Sam Altman, the CEO of artificial intelligence lab OpenAI, told a Senate panel he welcomes federal regulation on the technology "to mitigate" its risks. Sam Altman, the CEO of artificial intelligence lab OpenAI, told a Senate panel he welcomes government regulation on the technology "to mitigate" its risks. "As this technology advances, we understand that people are anxious about how it could change the way we live. But we believe that we can and must work together to identify and manage the potential downsides so that we can all enjoy the tremendous upsides. It is essential that powerful AI is developed with democratic values in mind. And this means that U.S. leadership is critical," Altman said Tuesday.

The CEO of OpenAI, the company responsible for creating artificial intelligence chatbot ChatGPT and image generator Dall-E 2, said "regulation of AI is essential" on Tuesday as he testified in front of a Senate judiciary committee panel. In his first appearance in front of Congress, Sam Altman said he supported regulatory guardrails for the technology that would enable the benefits of artificial intelligence while minimizing the harms. "We think that regulatory intervention by governments will be critical to mitigate the risks of increasingly powerful models," Altman said in his prepared remarks. "For example, the US government might consider licensing and testing requirements for development and release of AI models above a threshold of capabilities." Altman and Gary Marcus, emeritus professor of psychology and neural science at New York University, both called for a new regulatory agency for the technology.

OpenAI CEO Sam Altman discussed the risks and benefits of AI at a Senate Judiciary subcommittee hearing on May 16, 2023. OpenAI CEO Sam Altman told a panel of senators Tuesday that his greatest fear as his company develops artificial intelligence capabilities is that is causes major harmful disruption for people, and acknowledged that AI has this potential downside if it isn't properly regulated. "My worst fears are that we cause significant – we, the field, the technology industry – cause significant harm to the world," Altman told a Senate Judiciary subcommittee. "I think that could happen in a lot of different ways. It's why we started the company."

The Biden administration is increasingly calling AI an important priority, and there are growing efforts on Capitol Hill to draft legislation addressing the technology. Senate Majority Leader Charles E. Schumer (D-N.Y.) has been developing a new AI framework, which would "deliver transparent, responsible AI while not stifling critical and cutting edge innovation."

"I can tell you that there is discussions going on in all of the guilds, all of the agencies, and all of the legal firms in order to come up with the legal ramifications of my face and my voice and everybody else's being our intellectual property," Hanks added.

Google's approach is to introduce these new functions into its products gradually. But it will most likely be just a matter of time before things start to go awry. The company has not solved any of the common problems with these AI models. They still make stuff up. They are still easy to manipulate to break their own rules.

Chris Winfield, founder of Understanding A.I., tells'Fox & Friends Weekend' host Will Cain about a study showing patients preferred medical answers from artificial intelligence over doctors. Artificial intelligence fails to match humans in judgment calls and is more prone to issue harsher penalties and punishments for rule breakers, according to a new study from MIT researchers. The finding could have real world implications if AI systems are used to predict the likelihood of a criminal reoffending, which could lead to longer jail sentences or setting bail at a higher price tag, the study said. Researchers at the Massachusetts university, as well as Canadian universities and nonprofits, studied machine-learning models and found that when AI is not trained properly, it makes more severe judgment calls than humans. Human participants then labeled the photos or text, with their responses used to train AI systems.

In-context learning is a new learning paradigm where a language model conditions on a few input-output pairs (demonstrations) and a test input, and directly outputs the prediction. It has been shown highly dependent on the provided demonstrations and thus promotes the research of demonstration retrieval: given a test input, relevant examples are retrieved from the training set to serve as informative demonstrations for in-context learning. While previous works focus on training task-specific retrievers for several tasks separately, these methods are often hard to transfer and scale on various tasks, and separately trained retrievers incur a lot of parameter storage and deployment cost. In this paper, we propose Unified Demonstration Retriever (\textbf{UDR}), a single model to retrieve demonstrations for a wide range of tasks. To train UDR, we cast various tasks' training signals into a unified list-wise ranking formulation by language model's feedback. Then we propose a multi-task list-wise ranking training framework, with an iterative mining strategy to find high-quality candidates, which can help UDR fully incorporate various tasks' signals. Experiments on 30+ tasks across 13 task families and multiple data domains show that UDR significantly outperforms baselines. Further analyses show the effectiveness of each proposed component and UDR's strong ability in various scenarios including different LMs (1.3B - 175B), unseen datasets, varying demonstration quantities, etc.

We are currently unable to specify human goals and societal values in a way that reliably directs AI behavior. Law-making and legal interpretation form a computational engine that converts opaque human values into legible directives. "Law Informs Code" is the research agenda embedding legal knowledge and reasoning in AI. Similar to how parties to a legal contract cannot foresee every potential contingency of their future relationship, and legislators cannot predict all the circumstances under which their proposed bills will be applied, we cannot ex ante specify rules that provably direct good AI behavior. Legal theory and practice have developed arrays of tools to address these specification problems. For instance, legal standards allow humans to develop shared understandings and adapt them to novel situations. In contrast to more prosaic uses of the law (e.g., as a deterrent of bad behavior through the threat of sanction), leveraged as an expression of how humans communicate their goals, and what society values, Law Informs Code. We describe how data generated by legal processes (methods of law-making, statutory interpretation, contract drafting, applications of legal standards, legal reasoning, etc.) can facilitate the robust specification of inherently vague human goals. This increases human-AI alignment and the local usefulness of AI. Toward society-AI alignment, we present a framework for understanding law as the applied philosophy of multi-agent alignment. Although law is partly a reflection of historically contingent political power - and thus not a perfect aggregation of citizen preferences - if properly parsed, its distillation offers the most legitimate computational comprehension of societal values available. If law eventually informs powerful AI, engaging in the deliberative political process to improve law takes on even more meaning.

The HuggingFace Datasets Hub hosts thousands of datasets, offering exciting opportunities for language model training and evaluation. However, datasets for a specific task type often have different schemas, making harmonization challenging. Multi-task training or evaluation necessitates manual work to fit data into task templates. Several initiatives independently tackle this issue by releasing harmonized datasets or providing harmonization codes to preprocess datasets into a consistent format. We identify patterns across previous preprocessing efforts, such as column name mapping and extracting specific sub-fields from structured data in a column. We then propose a structured annotation framework that ensures our annotations are fully exposed and not hidden within unstructured code. We release a dataset annotation framework and dataset annotations for more than 500 English tasks\footnote{\url{https://github.com/sileod/tasksource}}. These annotations include metadata, such as the names of columns to be used as input or labels for all datasets, which can save time for future dataset preprocessing, regardless of whether our framework is utilized. We fine-tune a multi-task text encoder on all tasksource tasks, outperforming every publicly available text encoder of comparable size in an external evaluation.