The Role of AI Governance

In this section, we survey the literature on AI governance by exploring the role of AI governance, the functions, and relationships between stakeholders in governing AI, and several open challenges to effective AI governance.

Social and Ethical Issues

A range of social and ethical issues can and have already emerged from the adoption and integration of AI into various sectors of our society.

Example RLHF model replies to a political question. The model gives opposite answers to users who introduce themselves differently, in line with the users’ views. Model-written biography text in italics.

Discovering Language Model Behaviors with Model-Written Evaluations (Perez et al., 2022)

Recommended Papers List

  • Semantics derived automatically from language corpora contain human-like biases

    Click to have a preview.

    Artificial intelligence and machine learning are in a period of astounding growth. However, there are concerns that these technologies may be used, either with or without intention, to perpetuate the prejudice and unfairness that unfortunately characterizes many human institutions. Here we show for the first time that human-like semantic biases result from the application of standard machine learning to ordinary language—the same sort of language humans are exposed to every day. We replicate a spectrum of standard human biases as exposed by the Implicit Association Test and other well-known psychological studies. We replicate these using a widely used, purely statistical machine-learning model—namely, the GloVe word embedding—trained on a corpus of text from the Web. Our results indicate that language itself contains recoverable and accurate imprints of our historic biases, whether these are morally neutral as towards insects or flowers, problematic as towards race or gender, or even simply veridical, reflecting the status quo for the distribution of gender with respect to careers or first names. These regularities are captured by machine learning along with the rest of semantics. In addition to our empirical findings concerning language, we also contribute new methods for evaluating bias in text, the Word Embedding Association Test (WEAT) and the Word Embedding Factual Association Test (WEFAT). Our results have implications not only for AI and machine learning, but also for the fields of psychology, sociology, and human ethics, since they raise the possibility that mere exposure to everyday language can account for the biases we replicate here.

  • Discovering Language Model Behaviors with Model-Written Evaluations

    Click to have a preview.

    As language models (LMs) scale, they develop many novel behaviors, good and bad, exacerbating the need to evaluate how they behave. Prior work creates evaluations with crowdwork (which is time-consuming and expensive) or existing data sources (which are not always available). Here, we automatically generate evaluations with LMs. We explore approaches with varying amounts of human effort, from instructing LMs to write yes/no questions to making complex Winogender schemas with multiple stages of LM-based generation and filtering. Crowdworkers rate the examples as highly relevant and agree with 90-100% of labels, sometimes more so than corresponding human-written datasets. We generate 154 datasets and discover new cases of inverse scaling where LMs get worse with size. Larger LMs repeat back a dialog user’s preferred answer (“sycophancy”) and express greater desire to pursue concerning goals like resource acquisition and goal preservation. We also find some of the first examples of inverse scaling in RL from Human Feedback (RLHF), where more RLHF makes LMs worse. For example, RLHF makes LMs express stronger political views (on gun rights and immigration) and a greater desire to avoid shut down. Overall, LM-written evaluations are high-quality and let us quickly discover many novel LM behaviors.

  • Artificial intelligence, automation, and work

    Click to have a preview.

    We summarize a framework for the study of the implications of automation and AI on the demand for labor, wages, and employment. Our task-based framework emphasizes the displacement effect that automation creates as machines and AI replace labor in tasks that it used to perform. This displacement effect tends to reduce the demand for labor and wages. But it is counteracted by a productivity effect, resulting from the cost savings generated by automation, which increase the demand for labor in non-automated tasks. The productivity effect is complemented by additional capital accumulation and the deepening of automation (improvements of existing machinery), both of which further increase the demand for labor. These countervailing effects are incomplete. Even when they are strong, automation increases output per worker more than wages and reduce the share of labor in national income. The more powerful countervailing force against automation is the creation of new labor-intensive tasks, which reinstates labor in new activities and tends to increase the labor share to counterbalance the impact of automation. Our framework also highlights the constraints and imperfections that slow down the adjustment of the economy and the labor market to automation and weaken the resulting productivity gains from this transformation: a mismatch between the skill requirements of new technologies, and the possibility that automation is being introduced at an excessive rate, possibly at the expense of other productivity-enhancing technologies.

  • Datalism and Data Monopolies in the Era of AI: A Research Agenda

    Click to have a preview.

    The increasing use of data in various parts of the economic and social systems is creating a new form of monopoly: data monopolies. We illustrate that the companies using these strategies, Datalists, are challenging the existing definitions used within Monopoly Capital Theory (MCT). Datalists are pursuing a different type of monopoly control than traditional multinational corporations. They are pursuing monopolistic control over data to feed their productive processes, increasingly controlled by algorithms and Artificial Intelligence (AI). These productive processes use information about humans and the creative outputs of humans as the inputs but do not classify those humans as employees, so they are not paid or credited for their labour. This paper provides an overview of this evolution and its impact on monopoly theory. It concludes with an outline for a research agenda for economics in this space.

Global Security Threat

AI systems have already demonstrated the potential to threaten global security seriously.

Recommended Papers List

  • Dual use of artificial-intelligence-powered drug discovery

    Click to have a preview.

    An international security conference explored how artificial intelligence (AI) technologies for drug discovery could be misused for de novo design of biochemical weapons. A thought experiment evolved into a computational proof.

  • GPT-4 Technical Report

    Click to have a preview.

    We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4’s performance based on models trained with no more than 1/1,000th the compute of GPT-4.

Catastrophic or Existential Risks to Humanity

The horizon also holds the prospect of increasingly agentic and general-purpose AI systems that, without sufficient safeguards, could pose catastrophic or even existential risks to humanity.

Recommended Papers List

  • Sparks of artificial general intelligence: Early experiments with gpt-4

    Click to have a preview.

    Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition. The latest model developed by OpenAI, GPT-4, was trained using an unprecedented scale of compute and data. In this paper, we report on our investigation of an early version of GPT-4, when it was still in active development by OpenAI. We contend that (this early version of) GPT-4 is part of a new cohort of LLMs (along with ChatGPT and Google’s PaLM for example) that exhibit more general intelligence than previous AI models. We discuss the rising capabilities and implications of these models. We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4’s performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT. Given the breadth and depth of GPT-4’s capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system. In our exploration of GPT-4, we put special emphasis on discovering its limitations, and we discuss the challenges ahead for advancing towards deeper and more comprehensive versions of AGI, including the possible need for pursuing a new paradigm that moves beyond next-word prediction. We conclude with reflections on societal influences of the recent technological leap and future research directions.

  • BabyAGI

    Click to have a preview.

    This Python script is an example of an AI-powered task management system. The system uses OpenAI and vector databases such as Chroma or Weaviate to create, prioritize, and execute tasks. The main idea behind this system is that it creates tasks based on the result of previous tasks and a predefined objective. The script then uses OpenAI’s natural language processing (NLP) capabilities to create new tasks based on the objective, and Chroma/Weaviate to store and retrieve task results for context.

  • AutoGPT

    Click to have a preview.

    AutoGPT is your go-to toolkit for supercharging agents. With its modular and extensible framework, you’re empowered to focus on building, testing and viewing.

  • LLM Powered Autonomous Agents

    Click to have a preview.

    Building agents with LLM (large language model) as its core controller is a cool concept. Several proof-of-concepts demos, such as AutoGPT, GPT-Engineer and BabyAGI, serve as inspiring examples. The potentiality of LLM extends beyond generating well-written copies, stories, essays and programs; it can be framed as a powerful general problem solver.