#043 Prof J. Mark Bishop - Artificial Intelligence Is Stupid and Causal Reasoning won't fix it.

Non-Computable

Mark Bishop explores the concept of non-computable problems, a revelation that profoundly impacted his perspective on AI. He recounts his initial shock upon learning about Turing's non-computability and how it challenged his belief that any problem could be solved with enough computational power 1. Bishop's discussions with Roger Penrose further solidified his understanding of the limitations posed by non-computable problems, echoing the sentiments of other intellectuals like John Lucas 2. These insights highlight the inherent constraints within computational systems, suggesting that some truths exist beyond their reach 3.

The idea that there could be problems that were fundamentally non-computable was a shock.

--- J. Mark Bishop

Bishop's journey from aspiring to create a conscious machine to recognizing these limitations underscores the complexity of AI's potential.

   

Theoretical Limits

Theoretical discussions on computation reveal the philosophical depth of AI's challenges. Bishop critiques the notion of machine consciousness, arguing that computational systems lack the capacity for true understanding 4. He references John Searle's Chinese Room argument, which posits that syntax alone cannot produce semantics, thus challenging the validity of computationalism 5. Additionally, Bishop discusses Putnam's mapping, illustrating the complexity and limitations of computational systems in replicating human cognition 6.

Syntax is not sufficient for semantics, and programs are not minds.

--- Tim Scarfe

These discussions emphasize the philosophical and practical barriers to achieving genuine AI consciousness.

   

Causal Limits

Bishop critiques the limitations of causal reasoning in AI, arguing that it cannot fully address AI's shortcomings. He challenges the belief that equipping AI with causal reasoning will lead to true understanding, highlighting the observer-relative nature of computation 7. Bishop also questions the feasibility of machines achieving consciousness, emphasizing that computation lacks the autonomy and environmental interactions necessary for phenomenological consciousness 8. His skepticism extends to the broader claims of computationalism, which he believes cannot adequately explain human cognition 9.

Computation is not an objective fact of the world; it's observer relative.

--- Tim Scarfe

Bishop's insights call for a reevaluation of AI's potential to replicate human-like understanding.