I’m quite new to ChatGPT 3 but have played with several other packages capable of writing code, creating general text and more (I’ve been fascinated with AI before the term was in general use). I have of course played with the various AI programs that create art from simple text descriptions but by comparison they seem harmless except to mediocre artists). I’m also looking forward to having a play with chatGPT 4 in March. At first glance the program seems able to write just about anything – and with luck it WILL put out of commission some of the people who write NONSENSE in, say, Amazon product comments sections.
For years now, positive articles and comments have been planted on commercial websites to encourage you to buy products – I have no problem with that unless they actually talk complete nonsense – after all everyone (and now – it seems – everything) is entitled to an opinion. There are MANY blogs and general websites out there deliberately spouting nonsense – I’ve no idea how many people are taken in by this – but what we are seeing now is machinery able to do the same thing at high speed but with little or no human effort.
So, to give ChatGPT a simple test, I asked it to talk about how great my home village is (in 200 words). Off it went and dutifully wrote a glowing report on how GOOD our village is. At this point I was taken in by the novelty value. I then changed one word, asking it to write a damning report about my village. Once again off it went and wrote a perfect report on how BAD our village is. The two reports utterly contradicted each other and the facts stated were sufficiently general or vague that someone from out of town could easily read into either report exactly what they wanted/expected to hear.
So, to the blog – I COULD (and I’m sure many WILL) use this software to save a lot of work and talk about, different products – in fact – as another test, I had it write for me a glowing report on the value of Sonoff temperature sensors (without even giving a specific product name).
The response came back – how great Sonoff temperature sensors are and how, armed with a mobile phone – you can easily control your central heating using the Sonoff sensors. All very nice and compehensive – I won’t bore you with the actual copy that ChatGPT produced, but sensors don’t actually control ANYTHING – they READ information. The report on second glance was utter gobbledegook.
Suffice it to say that what you read in here will continue to be written by me, about products and services I’ve actually used – and they will continue to be produced with no guarantee of being positive – manufacturers and suppliers who send me samples know that how good or bad the review is will depend on how well the product performs. I have no intention of changing this format and that attitude has cost me VERY little in terms of falling out with suppliers over the years – many WANT writers to be critical – and honest.
I don’t know about you guys but MANY times over the years I’ve bought products that turned out to be duff because I’d never seen bad reviews and hence assumed the product was good. Designers and manufacturers who fail to encourage honest reviews invariably learn nothing and hence their products may not improve with time – and of course if you buy a duff product not only do you NOT buy it again – you often assume that everything else that company produces is suspect and you may wll tell that to others in your business and/or social circle. We will all of course have to get more savvy so we can better differentiate between auto-produced and hand-written content – but then people are regularly duped by human-produced nonsense every day – that’s the world in which we live.
I cannot promise that everything I write will be 100% accurate but it will always be the product of my best efforts. That I can promise as we move into this new, unfathomable world of AI.
While I’m here – as lots of text can be a tad boring – so here’s some art I “made” with the considerably less worrying DALL-E 2 – all good except for the absolutely rubbish cat on the table.
For a little background (and this will likely only mean anything to older readers) – last century I wrote my own pretty neat version of Eliza (the auto-psychiatrist which was anything BUT AI – but great fun non-the-less as it produced nonsensical responses to human input). Some may also recall a program from 1981 called The Last One, implying, somewhat ridiculously in hindsight, that programmers were about to start losing their jobs. If you’re new to all of this – check out the link above...
Meanwhile, thanks to NightCafe, here’s one more ‘cavehome’ – with China locked down for their new year and Covid (what a combination) I’m keeping myself busy creating ‘art’ for fun. More gadgets soon – I promise.