Member-only story
Chatbots are Stochastic Parrots by Definition (and what it really means)
In this article I will comment about the Quanta Magazine article, where some researchers claim that Chatbots have skills beyond Stochastic Parrot’s based on funky evidence (which kind of shows they didn’t even understand the power of Stochastic Processes in general). Then I will show that the claim of Chatbots being more than Stochastic Parrots goes fundamentally against the foundations of computer science, however, this also shows the core of the problem: we need to separate the domains of computer science and human-computer interaction.
It seems like computer science as exact non-empirical mathematical science has lived way beyond it’s due date and what we actually want from modern software engineers would benefit more from philosophical, humanistic and empirical framing of computational problems, so they don’t mix-up the metaphysics of humans to the physics of computers.
Many claims of LLM capabilities are based on false attribution of capabilities used in computer-human interaction, which is not dissimilar to the observer effect from quantum mechanics. The historical reasons for why this is very typical are insteresting, but I will dive deeply to those in another article. There I will explore the full story of Cartesian-Newtonian apparatus and how it gives birth to digital computers and later…
