We propose Reinforcement Learning from Contrast Distillation (RLCD), a method for aligning language models to follow natural language principles without using human feedback. RLCD trains a preference model using simulated preference pairs that contain both a high-quality and low-quality example, generated using contrasting positive and negative prompts. The preference model is then used to improve a base unaligned language model via reinforcement learning. Empirically, RLCD outperforms RLAIF (Bai et al., 2022b) and context distillation (Huang et al., 2022) baselines across three diverse alignment tasks--harmlessness, helpfulness, and story outline generation--and on both 7B and 30B model scales for preference data simulation. Reinforcement Learning from Human Feedback (RLHF) has recently been used to great effect to align pretrained large language models (LLMs) to human preferences, optimizing for desirable qualities like harmlessness and helpfulness (Bai et al., 2022a) and achieving ...

Generative artificial intelligence (AI) is a widely popular technology that will have a profound impact on society and individuals. Less than a decade ago, it was thought that creative work would be among the last to be automated - yet today, we see AI encroaching on many creative domains. In this paper, we present the findings of a survey study on people's perceptions of text-to-image generation. We touch on participants' technical understanding of the emerging technology, their fears and concerns, and thoughts about risks and dangers of text-to-image generation to the individual and society. We find that while participants were aware of the risks and dangers associated with the technology, only few participants considered the technology to be a personal risk. The risks for others were more easy to recognize for participants. Artists were particularly seen at risk. Interestingly, participants who had tried the technology rated its future importance lower than those who had not tried it. This result shows that many people are still oblivious of the potential personal risks of generative artificial intelligence and the impending societal changes associated with this technology.

Driverless cars may be worse at detecting children and people with darker skin, tests on artificial intelligence systems suggest. The researchers who carried out the work say that tighter government regulation is needed and that car-makers must be transparent about the development and testing of these vehicles. Jie Zhang at King's College London and her colleagues assessed eight AI-based pedestrian detectors used in driverless car research.

The statement said that Japan's copyright law is more favorable to AI learning than those of other countries.

Newly unsealed court filings reveal how much data Xwitter has handed over to the January 6 investigation. This includes all tweets sent, drafted, liked and retweeted – even if they were subsequently deleted – by Donald Trump's official account. This cache also included DMs sent, received or stored in draft form, as well as linked accounts used on the same device. Even more interesting is the company handed over records of all searches made by the account, too. We already knew Xwitter had fought the order tooth-and-nail, leading to a court battle and a hefty fine. But the list of what was available should also serve as a warning to everyone else that the platform stores a lot more data on its users than you might expect.

Fox News correspondent Gillian Turner has the latest on the president's focus amid calls for an impeachment inquiry on'Special Report.' House Democrats are launching a working group aimed at crafting artificial intelligence policy, the latest attempt by federal lawmakers to wrap their heads around legislating the rapidly-advancing sector. The New Democrat Coalition, a group of nearly 100 House Democrats that touts itself as "pragmatic," unveiled the new initiative this week. Rep. Don Beyer, D-Va., one of the initiative's vice chairs, told Fox News Digital he hopes the working group will "help develop real, practicable ideas that will put guardrails in place for AI. "I continue to be focused on a variety of areas related to AI, including safety and security, transparency, the future of work, preventing civil rights abuses, health care and suicide prevention, and more, and have discussions ongoing about legislation in these areas with members of both parties," Beyer said. "Congress has to get up to speed on this issue, and I think the New Dems' AI working group will be a constructive setting for progress." The Biden administration and Congress are examining how to regulate AI. Working group Chair Rep. Derek Kilmer, D-Wash., suggested it could lay the groundwork for an AI regulatory framework in the House of Representatives. "We are already seeing how breakthroughs in this emerging technology present both great opportunities and challenges with potential disruptions for workers, for democracy, and for national security," Kilmer said. "As AI's applications expand and change, it is incumbent on lawmakers to address its unique opportunities and challenges by creating a regulatory framework that both encourages growth while guarding against potential risks." WHAT IS ARTIFICIAL INTELLIGENCE (AI)? Rep. Seth Moulton, D-Mass., another member of the working group and a Marine veteran, said he was concerned with how AI would "transform warfare" and called on Congress to put up responsible guardrails against the technology's most devastating possibilities. "It's going to be impossible for Congress to really stay ahead of AI, but what we can and should do is to take very seriously AI's most dangerous use cases and develop solutions and safeguards that apply directly to those cases," Moulton told Fox News Digital. "I'm also particularly concerned about how AI will transform warfare.

In the 2023-2024 academic year, the widespread availability of generative artificial intelligence, exemplified by ChatGPT's 1.6 billion monthly visits, is set to impact academic integrity. With 77% of high school students previously reporting engagement in dishonest behaviour, the rise of AI-driven writing assistance, dubbed 'AI-giarism' by Chan (arXiv:2306.03358v2), will make plagiarism more accessible and less detectable. While these concerns are urgent, they also raise broader questions about the revolutionary nature of this technology, including autonomy, data privacy, copyright, and equity. This paper aims to explore generative AI from a social justice perspective, examining the training of these models, the inherent biases, and the potential injustices in detecting AI-generated writing.

The integration of emotional support into various conversational scenarios presents profound societal benefits, such as social interactions, mental health counseling, and customer service. However, there are unsolved challenges that hinder real-world applications in this field, including limited data availability and the absence of well-accepted model training paradigms. This work endeavors to navigate these challenges by harnessing the capabilities of Large Language Models (LLMs). We introduce an innovative methodology that synthesizes human insights with the computational prowess of LLMs to curate an extensive emotional support dialogue dataset. Our approach is initiated with a meticulously designed set of dialogues spanning diverse scenarios as generative seeds. By utilizing the in-context learning potential of ChatGPT, we recursively generate an ExTensible Emotional Support dialogue dataset, named ExTES. Following this, we deploy advanced tuning techniques on the LLaMA model, examining the impact of diverse training strategies, ultimately yielding an LLM meticulously optimized for emotional support interactions. An exhaustive assessment of the resultant model showcases its proficiency in offering emotional support, marking a pivotal step in the realm of emotional support bots and paving the way for subsequent research and implementations.

We introduce EgoSchema, a very long-form video question-answering dataset, and benchmark to evaluate long video understanding capabilities of modern vision and language systems. Derived from Ego4D, EgoSchema consists of over 5000 human curated multiple choice question answer pairs, spanning over 250 hours of real video data, covering a very broad range of natural human activity and behavior. For each question, EgoSchema requires the correct answer to be selected between five given options based on a three-minute-long video clip. While some prior works have proposed video datasets with long clip lengths, we posit that merely the length of the video clip does not truly capture the temporal difficulty of the video task that is being considered. To remedy this, we introduce temporal certificate sets, a general notion for capturing the intrinsic temporal understanding length associated with a broad range of video understanding tasks & datasets. Based on this metric, we find EgoSchema to have intrinsic temporal lengths over 5.7x longer than the second closest dataset and 10x to 100x longer than any other video understanding dataset. Further, our evaluation of several current state-of-the-art video and language models shows them to be severely lacking in long-term video understanding capabilities. Even models with several billions of parameters achieve QA accuracy less than 33% (random is 20%) on the EgoSchema multi-choice question answering task, while humans achieve about 76% accuracy. We posit that \name{}{}, with its long intrinsic temporal structures and diverse complexity, would serve as a valuable evaluation probe for developing effective long-term video understanding systems in the future. Data and Zero-shot model evaluation code are open-sourced for both public and commercial use under the Ego4D license at http://egoschema.github.io

While data-driven predictive models are a strictly technological construct, they may operate within a social context in which benign engineering choices entail implicit, indirect and unexpected real-life consequences. Fairness of such systems -- pertaining both to individuals and groups -- is one relevant consideration in this space; it arises when data capture protected characteristics upon which people may be discriminated. To date, this notion has predominantly been studied for a fixed model, often under different classification thresholds, striving to identify and eradicate undesirable, discriminative and possibly unlawful aspects of its operation. Here, we backtrack on this fixed model assumption to propose and explore a novel definition of cross-model fairness where individuals can be harmed when one predictor is chosen ad hoc from a group of equally-well performing models, i.e., in view of utility-based model multiplicity. Since a person may be classified differently across models that are otherwise considered equivalent, this individual could argue for a predictor granting them the most favourable outcome, employing which may have adverse effects on others. We introduce this scenario with a two-dimensional example and linear classification; then, we present a comprehensive empirical study based on real-life predictive models and data sets that are popular with the algorithmic fairness community; finally, we investigate analytical properties of cross-model fairness and its ramifications in a broader context. Our findings suggest that such unfairness can be readily found in the real life and it may be difficult to mitigate by technical means alone as doing so is likely to degrade predictive performance.