Originality and Uncertainty
 
 
 
 

I’ve just come across the site - https://dalledemo.com/ - in which you can enter a description (in words) of what you would like depicted and it will produce a picture for you. I entered the instruction: ‘an oil painting of a white mouse reading a book - in the style of Picasso’. It generated 4 images, the best of which is:

Not bad as a piece of cubist art, I thought.

I tried various other combinations of subjects and styles. The other picture shown here is its version of a Monet-style haystack. It has the right sort of colour pallet, although it’s a bit less impressionist and more realist than the master’s works. Both however are obviously a lot better than either I or most other people could paint.

Which of his  many haystack paintings it’s based on I have no idea, but the software is designed to to use the data it finds on the web. And I imagine that every one of his works is there somewhere.

As far as I know, though, Picasso didn’t actually paint a picture of a mouse reading a book and so the mouse image shows a striking capacity to handle data in a way that looks like creativity, but is in fact the ability to create a pastiche. We do not see originality.

What really made the news in AI recently, however, was ChatGPT. Commentators told us that school and university coursework was dead, because the software could do it all for us. I shan’t trouble you with the essay which I asked ChatGPT to write for me on the subject of freedom of speech. I awarded it 5 out of 10. It would make a rather dull encyclopaedia entry, providing nothing more than anodyne comments on the difficulties inherent in the concept.


But clearly AI is making progress - of a sort. The difficulty is that it is essentially plagiarising the data to be found on the world-wide web. It is not providing us with anything new. It is reshaping what is already there. And so I don’t expect to see a new and compelling school of art emerging anytime soon. Neither do I expect any novel philosophical progress from ChatGBT. That software even comes with a warning that its data resource stops at the end of 2021. Anything added to the web after that is not taken into account.

And the real killer is that it takes the string of words it's provided with and then from its immense database decides, on a statistical basis, which word should come next and then which word should follow that and so on. As a number of computer scientists have now said it is designed to produce text which is 'plausible' rather than accurate. Not a good start.

Neither is it capable of checking the data it’s using. It is regurgitating whatever it finds amongst the writings of all and sundry on the web, from the misognynist Andrew Tate, to the Pope, to Donald Trump. So then, I don’t think that thinking or indeed essay writing is quite dead - yet.


What is true, however, is that we are relying more and more on computers in so many aspects of our lives. They are there to increase output and provide us with more certainty that the decisions we make are right. In many respects, this has worked well. It does though depend on both data input and software being correct. Errors in commercial software require frequent updates for this reason and data input errors can go unnoticed for long periods – hence our general distrust of computers in the hands of government.

In science and engineering, however, where things come down to formulae, computers are certainly able to provide massive assistance. This is particularly so with quantum computers, which can engage in astonishing feats of parallel computing and so test multiple hypotheses at the same time. It means for example that new materials and their properties can be predicted at great speed and with immense accuracy.

We know that there is a distrust of the ability of computers to model the climate and so enable accurate predictions of global warming to be made. Fortunately, those models are being continuously modified and so the likelihood of significant error is diminished. But, although we can hope, we cannot be certain that the likelihood of error is negligible: the data which is not a part of the model may have an unexpectedly greater effect than we can know. Models are necessarily not full descriptions of any system and so by definition have unknown limitations.


With climate, however, we are at least still using physical quantities which can in principle be described by formulae. A greater difficulty arises when the impression of computer accuracy is presumed to apply to systems where formulae do not really apply.

Here, I am thinking of economics or, as we should describe it, the psychology of individuals and markets. I’m afraid we’re nowhere near being able to encapsulate any sort of psychology in formulae. And so all that computers can do is work on the basis of the assumptions of the economists. Those assumptions are based on what has happened in the past, in what they would describe as previous, similar economic cycles.

But imagining that history repeats itself is, itself, madness. It may seem that there was an inevitability about what we see as having happened in times past. Those who lived through it were not so sure. Just like us over the last few years, they (and we) had no idea how things would turn out. That unforeseen events, such as the arrival of a new virus, will happen from time to time is inevitable, but we know not when, where or how it will happen. Neither can we know what the consequences will be.

In this case, vaccines were developed in record time and deaths on the scale of the last major pandemic - Spanish flu in 1918 - averted. No economic model could have taken that into account.

Neither could it have taken into account the election of at least two hopeless Prime Ministers who would take decisions so damaging to our economy. After all, Liz Truss actually had supporters amongst academic economists. And Johnson’s views on Brexit are still supported by our existing Prime Minister and most of the party faithful.


It is therefore clear to me that we live in a world of uncertainty. This though leaves me in a strange position. Although I am someone who believes in a basically cause and effect, deterministic, world, actually to foretell with certainty what will happen next is impossible.

To do so would require being able to examine a model having every single one of the features of our real world i.e. a complete second version of it at atomic level. After all, the butterfly effect tells us that even a minor perturbation on one side of the world can have a major effect on the other side. But having examined the model to see the future, our knowledge of it would have to be taken into account and so we would need to amend the model to allow for that and so on.

So then, all that we can actually do is watch our personal history unfold. But many unforeseeable events which we see occur entail our making choices as to how we will react. In turn, our choices, tell us that we have taken an active part in our development. And it is not an illusion. But the apparent paradox remains: we cannot know what will happen next, but our choices are a part of an apparently deterministic chain that produces what happens next.

Messages, however, pass between our synapses at molecular level and so are subject to interference by quantistic, random, noise. It may be therefore that our thinking is actually interfered with by indeterministic, random quantum events. It is then, we hope, subjected to our reasoning processes and rejected or modified before being fed back into the chain of events. That would create an indeterministic, but rational decision personal to us
and so the possibility of original thought and creativity.

Whatever may be the case, I can look back at many turning points in my life. In the 6th form, at the age of 15, I took the unusual option in my school of learning a subject then quite foreign to me - philosophy and logic. Forty years ago, we were asked by a friend from Church (who happened to be Coleshill Town Clerk), in desperate need of hosts for French people (from what was to become our twin town), if we could possibly help. This led to my taking up language learning again after abject failure in school. That in turn led to writing essays with a philosophical bent including, of course, this one.

All of us have made decisions which have been key in our lives. The twists and turns are quite mind-boggling. I’m still not sure that I’ve managed to process it.


Paul Buckingham

11 January 2023
 

 
 

Home      A Point of View     Philosophy     Who am I?      Links     Photos of Annecy