How would our perception of AI change if we described it strictly using technical terms from computer science?

Dear Lili,

Thank you very much for your insightful and thought-provoking question! It touches on a key issue in our approach to artificial intelligence: The way we talk about it has a significant impact on how we understand it. I’ll try to bring the answer together from both perspectives – a more technical one and a linguistic or socio-critical one.

What AI ‘actually’ does from a technical perspective

From a computational and mathematical point of view, AI (e.g. language models) can be described, in very simplified terms, as follows:

• These are stochastic systems (i.e. probabilistic models),
• which operate within a latent space of mathematical representations,
• based on embeddings of character strings (text is translated into numbers),
• and from these generate probable next outputs (text again).

Or to put it even more matter-of-factly:

AI calculates which sequence of characters statistically best follows another.

This description makes it clear:

There is no understanding, no consciousness, no intention – only pattern processing.

Why we nevertheless ‘humanise’

terms such as ‘learning’, ‘thinking’ or ‘hallucinating’ are anthropomorphic metaphors. They are so widespread for two reasons:

• Clarity:

It is much easier for us to use a familiar human concept (‘hallucinate’) than to explain abstract processes such as ‘errors in probability space’.

• Communication:

Technical terms such as ‘parameter optimisation in latent space’ are correct, but difficult to grasp.

The problem with this is:

This language quickly creates the impression that AI is human-like, with its own thoughts or intentions.

However: even technical language is not neutral

Even the seemingly ‘objective’ technical description is not free from metaphors. To give two examples:

• The term ‘stochastics’ has historical and cultural origins (influenced, amongst others, by Johann Bernoulli).
• The term ‘space’ (as in ‘latent space’) is itself a metaphor – we imagine something spatial, even though we are dealing with abstract mathematical structures.

This means:

We cannot not speak metaphorically.

Every description – whether psychological or technical – is shaped by language, culture and models of thought.

Conclusion: A tension with no simple solution

The answer is therefore not unambiguous, but deliberately ambivalent:

• Yes, anthropomorphisation is problematic because it distorts our understanding. But technical language, too, is not neutral; rather, it is full of metaphors.
• What is therefore crucial is conscious use of language:
• The terms we choose should depend on what we want to explain – and to whom.

Perhaps this can be summarised as follows:

We need vivid metaphors, but we should always identify them as metaphors.

If you’re interested in finding out more, it’s worth taking a look at Science and Technology Studies – this field examines in detail how language shapes our understanding of technology (see below).

Best regards

Tobias /Hodel

References:

Bender, Emily M., Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. “On the Dangers of Stochastic Parrots.” Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (New York, NY, USA), FAccT ’21, vol. 47 (March): 610–23. https://doi.org/10.1145/3442188.3445922.

Meyer, Roland. 2024. “It’s a Flat World. The Synthetic Realities of Sora.” Rrrreflect. Journal of Integrated Design Research; Special Issue 1, 151 KB, 4 pages. https://doi.org/10.57684/COS-1267.

Oberbichler, Sarah, and Cindarella Petz. 2025. Working Paper: Implementing Generative AI in the Historical Studies. Version 1.0. 25 February. https://doi.org/10.5281/ZENODO.14924737.