As someone who uses AI and has become familiar with its language, I can confidently say that I can recognise the language of AI. It is very particular: a little flowery, overcooked. AI tries too hard to write excellent text and produces language that is slightly unnatural.
Yes, it is excellent in terms of technical quality. There are no spelling mistakes and the grammar is perfect but: as mentioned it is not truly human and therefore slightly unnatural.
It is close to being human but it seems that the programmers have failed to really get to the bottom of what human writing looks like.
Look, AI written text from ChatGPT is very impressive and it can summarise complex information brilliantly which makes it attractive to use when turning out an article or essay. It does all the heavy lifting.
And it can be directed i.e. managed accurately to create an article that one would write, even by changing the style of the text. However, the language sometimes reveals the article’s source.
I have to admit that the signs can be subtle but the more one uses AI the more obvious they become.
Therefore over time internet users, teachers and lecturers will be able to recognise AI text. Does this matter?
Well, I think it does because another aspect of AI text is that it looks sort of homogenized by which I mean slightly bland and vanilla. Almost too perfect and rounded. To which you have to add in the potentially incestuous nature of AI when it comes to research. It will find its own articles on the internet and use them for the next article. That is ‘AI inbreeding’! It must result in dumbing down and more homogenisation.
AI text can be too balanced and sensible to be human. And there is a trust issue here. Can we always trust AI-written articles?
In general I think we can as much as we can trust human-written articles in terms of accuracy of content but this is more about text which does not include original thought. AI text does not include ideas. It is beautifully crafted wallpaper-text! Except for the much reported ‘hallucinations’ (fictions).
I am sure that expert internet users will in due course be able to recognise AI generated text. When this happens it may become less popular although the programmers will probably be able to better disguise this subtle defect in the months and years to come.
There are many AI-generated videos online these days by which I mean both the video imagery and the words. Once again the commentary is often clearly recognisable as AI-generated.
I sense that in the long term AI will damage human writing skills because skills need practice and delegating to AI precludes writing practice. It might be a negative force for humans in this respect. Perhaps it will dumb down humans further. We will stop advancing and start regressing.
If that is true it will support the notion that AI might one day be managing obedient and stupid humans.
More: AI
