Google DeepMind's Senior Scientist Alexander Lerchner challenges the idea that large language models can ever achieve consciousness(not even in 100years), calling it the 'Abstraction Fallacy.'

In a significant intervention into one of AI's most heated philosophical debates, Google DeepMind senior research scientist Alexander Lerchner has publicly challenged the idea that large language models (LLMs) can achieve consciousness. Lerchner argues that the belief stems from an 'Abstraction Fallacy'—the mistaken assumption that because LLMs like GPT-4 or Llama 3 can generate coherent, seemingly understanding text, they must possess some form of internal subjective experience. He contends that these models are fundamentally sophisticated statistical engines for pattern prediction, not entities capable of feeling or awareness, and suggests this gap is not one of scale but of kind.

Lerchner's stance, which he asserts holds true even on a 100-year timeline, places him in direct opposition to other prominent thinkers in the field. Some theorists, often aligned with functionalist or emergentist views of consciousness, suggest that sufficiently complex information processing systems could give rise to subjective experience. By dismissing this possibility for LLMs, Lerchner's argument has immediate practical implications. It influences critical discussions around AI rights, ethical treatment, and the allocation of safety research resources, steering focus toward AI's measurable capabilities and impacts rather than speculative internal states.