There’s been a lot about AI in the news lately, though so far as I know, not much has been written about the relationship between AI and the doctrine of transubstantiation. Granted, some people have commented on ChatGPT’s ability to produce a perfectly sound account of this doctrine, but I’m not aware of anyone drawing a direct connection between the nature of artificial intelligence itself and transubstantiation. But I think there is a connection, and this is to do with what is meant by the word artificial and what can be accomplished artificially.
St Thomas Aquinas considers this question in the Summa Theologiae in response to an objection concerning transubstantiation. The doctrine of transubstantiation depends on a distinction between substance and accidents. Substance refers to what can exist in itself, so things like water, trees, and human beings are substances. Accidents, on the other hand, refer to what can normally only exist in something else.
For example, colour exists in a ball, so colour is an accident. Now the doctrine of transubstantiation depends on bread being a substance: at the consecration, the substance of bread becomes the substance of Christ’s body, while all the accidents such as the taste of bread and its colour and shape remain. So this means that although the host looks and tastes like bread, it is not bread at all, but is in reality Christ’s body.
Now one of the objections to this doctrine is that bread is by nature something artificial, for to be artificial is just to be something humans produce via some skill (literally something made by art). The objector claims that in the production of artificial things, we do nothing more than introduce some accidental form into the things we produce. For instance, in the production of a bronze statue, the sculptor only changes the shape of the bronze, its new shape being an accidental form – the sculptor doesn’t change the bronze into another substance. The objector then goes on to claim that since bread is artificial, all the baker does is impose the accidental form of bread onto the flour and water out of which it is made. The objector thus concludes that after the consecration, the host is still bread in the same way as the host is still round since the form of bread and the form of roundness are accidental forms.
In response to this objection, St Thomas Aquinas argues that in the production of artificial things, we can do more than introduce accidental forms to the things we produce. For by harnessing the powers of nature, it may be possible for us to bring about a change of substance. So in the case of bread making, by harnessing the power of heat, we can change the substances of flour and water into the substance of bread. And it is this substance of bread that by God’s power becomes the substance of Christ’s body at the consecration.
Now if we can produce new substances by harnessing the powers of nature, it doesn’t seem that implausible that we could one day manufacture a substance that possesses intelligence. One way, albeit highly immoral, would be to artificially synthesise the human genome and use this to create a human being without biological parents. Such a person would be intelligent but also artificial insofar as the person was produced via scientific skill. The person would therefore be an example of artificial intelligence.
So what about ChatGPT? Is this an example of artificial intelligence? I would say no, and for two reasons. Firstly, it seems that most of the work involved in producing something like ChatGPT only requires the configuration of components, and so the engineers are only introducing accidental forms to the components out of which ChatGPT is made. It therefore doesn’t have the requisite unity to be a genuine subject of intelligence.
But my second reason for denying that ChatGPT is intelligent is because of my experiences of chatting with it. An important sign of intelligence is the ability to apply abstract principles to particular situations, but as far as I can tell, ChatGPT lacks this ability. For instance, although ChatGPT can churn out a definition of what a valid argument is (namely, an argument in which the truth of its conclusion follows from the truth of its premises) ChatGPT finds it very difficult to evaluate whether a particular argument is valid. What ChatGPT excels at is producing answers to questions that sound intelligent. But sounding intelligent is very different from genuine intelligence.
Recently, the so-called “godfather of AI”, Geoffrey Hinton, resigned from Google because of his fears that biological intelligence will soon be replaced by digital intelligence. I don’t share his fear, primarily because I don’t think there is such a thing as digital intelligence. But it is nevertheless very dangerous for society if we start believing something is intelligent when in reality it is devoid of intelligence. When it comes to producing artificial things, humanity is currently much better at producing bread than producing intelligence.
Areas of Catholic Herald business are still recovering post-pandemic.
However, we are reaching out to the Catholic community and readership, that has been so loyal to the Catholic Herald. Please join us on our 135 year mission by supporting us.
We are raising £250,000 to safeguard the Herald as a world-leading voice in Catholic journalism and teaching.
We have been a bold and influential voice in the church since 1888, standing up for traditional Catholic culture and values. Please consider donating.