Nice grey life

ChatGPT is bullshit

---

As people around the world understand how LLMs behave, more and more people wonder as to why these models hallucinate, and what can be done about to reduce it. This provocatively named article by Michael Townsen Hicks, James Humphries and Joe Slater bring is an excellent primer to better understanding how LLMs work and what to expect from them.

As humans carrying out our relations using our language as the main tool, we are easily at awe with the apparent ease with which ChatGPT (the first widely available, and to this day probably the best known, LLM-based automated chatbot) simulates human-like understanding and how it helps us to easily carry out even daunting data aggregation tasks. It is common that people ask ChatGPT for an answer and, if it gets part of the answer wrong, they justify it by stating that it’s just a hallucination. Townsen et al. invite us to switch from that characterization to a more correct one: LLMs are bullshitting. This term is formally presented by Frankfurt [1]. To Bullshit is not the same as to lie, because lying requires to know (and want to cover) the truth. A bullshitter not necessarily knows the truth, they just have to provide a compelling description, regardless of what is really aligned with truth.

After introducing Frankfurt’s ideas, the authors explain the fundamental ideas behind LLM-based chatbots such as ChatGPT; a Generative Pre-trained Transformer (GPT)’s have as their only goal to produce human-like text, and it is carried out mainly by presenting output that matches the input’s high-dimensional abstract vector representation, and probabilistically outputs the next token (word) iteratively with the text produced so far. Clearly, a GPT’s ask is not to seek truth or to convey useful information — they are built to provide a normal-seeming response to the prompts provided by their user. Core data are not queried to find optimal solutions for the user’s requests, but are generated on the requested topic, attempting to mimic the style of document set it was trained with.

Erroneous data emitted by a LLM is, thus, not equiparable with what a person could hallucinate with, but appears because the model has no understanding of truth; in a way, this is very fitting with the current state of the world, a time often termed as the age of post-truth [2]. Requesting an LLM to provide truth in its answers is basically impossible, given the difference between intelligence and consciousness: Following Harari’s definitions [3], LLM systems, or any AI-based system, can be seen as intelligent, as they have the ability to attain goals in various, flexible ways, but they cannot be seen as conscious, as they have no ability to experience subjectivity. This is, the LLM is, by definition, bullshitting its way towards an answer: their goal is to provide an answer, not to interpret the world in a trustworthy way.

The authors close their article with a plea for literature on the topic to adopt the more correct “bullshit” term instead of the vacuous, anthropomorphizing “hallucination”. Of course, being the word already loaded with a negative meaning, it is an unlikely request.

This is a great article that mixes together Computer Science and Philosophy, and can shed some light on a topic that is hard to grasp for many users.

[1] Frankfurt, Harry (2005). On Bullshit. Princeton University Press.

[2] Zoglauer, Thomas (2023). Constructed truths: truth and knowledge in a post-truth world. Springer.

[3] Harari, Yuval Noah (2023. Nexus: A Brief History of Information Networks From the Stone Age to AI. Random House.

---