Yes, right now, I will ask you about the price of gold, it would not be too difficult to solve the issue. You would probably put it in Google or some search engine and you would have it, or maybe you would solve it through some Generative artificial intelligencehow ChatGPT. However, everything would change if I asked you for the value because, or Chatgpt No Sabe to clarify iteither vilely deceives you. And, value and price are not the same, and this for generative AI should be a serious problem.
The price of things is the monetary or in kind convention that we have achieved to determine equivalences. It must be concrete and can be negotiablebut we talk about an amount. However, when we talk about value, we talk about a very different predicate, because we are talking about the non-indifference that the world generates in us, which entails a hierarchy of what is part of the world. Just as in the price we can know the convention that specifies a precise calculation, in the case of the values we are attended that they are not demonstrable as demonstrated or broken down by the price of an object. The same thing is to give value in the world and generate against value in contrast. And this generates a problem for generative AI: The determination of what is valuable or of what is not.
The own thing is to give value in the world and generate against value as opposed
Now, is it possible to expose something without apply values? Certainly, at the time we take different sources to elaborate a content we are already applying A value to a source above the one we assume for other sources, so that we prefer some sources to others depending on the value we apply. The “neuronal networks” of the generative AI, with their tokens and their algorithms, constantly execute this process, applying a series of values to the production of information. Perhaps this for the reader is not problematic, but here we have the germ of a highly dangerous process.
Is it possible to expose something without applying values?
The human being is capable of discover values and apply them from their own freedom, and is even able to ignore these same values. The ability to value is inherent to the human being, and this is something that leads him, almost automatically, to concatenate value judgments that allow him to discover the world and its configuration. This is unavoidable to human activity and we assume it in our personal interactions. However, how does the assumption of values manage the generative AI? Although this question may seem of minor importance, in it we play the credibility we give to the work carried out by this tool of high effectiveness in the development of our daily lives.
The ability to value is inherent in the human being
Values cannot be inherent to the information transmitted by the generative AI, although they are contained in that information. That is why, necessarily, AI must do an assessment process when preparing the requested contents. Now, Who has transmitted these values? Where do they come from? Has she given them herself? Obviously not, as long as she herself gives an exponential improvement in content elaboration, but cannot generate functions beyond what she is. If this is so, the values are transmitted to them and, with them, in a subtle way, informative biases that are transmitted to us in imperceptible ways, or sometimes, absurdly.
How many subtle biases apply in the generative AI without us noticing?
I remember an example of a youtuber specialized in computer science that tells how he asked a generative of images a Illustration of a Nazi soldier And, since the tool was scheduled to be inclusive, it generated an image of a Nazi whose skin color was opposed to what we understand by “ario.” The error was not of AI, since she had to generate inclusive content, but, without knowing, generated an ironic anachronism because of the inclusive bias with which she had been programmed. This example is of the gross, but how many subtle biases apply in generative AI without us noticing?
We all know that, if we trained a generative AI with the journalistic contents of the Spanish press, the result would vary greatly depending on the training field. In the same way, the assumption by the generative information is not something innocent or integral: our assistant, together with the information provided, also transmits us Elements hidden in metaversus That, beyond cyber conspiracies, they configure us more than us.
Tomorrow, the generative AI advances in the management of values
Tomorrow, the generative AI advances in the management of values and how they influence the information it generates, or that its understanding is increasingly sophisticated. However, the exercise of freedom that claims the assumption of values is something that I do not think is on the horizon of this tool that, although it is highly sophisticated, I doubt it is able to determine what has more value: if a walk along the beach or a film afternoon, regardless of the price.
- Domingo Pacheco He is a diocesan priest of Valencia, chaplain and director of the Joseph Ratzinger Theology Chair of the CEU Cardenal Herrera University and Diocesan consiliary of Juniors MD Degree in Historical Theology and Master in Ethics and Democracy. It is also coordinator of the reflection group on the area of Philosophy and Ethics of the CEU.
