GPT-3 and InstructGPT: technological dystopianism, utopianism, and “Contextual” perspectives in AI ethics and industry

Abstract This paper examines the ethical solutions raised in response to OpenAI’s language model Generative Pre-trained Transformer-3 (GPT-3) a year and half from its release. I argue that hype fear about GPT-3, even within Natural Language Processing (NLP) industry AI ethics, have often been underpinned by technologically deterministic perspectives. These perspectives emphasise autonomy of rather than human actors systems. highlight existence current discourse (which range technological utopianism dystopianism), with specific focus on two issues of: (1) GPT-3’s potential intentional misuse for manipulation (2) unintentional harm caused bias. In response, find contextual approach which is centred upon wider ecologies societal benefit, autonomy, values, illuminates practical concerns Additionally, although newest 2022 InstructGPT represents small step reducing toxic aligning GPT-3 user intent, it does not provide any compelling or Therefore, address these must organisational settings as precondition decision-making AI, high-quality curated datasets less harmful outputs.