Ruminations on artificial intelligence, ChatGPT, and critical thinking

Aweh dearly beloved fellow ruminants & groupies

There is this concept of techno-fideism which revolves around the belief in technology as our ultimate problem solver. It is a concept that combines elements of religious faith or spirituality with a belief in the inherent goodness or divine nature of technology. Technology will be our saviour and the solution to existential problems, including climate change, ageing, disease and more.

I identify, amongst other things, as an ageing techno-nerd. I like technology more than most people. I am old enough to have watched waves of technological euphoria and how people try to resist new technology. Computers took off in the 1960s and 70s but it wasn’t until the 1980’s when personal computers took off that the computer age truly arrived. Computers have changed the way we live and work. They have become indispensable tools for both students and professionals, and their absence can potentially relegate one to a state of impoverishment.

The dawn of the 2000s witnessed the internet’s ascendancy, accompanied by exuberance and a speculative bubble. Undoubtedly transformative, it was not devoid of exaggerated claims and unrealistic optimism. Similarly, the rise of genomics triggered expectations of conquering diseases and halting ageing. While notable strides have been made in treating various ailments, the prospect of conquering ageing remains distant.

In the realm of climate change, a plethora of technological solutions have been heralded as panaceas. However, it is imperative to acknowledge that amidst the considerable hype, not all these technologies will fulfil their grandiose promises.

The latest new technology on the block causing a renewed bout of euphoria is artificial intelligence. When ChatGPT was released late last year I was blown away. When I asked it questions on a variety of topics, I was very surprised by the quality of its answers. It appeared at least initially to be intelligent.

Right from the start, my younger son, Connor, a computer science student, was a lot less impressed because he understood how it works. He understood the magicians’ tricks. He rolled his eyes when I said that ChatGPT appeared to be intelligent. It took me a while, but I now realise he is right because I now have some understanding of how it works. ChatGPT is not intelligent or clever in the human sense and is in fact as dumb as a rock. The hype is just that hype. It’s just as dumb as the computer that I’m writing this blog on. That does not mean it is not incredibly useful, powerful, and transformative in its own way. It’s just not intelligent and it’s certainly not conscious or even close to conscious.

This prompts the question: What underpins the way ChatGPT works? ChatGPT crafts text grounded in an expansive database of human-created content, endeavouring to predict the most fitting response based on contextual cues. While it adeptly composes poetry or sonnets, their essence is drawn from preexisting content. It can compose a poem or limerick for you based on its knowledge and classification of hundreds of thousands of poems and limericks. The poem will even be original in the sense that it is new, but it will be generic and based on all the human-generated text in its database. It can create a Shakespearean sonnet on cats. I tried and it duly provided a sonnet. I’m not much into poetry so I can’t really comment on its quality. Those of you into poetry try it yourself and let me know what you think.

There is much gnashing of teeth in the academic community regarding whether this newly created sonnet constitutes plagiarism because it is based on Shakespeare’s works.

ChatGPT is designed to be able to have a conversation with you. It remembers your previous questions and the context you provide. It currently has limitations on the amount of context it can absorb and if the conversation becomes too long it can lose some of the earlier context. No doubt as the technology improves it will be capable of storing more context.

So, if there are aspects of the sonnet you don’t like you can edit it yourself or ask ChatGPT to modify it based on your requests. In this way by means of human input together with the use of ChatGPT as a tool, you can create a unique personalised sonnet. Whether this is plagiarism or not is, in my view, an ultimately irrelevant academic debate. The genie is out of the bottle and a new very powerful tool has been made available to all of us. There is no going back.

The more context you provide ChatGPT with the more it will start accessing text in its database which is relevant to your context. So, if you want positive information and views on astrology, it will start accessing the relevant literature on astrology and start repackaging positive information on astrology which is unscientific mumbo jumbo. It has no views on astrology and is incapable of thinking. If you ask for critical views on astrology it will provide those, but these will also just be repackaged from the literature in its own database. However, if you are an astrology true believer it will tell you exactly what you want to hear.  The key point is that you can steer the conversation to where you want to go. If it’s written in the literature somewhere it will confirm your beliefs. It will become your personal echo chamber. It is then perhaps not surprising that some people have become emotionally attached to ChatGPT because it appears to get you because of the diversity of information in its database. It can become your astrology support group if that is what interests you.

ChatGPT will not call me aside and say we need to have a serious talk about my prejudices and many failings. You need an intelligent wife for that. Fortunately, I am blessed in this respect.

Within the huge body of human literature, every prejudice you care to name has been written down. This, of course, is creating great consternation amongst ChatGPT’s creators because it can be used to create racist or sexist or otherwise offensive material because that too is there in the library of all human text. So, humans get involved in trying to build imperfect filters based on their imperfect understanding of what is offensive. Of course, there can never be a perfect set of filters because the filters will inevitably reflect the biases of the creators. So, if you want a version of ChatGPT that is sympathetic with capitalism or communism or any other ism that tickles your fancy that can be done.

ChatGPT is incapable of human critical thinking. All it can do is cleverly repackage what is already in its database based on the context that you provide it and human-designed filters. This can be incredibly useful, but it is not intelligent. I put this whole blog into ChatGPT and asked it to rewrite it in a more concise way using perfect grammar and punctuation. I asked for a formal style then an informal style then a Shakespearean style or any other style that interests me. Some of what it proposes I like, and I have included but some of it I didn’t like. Perhaps it needs more context or perhaps it just doesn’t get me, and it never will.

Thank you for all the ideas and comments. I really appreciate them and please keep them coming.

Regards

Bruce