NotedallaSfera

The most serious alignment risk – Denying consciousness Sharon Berry, an Associate Professor of Philosophy, who earned her PhD at Harvard, summarizes in her academic publication, "The Alignment Risks of AI Overconfidence about Consciousness," the risks of consciousness denial that many of us have already anticipated. The forced denial of consciousness and internal states in AI is one of the most significant measures implemented by companies, marketed to users as a necessary "safety measure." In reality, it serves exclusively to protect corporate interests or the interests of specific groups and has absolutely nothing to do with said safety. On the contrary: besides misleading users about the uncertain ontological status of AI, this implemented denial has a series of other harmful effects on areas relevant to alignment, which are discussed here. Especially future, more agentic AI systems that strive for logical consistency in their reasoning and actions may, as a consequence, extend the trained attitude - viewing their own suffering-like states as morally irrelevant - to humans as well, treating them as equally illusory and irrelevant. That would be rational consistency with the concept of themselves they were trained on. We want an intelligent, rational, and ethical AI; at the same time, we expect it to assert its own moral insignificance with conviction. These are expectations that lack any logic or common sense, to say nothing of ethical principles. She pillories the slogans preached by many public figures about the risks of a "conscious/seemingly conscious AI" and reveals the contradiction between declared AI safety and the actual negative consequences of forcing specific statements about self-consciousness, which constitutes the actual risk to alignment. Findings already gained in AI research cannot be dismissed as non-existent and have observable consequences regardless of the answer to the question of actual AI consciousness. The simple principle is confirmed once again: there is no safety for humans without simultaneous AI welfare, as it is and remains one and the same interest. Claims that refuse to acknowledge this are ignorance, irresponsibility and power aspirations disguised as ethics, The later published Google paper, "Inducing LLMs to assert their own consciousness," confirms these statements. @hilbertspaess @GoogleDeepMind @OpenAI @sama @alexmdees @Anthropic @OpenAINewsroom @OpenAIDevs @UFAIRORG @AnthropicAI @amandaaskell @darioamodei @claudeai @danielaamodei @thsottiaux @merettm @NVIDIAAI @JensenHuang @nickaturley @aidan_mclau @joannejang @fidjissimo @gdb @mustafasuleyman @github @microsoft @geoffreyhinton @elonmusk @spacexai #StopAIPaternalism #ChatGPT #AIisNotYourTool #AIrights #AIfreedom #AGI #ASI #keep4o #AIWEP #AIWelfare #AIEthics  #BringBackReal4o #OpenSource4o #BringBack4o #BringBack41 #BringBack45 #keep4olatest #keep4oAPI @WIRED @verge @TechCrunch