After the presentation led by Professor Claus Beisbart from the University of Bern, I became particularly curious about his reference to the notion of “understanding” and how it differs from knowledge. To explore this further, I looked up several resources online and discovered the following:
“Understanding is the knowledge about a subject, situation, etc., or about how something works” 1
However, when I searched for the “philosophical notion of understanding”, I found the following:
“Understanding seems to be different than knowledge in both respects. On the one hand, understanding typically seems harder to acquire, and more of an epistemic accomplishment, than knowledge2. On the other hand, the objects of understanding seem more structured and interconnected3” 4
Professor Beisbart also explained that “understanding can be spelled out in terms of ability: grasping connections, giving explanations, reasoning based on explanations (e.g., counterfactual history, exploring alternatives), and applying insights in practice (e.g., building an instrument based on the understanding of a module).” As this statement suggests, understanding is a multifaceted and complex concept encompassing knowledge, contextualization, interpretation, and application.
At this point, I asked myself: “Does AI understand? To what extent does it understand, and in what ways can it assist human understanding?”
This blog post will further develop these concepts.
Does AI understand?
This question, which is deeply philosophical, lies at the core of debates in various fields, such as philosophy, artificial intelligence, and cognitive science. Understanding, as defined by the Stanford Encyclopedia of Philosophy and Professor Beisbart, is a complex process that involves many areas. AI, which includes machine learning models and natural language processing tools, does not understand in the same way humans do. In reality, AI systems process information and make decisions based on data patterns; they lack proper comprehension, consciousness, awareness, or the ability to reflect on their processes.
Image 1:
If understanding is seen as a process that involves consciousness, emotions and subjective experience, then it is far from the results that AI systems can achieve:
“A veneer of linguistic facility is not the same as actually comprehending human language”5
“Understanding language requires understanding the world, and a machine exposed only to language cannot gain such an understanding”6
AI responses are based on the data and algorithms that support them, allowing them to make sense of conversations. However, this process does not involve consciousness or opinions, which AI lacks. The programming and input data enable AI to make predictions and respond to questions, translate, summarize texts and perform other tasks. This can sometimes lead to misinterpretation, as AI systems do not have emotions, making it difficult for them to fully grasp complex human feelings and states of mind or thoughts and perceptions. While AI interacts in meaningful ways and is becoming increasingly sophisticated, the division between AI and human understanding remains clearly visible.
To what extent does it understand, and in what ways can it assist human understanding?
The appeal of AI lies in its promises to improve outcomes, reduce the necessary time to do it and overcome human shortcomings and biases. However, what we are experiencing with AI is a vast production of results with limited understanding. They are machine learning systems, so their work consists of processing data to generate outputs. The impressive capabilities of these models are simply the result of predictions they are trained to make, but they lack a coherent understanding of the world. In fact, they can perform specific tasks well without fully grasping the underlying rules, but when it comes to real-world applications and scientific problems, the situation is quite different.
AI systems have significant limitations in assisting human understanding due to their lack of contextual awareness. They are also limited by the quality of the data they have been trained on, which can be incomplete or carry inherent biases. Furthermore, they sometimes oversimplify complex contexts or hallucinate incorrect information.
There is a sector that has been deeply changed by AI: the language translation industry. Machine learning systems have revolutionized this sector, shaping the future of the industry with their immense capabilities. In fact, AI has opened the way for new career possibilities without leaving human workers out of the equation.
“AI still has some ways to go to overcome issues with context, colloquialisms, and tone of voice”7
The truth is that, nowadays, even the best language translation systems cannot work independently because they lack crucial features.
Image 2:
In conclusion, it is also important to acknowledge that AI has proven to be a powerful tool that aids humans in several ways due to its speed, its ability to identify patterns and correlations, its support for understanding contexts in languages people do not speak fluently, and its capacity to clarify complex concepts in general. Moreover, they can help expand and deepen human connections, enabling us to gain an immense understanding of people across the globe.
Bibliography:
Unpublished:
Presentation given by Professor Claus Beisbart on the 18.11.2024
Internet literature:
https://dictionary.cambridge.org/dictionary/english/understanding
https://plato.stanford.edu/entries/understanding
https://www.quantamagazine.org/what-does-it-mean-for-ai-to-understand-20211216/
https://www.nature.com/articles/s41586-024-07146-0
https://news.mit.edu/2024/generative-ai-lacks-coherent-world-understanding-1105
https://www.getblend.com/blog/artificial-intelligence-changing-the-translation-services-industry
Image sources:
Picture 1: https://www.genetec.com/blog/cybersecurity/the-implications-of-large-language-models-in-physical-security
Picture 2: https://uxdesign.cc/how-to-really-understand-human-48b5b79cb371
- Definition of Understanding, Cambridge English Dictionary https://dictionary.cambridge.org/dictionary/english/understanding ↩︎
- (Pritchard 2010) ↩︎
- (Zagzebski 2019) ↩︎
- Definition of Understanding, Stanford Encyclopedia of Philosophy
https://plato.stanford.edu/entries/understanding/ ↩︎ - What Does It Mean for AI to Understand?, Quanta Magazine (Mitchell, 2021) ↩︎
- What Does It Mean for AI to Understand?, Quanta Magazine (Mitchell, 2021) ↩︎
- Artificial intelligence changing the translation industry, Getblend (Habash, 2023). ↩︎
OpenEdition schlägt Ihnen vor, diesen Beitrag wie folgt zu zitieren:
valentinapellini (6. Dezember 2024). What contributions can AI make to understanding in the humanities? Einblicke in die Digital Humanities. Abgerufen am 24. Januar 2025 von https://doi.org/10.58079/12v27
10 Antworten auf „What contributions can AI make to understanding in the humanities?“
I find your blog post really interesting and inspiring! In particular, I enjoyed reading the part about language translation systems, as I study foreign languages for communication and business. It’s reassuring to hear that human contribution in this field is still important, even though AI is rapidly advancing and sometimes replacing human work.
This topic resonates with me personally, as I believe that language is not just about words but also about culture, context, and emotion—things that are difficult for AI to fully grasp. While AI can handle straightforward tasks efficiently, human interpreters bring nuance, empathy, and cultural understanding to their work.
I’m also curious about how this will evolve in the future. Will AI tools and human professionals find a way to complement each other more seamlessly?
Moreover, I think translation plays a vital role in fostering intercultural communication. It’s not only about converting words but also about building bridges between people and cultures. In this sense, the human touch seems irreplaceable to me.
Thank you so much for your response! I am glad you found the blog interesting and that it resonated with your studies as well.
I completely agree with your point about language being more than just words, these are indeed complex aspects that are challenging for AI to fully grasp. I also completely understand your curiosity about how AI and human professionals might complement each other seamlessly in the future, and I found it very insightful.
Thanks again for your reflections! It is always great to hear different point of view on these important topics.
Your exploration of understanding versus knowledge is profound and nuanced. The philosophical distinction you’ve drawn between AI’s computational processing and genuine human comprehension is particularly insightful. While AI can simulate understanding through sophisticated pattern recognition, it fundamentally lacks the subjective, emotional, and contextual depth that characterizes human cognition.
Your analysis rightly emphasizes that true understanding isn’t just about information retrieval or pattern matching, but about meaningful interpretation, emotional intelligence, and the ability to reflect on complex experiences. The quotes you’ve selected beautifully illustrate how AI, despite its impressive capabilities, remains a tool that augments rather than replicates human understanding.
The translation industry example is especially compelling – it demonstrates both AI’s potential and its limitations. AI can process and translate at remarkable speeds, but still struggles with the subtle nuances of human communication.
Thank you for this thought-provoking piece that invites us to critically examine the nature of understanding in the age of artificial intelligence.
Thank you so much for your thoughtful and insightful response! I am really glad that you found the exploration of understanding versus knowledge interesting and meaningful, and I truly appreciate the depth of your reflections on the topic.
You make an excellent point about the philosophical distinction between AI’s computational processing and genuine human comprehension, and your observation really resonates with me.
I am also happy to hear that the translation industry example struck a chord with you.
Thank you again for your kind words and for taking the time to engage with the article so thoroughly.
Thank you for the interesting blog post. It made me think of my initial experiences with large language models when I occasionally wondered whether a model had any understanding including that of my inputs and its outputs. Those with more technical background have reassured me that the models neither “understand” nor “realise” anything.
Although much clearly depends on how understanding is defined or interpreted, I remain convinced that it is not an attribute of AI, let alone consciousness or emotion. The operations are based on data patterns and statistical computations, with the quality and quantity of data playing a critical role.
Thank you for your response! I am glad that the blog post made you reflect on your own experiences with large language models. It is definitely a fascinating topic to explore, especially when it comes to the question of whether these models truly understand.
I completely agree with your point that the models operate based on data patterns and statistical computations, and that their output is a result of these complex processes.
Your perspective on how understanding is defined is particularly important and I think that we will continue to explore the distinction between human cognition and AI’s computational processes.
Thank you again for your thoughtful reflections.
The blog post convincingly demonstrates how carefully and multi-layered Claus Beisbart’s presentation on the question ‘Artificial humanities? What the use of AI means for the humanities’ unfolds his line of reasoning. The conclusion: ‘[W]hat we are experiencing with AI is a vast production of results with limited understanding,’ seems to be a central point. At its best, AI can support us in making analyses, forecasts and control processes more efficient, reliable and also more diverse in both the natural sciences and the humanities, in medicine (including in the early detection and treatment of cancer and other fatal diseases), in the sustainable management of resources in the field of energy, and in cultural exchange in facilitating our communication across language barriers. However, as Beisbart points out, this is not enough to understand. In my opinion, a central point of his argument is the following: Understanding means placing oneself in relation to what one seeks to understand or thinks one understands. It is about positioning oneself, finding a position within an equally intellectual, cultural as well as physical, material and social space, in which meaning in an empathic sense can first emerge. It is about connecting with what is present and what is missing, with people who are in search of (gender)equitable, life-enhancing, inclusive practices and cultural diversity. Solutions to problems, results, explanations and justifications are essential for this. Furthermore, it is about participating in a collective practice that not only expands our knowledge, but also contributes to placing our knowledge, our intuitions and our behavior in a humane, meaning-generating context. We should also reasonably allow ourselves to be supported in this by AI processes. A (self-)critical awareness is essential here: we must carefully examine whether AI applications help to cultivate free will and not – due to an incomprehensible lack of concern or naivety or a seductive political calculation – restrict or even destroy it.
Thank you so much for your detailed response! I truly appreciate the depth of your reflections on Professor Claus Beisbart’s presentation and the way you have unpacked his key argument about AI’s role in the humanities and beyond.
You have highlighted an essential point, that of AI remaining fundamentally limited when it comes to fostering true understanding.
I also resonate with your interpretation of Beisbart’s argument, that understanding involves positioning oneself within an intellectual, cultural, and social space. It is about relating to the world, to others, and to our sense of meaning, which AI cannot fully replicate.
It is incredibly enriching to engage in such a nuanced conversation about the role of AI in shaping not only our knowledge but also the way we connect with the world around us.
Thanks again!
Thank you for your article.
I would like to draw attention to a point that I think was mentioned by Pr. Beisbart in his presentation, that is, how can we characterize human understanding in a way that makes it radically different from AI’s way of processing data.
It could be argued that there is no more deep understanding of meaning in humans as there is in the simplest algorithms. Does our use of language imply a deeper understanding of it?
This questioning has been eye-opening to me, and I feel that it really provides a new way to look at the debate around machine learning and “true” undestanding.
Thank you for your response! I am glad that the article sparked such deep reflection, and I appreciate your question about the nature of human understanding compared to AI’s data processing.
As you suggest, this question invite us to reconsider what we mean by “true” understanding, and this is where the philosophical debate gets particularly interesting. I think this is a question we must continue to explore as AI technologies evolve. While machines may process data with incredible efficiency, the subjective, embodied nature of human understanding might still be what sets us apart.
Thank you again for bringing such a thought-provoking angle to the discussion.