If the unofficial debut of Google's Bard chatbot is any indication, misinformation is about to get a lot worse. The company posted an ad to Twitter this week showing off the natural-language AI model displaying false information about the James Webb Space Telescope (JWST). In the advertisement (via Reuters), a short GIF shows an example of a Q&A with Bard. "What new discoveries from the James Webb Space Telescope can I tell my 9-year old about?" the query reads. The machine quickly spits out three ideas, including the last one that says, "JWST took the very first pictures of a planet outside of our own solar system.

Fox News Flash top headlines are here. Check out what's clicking on Foxnews.com. Firefighters worked for a second day Wednesday to fully extinguish a blaze at a U.S. company's drone plant in Latvia. Local police said nothing had been found so far to indicate sabotage. Latvia's State Fire and Rescue Service was alerted Tuesday afternoon that a fire had broken out at Edge Autonomy's drone production plant in Marupe, a town that borders the capital, Riga. The Baltic News Service said that although the blaze was largely contained by 7 p.m. on Tuesday, firefighters continued work to fully extinguish the fire early Wednesday.\

Google and Microsoft are going head to head over the future of search by embracing the technology behind artificial intelligence chatbots. Google announced on Monday that it is testing Bard, a rival to the Microsoft-backed ChatGPT, which has swiftly become a sensation, and will roll it out to the public in the coming weeks. And on Tuesday, Microsoft announced it is increasing its focus on artificial intelligence, boosting funding for new tools and integrating the technology underpinning ChatGPT into products including its Bing search engine and Edge browser, with the goal of making search more conversational. ChatGPT, developed by San Francisco company OpenAI, has reached 100 million users since its public launch in November, becoming by some estimates the fasting growing consumer app of all time. Here are some questions about Google and Microsoft's AI plans and their likely impact.

AI-generated text, from tools like ChatGPT, is starting to impact daily life. Teachers are testing it out as part of classroom lessons. Marketers are champing at the bit to replace their interns. Memers are going buck wild. It would be a lie to say I'm not a little anxious about the robots coming for my writing gig.

The "news broadcasters" appear stunningly real, but they are AI-generated deepfakes in first-of-their-kind propaganda videos that a research report published Tuesday attributed to Chinese state-aligned actors. The fake anchors -- for a fictious news outlet called Wolf News -- were created by artificial intelligence software and appeared in footage on social media that seemed to promote the interests of the Chinese Communist Party, US-based research firm Graphika said in its report. "This is the first time we've seen a state-aligned operation use AI-generated video footage of a fictitious person to create deceptive political content," Jack Stubbs, vice president of intelligence at Graphika, told AFP. In one video analyzed by Graphika, a fictious male anchor who calls himself Alex critiques US inaction over gun violence plaguing the country. In the second, a female anchor stresses the importance of "great power cooperation" between China and the United States.

Recent advancements in large scale text-to-image models have opened new possibilities for guiding the creation of images through human-devised natural language. However, while prior literature has primarily focused on the generation of individual images, it is essential to consider the capability of these models to ensure coherency within a sequence of images to fulfill the demands of real-world applications such as storytelling. To address this, here we present a novel neural pipeline for generating a coherent storybook from the plain text of a story. Specifically, we leverage a combination of a pre-trained Large Language Model and a text-guided Latent Diffusion Model to generate coherent images. While previous story synthesis frameworks typically require a large-scale text-to-image model trained on expensive image-caption pairs to maintain the coherency, we employ simple textual inversion techniques along with detector-based semantic image editing which allows zero-shot generation of the coherent storybook. Experimental results show that our proposed method outperforms state-of-the-art image editing baselines.

Event prediction is the ability of anticipating future events, i.e., future real-world occurrences, and aims to support the user in deciding on actions that change future events towards a desired state. An event prediction method learns the relation between features of past events and future events. It is applied to newly observed events to predict corresponding future events that are evaluated with respect to the user's desired future state. If the predicted future events do not comply with this state, actions are taken towards achieving desirable future states. Evidently, event prediction is valuable in many application domains such as business and natural disasters. The diversity of application domains results in a diverse range of methods that are scattered across various research areas which, in turn, use different terminology for event prediction methods. Consequently, sharing methods and knowledge for developing future event prediction methods is restricted. To facilitate knowledge sharing on account of a comprehensive classification, integration, and assessment of event prediction methods, we combine taxonomies and take a systems perspective to integrate event prediction methods into a single system, elicit requirements and assess existing work with respect to the requirements. Based on the assessment, we identify open challenges and discuss future research directions.

Many open-domain dialogue systems rely on multiple response generators, any of which can contribute a response to the dialogue in a particular context. Thus the ability to compare potential responses and then select the best plays an important role in ensuring a dialogue system is coherent and engaging. Dialogue coherence goes beyond simply remaining on topic -- some trivia may be on topic and engaging when mentioned out of the blue, but may not be coherent and grounded in the context of the conversation. We carry out experiments on response selection in the Athena system, an Alexa Prize SocialBot that has dedicated content and multiple topic-specific response generators for a large number of topics. First, we collect a corpus of Athena conversations with live human traffic, where potential responses from all enabled response generators are logged and subsequently annotated for response quality. We compare several off-the-shelf response ranking methods for open-domain dialogue to Athena-Heuristic, a heuristic response ranker that was field-tested in Athena during the third Alexa Prize competition. We also compare these to a transformer-based response ranker we call Athena-RR, that we train on our Athena conversations. Athena-RR uses both the conversational context and the dialogue state to rank the potential responses. We find that Athena-RR with a Recall@1 of 70.79\% outperforms Athena-Heuristic and all of the off-the-shelf rankers by a large margin. We then conduct a live A/B study comparing Athena-Heuristic to Athena-RR in a 6,358 conversations with Alexa users. We show that Athena-RR leads to significantly longer conversations that receive significantly higher user ratings than the heuristic rule-based ranker.

Recent Transformer-based summarization models have provided a promising approach to abstractive summarization. They go beyond sentence selection and extractive strategies to deal with more complicated tasks such as novel word generation and sentence paraphrasing. Nonetheless, these models have two shortcomings: (1) they often perform poorly in content selection, and (2) their training strategy is not quite efficient, which restricts model performance. In this paper, we explore two orthogonal ways to compensate for these pitfalls. First, we augment the Transformer network with a sentence cross-attention module in the decoder, encouraging more abstraction of salient content. Second, we include a curriculum learning approach to reweight the training samples, bringing about an efficient learning procedure. Our second approach to enhance the training strategy of Transformers networks makes stronger gains as compared to the first approach. We apply our model on extreme summarization dataset of Reddit TIFU posts. We further look into three cross-domain summarization datasets (Webis-TLDR-17, CNN/DM, and XSum), measuring the efficacy of curriculum learning when applied in summarization. Moreover, a human evaluation is conducted to show the efficacy of the proposed method in terms of qualitative criteria, namely, fluency, informativeness, and overall quality.

Recommender system practitioners are facing increasing pressure to explain recommendations. We explore how to explain recommendations using counterfactual logic, i.e. "Had you not interacted with the following items, we would not recommend it." Compared to the traditional explanation logic, counterfactual explanations are easier to understand, more technically verifiable, and more informative in terms of giving users control over recommendations. The major challenge of generating such explanations is the computational cost because it requires repeatedly retraining the models to obtain the effect on a recommendation caused by the absence of user history. We propose a learning-based framework to generate counterfactual explanations. The key idea is to train a surrogate model to learn the effect of removing a subset of user history on the recommendation. To this end, we first artificially simulate the counterfactual outcomes on the recommendation after deleting subsets of history. Then we train a surrogate model to learn the mapping between a history deletion and the corresponding change of the recommendation caused by the deletion. Finally, to generate an explanation, we find the history subset predicted by the surrogate model that is most likely to remove the recommendation. Through offline experiments and online user studies, we show our method, compared to baselines, can generate explanations that are more counterfactually valid and more satisfactory considered by users.