong. As things stand, artificial intelligence’s effectiveness lies in providing an initial draft that has to be rewritten by biological intelligence (ie me). If the copy is meant as pap to fill space, the rewrite will be brief. If it’s an investigative piece for a sophisticated audience then the rewrite will be thorough and require additional research.

On the plus side, the spelling is better than mine (albeit American) as is the placement of commas. It also puts a topic into some kind of order at the click of a button whereas I tend to approach a new subject by writer’s-fogging my brain by zig-zagging around the internet with half a dozen windows open on my screen simultaneously.

Therefore, with AI I can spend more time on what I’m good at and what AI struggles with such as making things for a model railway, cutting the hedges and, of course, digging deeper for an article. The bot also usually brings up one or two things that I hadn’t thought of which can then be followed up by myself or, obviously, explored as a supplementary AI request. So how does this work for the user?

Writesonic is entirely online. There is nothing to download. Having logged on to your account you’re presented with a number of different templates to use for your writing. From LinkedIn ads to product descriptions, from responding to negative reviews (take note Fiona Bruce) to Facebook ads.

For my purposes, I use the Chatsonic template which researches simple questions into a fairly brief paragraph or two of text as if you were chatting on the subject with a semi-informed layman.

This starts me off on G-P’s regular Saturday evening family album-based nostalgia slot (tell me about Aberystwyth in 1932) and helps with the panellist biographies on Question Time Review (remind me about Lisa Nandy, yet again, and who the heck is Robert Jennrick?). It defined mass psychosis for The Real Science Behind Global Warming and, for the same article and after a fair bit of coaxing and carefully crafted input, was bullied and tricked into questioning the global warming agenda.

However, I’m struggling to find signs of intelligence. It simply repeats what it finds elsewhere by skimming Wiki and a few other usual-suspect internet sites and putting some of what they contain into half-decent but clanky sentences.

Puffins have already rumbled Steven Fry, whose memory and recall of fact tricks lesser mortals into thinking him very clever. Likewise, Jacob Rees-Mogg’s country house and Eton education don’t make him posh, it merely reminds the well-informed that his father was rich. While we’re on the subject, this award-winning author is aware of an overdue day when Puffins realise his raconteurial flowery prose makes him a semi-literate near-fantasist rather than a convincing writer.

Likewise with Mr AI Bot. As things stand, his grasp of the written word is over-praised and he will not be replacing those capable of insightful perspective and distinctive writing. So why all the fuss in mainstream media? Because AI will replace the not-so-clever (or clever but lazy) mainstream hacks who’ve been stealing a living for too long while sleepwalking through rehashing press releases and pasting up pre-prepared elitist agendas handed down to them from on high.

Amongst our early impressions, we need a definition of intelligence. Shall we ask Mr Bot? I think we should:

Intelligence is a multifaceted concept, often described as the ability to learn, understand, and apply knowledge. It encompasses cognitive skills such as problem-solving, reasoning, memory, and the ability to adapt to new situations. Intelligence also involves emotional understanding, creativity, and the capacity to absorb and interpret information from the surrounding environment. Though traditionally measured by IQ tests, the understanding of intelligence has broadened to include multiple types such as emotional, social, and creative intelligence.

Do we believe him? His definition is very wordy and remember Mr Bot is only churning what’s written elsewhere on the mainstream web. More simply, intelligence is the ability to work out what you haven’t been told, making correct conclusions from incomplete evidence. The less evidence you need to extrapolate correctly, the more intelligent you are. This is measurable via an IQ test where you’re given some information and asked to conclude from it a correct solution.

As with everything measurable, this becomes hierarchical. With the less quantifiable concepts of emotional, social and cultural intelligence shoehorned into the issue by AI, it’s easier to create a fog around the woke blasphemy of a league table of intelligence that sees Eastern Asiatics at the top and the poor old Abos and Sub-Saharan Africans at the bottom.

As for a definition of artificial intelligence, Mr Bot remains wordy but is closer to the mark,

Artificial intelligence (AI) is a branch of computer science that aims to create systems capable of performing tasks that normally require human intelligence. These tasks include learning from experience, understanding natural language, recognizing patterns, solving problems, and making decisions. AI can be categorized into two types: narrow AI, which is designed to perform a narrow task (like facial recognition), and general AI, which can perform any intellectual task that a human being can. AI is not just about simulation of human intelligence, but also about the augmentation of human capabilities

Arguably, in the facial recognition example given, narrow AI isn’t intelligent at all but mechanically matching images at the pixel level. When a brain surgeon looks at a scan of a tumour, he recognises a tumour. When narrow AI looks at the same image, it recognises nothing but rather compares the patterns of pixels to thousands of patterns of pixels of images of tumours and not tumours – looking for a match or an average match.

The process might be improved by using an algorithm regarding the dimensions and shapes of tumours, but again the AI does not recognise anything rather it measures the distances between different types of pixels and compares them to the ‘likely a tumour’ formula in its algorithm. The advantage of AI is that it can come up with a suggestion quickly because it is computer-based.

What, you thought the brain was a computer too? You’re wrong. And so was I. Calculate the following in your head (((12/4.1)*1.45)-11)*1/2. Even if you did, you did it much slower than the simplest computer. The processor in the computer runs very fast. It reduces the numbers to binary then one at time, but quickly, adds them together or takes them away through those AND NAND NOR and OR gates your physics teacher told you about while you were playing hangman with a telex machine at your local technical college.

The brain is very not like that. I found out the hard way because of a family illness. Baffled by my father’s behaviour after a stroke, and before the invention of the internet or Amazon, I went to our local library and from the small section available took out and devoured an academic work about the brain. It wasn’t an easy read but in time I began to make sense of it (as brains do).

Just as Stephen Fry is not a genius and Jacob Rees Mogg is not minor royalty, I’m not an expert in neuroscience after half-understanding one thirty-year-old book but what did strike me is that the brain is an organ of awareness that helps us to successfully interact with our environment – not a big calculator.

For example, more than half of the surface of our brain (the cortex) is dedicated to processing visual stimuli. Different parts of the brain deal with up, down, left, right, movement in two dimensions and movement in three dimensions. If any of these parts of the brain are damaged, the victim may not see things in a certain place, or when they move, or until they move. This is what happened to my father.

Likewise, occupation of space is self-evident because part of the brain is perceiving it. If this part of your brain is damaged you, like my father, will pile all the washing up onto exactly the same spot on the draining board, smashing all the glasses as you jam them on top of each other.

You may also stand beside an open front door, not knowing whether you are outside or inside the house. Hilariously, and simultaneously heartbreakingly, unable to recognise the features on your son’s face, you will begin to call him Doris. It is this kind of biological mediation between the self and the environment, in a survival of the fittest immediacy, that ‘dead’ computer-based AI isn’t suitable for and can only unconvincingly attempt to mimic.

In intellectual task performing, AI is on firmer ground but is still merely churning, Steven Fry style, data it finds elsewhere, making it a danger to the jobs of those who work in the lazy-content industry – and to the sanity of those who rely upon the fake news for their (mis)information.

As things stand, with G-P positioned at the reporting harsh reality end of publishing, Puffins have little to fear from AI beyond the occasional flat Question Time Review biography or cheesy introductory paragraph to Aberystwyth seafront in the 1930s.
 

© Always Worth Saying 2023