GenAI & The Dual-Deception Process in Web Science

Rapid adoption of GenAI has sparked interest once again in matters related to not just machine intelligence but human intelligence as well. Unfortunately, GenAI has created what can be considered a dual-deception process. On one hand, since Large Language Models (LLMs) output hallucinations and inaccurate information, these thereby deceive end users with content in response to prompts, queries, etc. The specific form of deception at work here is pandering; and the specific form of pandering in question derives from the fact that LLMs “want” the perceived agendas of those in their audience to be advanced, independent of truth. On the other hand, since the public has largely accepted the idea that this latest wave of Machine Learning (ML) will lead us to Artificial General Intelligence (AGI), i.e., to machine intelligence that matches human intelligence, humans are being deceived as to what genuine intelligence in our case is. Many are now convinced that these LLMs think as humans do, despite the inability of LLMs to reason abstractly, on novel information, not chained in any way to prior patterns enshrined in data regarding what has been produced in the past. This dual-deception process, enabled by the Web, should be analyzed and, perhaps, in some way regulated.