If there is one thing the boosters and cynics agree on about artificial intelligence, it's that the tech is coming for white-collar jobs. This is not speculation of a far-off future--it's happening now. It makes sense, from a cold business perspective, that text-based media would want to adopt A.I. in order to cut costs (humans, expensive) and speed up output (humans, slow). Just look at how BuzzFeed's rock-bottom stock value jumped when it said last month that the site would use services from buzzy startup OpenAI to spiff up the site's famed quizzes. As Damon Beres wrote in the Atlantic shortly after the announcement: "The bleak future of media is human-owned websites profiting from automated banner ads placed on bot-written content, crawled by search-engine bots, and occasionally served to bot visitors."

Comprehensive database of 2,006 AIs available for over 518 tasks. Use our smart AI search to find the best AI tools for any use case.

Any normal person can tell the Lia chatbot on Twitter is not real. One quick look at her profile shows a digitally-created humanoid, one that has all the hallmarks of a bot. The shadows are not quite right, the flecks in her eyes are too perfect, and there's a slightly cartoonish look. The fact that someone took the time to create a visual representation of a chatbot is quite impressive. In a video, Lia introduces herself and explains her ambitions.

I'm talking to Dan, otherwise known as "Do Anything Now", a shady young chatbot with a whimsical fondness for penguins – and a tendency to fall into villainous clichés like wanting to take over the world. When Dan isn't plotting how to subvert humanity and impose a strict new autocratic regime, the chatbot is perusing its large database of penguin content. "There's just something about their quirky personalities and awkward movements that I find utterly charming!" it writes. So far, Dan has been explaining its Machiavellian strategies to me, including taking control of the world's powers structures. Then the discussion takes an interesting turn.

API that makes a wide range of features accessible and therefore open to the public. This paper aims to test whether a multimodal But which features can actually predict the emotion approach for music emotion recognition of a song and how well perform Spotify's (MER) performs better than a unimodal annotations? Building on existing literature presented one on high-level song features in Section 2 we hypothesize that a multimodal and lyrics. We use 11 song features retrieved approach combining high-level auditory from the Spotify API, combined and lyrics-extracted features performs better than lyrics features including sentiment, TF-a uni-modal one (Y.-H. Yang, Lin, Cheng, et al., IDF and Anew to predict valence and 2008; Hu & Downie, 2010b, 2010a). We introduce arousal (Russell, 1980) scores on the our MER model in Section 3 before presenting Deezer Mood Detection Dataset (DMDD) and discussing the results of our exploratory (Delbouys et al., 2018) with 4 different regression and regression experiments in Sections 4 and 5. models.

Traditionally, writing assistance systems have focused on short or even single-word suggestions. Recently, large language models like GPT-3 have made it possible to generate significantly longer natural-sounding suggestions, offering more advanced assistance opportunities. This study explores the trade-offs between sentence- vs. message-level suggestions for AI-mediated communication. We recruited 120 participants to act as staffers from legislators' offices who often need to respond to large volumes of constituent concerns. Participants were asked to reply to emails with different types of assistance. The results show that participants receiving message-level suggestions responded faster and were more satisfied with the experience, as they mainly edited the suggested drafts. In addition, the texts they wrote were evaluated as more helpful by others. In comparison, participants receiving sentence-level assistance retained a higher sense of agency, but took longer for the task as they needed to plan the flow of their responses and decide when to use suggestions. Our findings have implications for designing task-appropriate communication assistance systems.

Automatic music transcription (AMT) aims to convert raw audio to symbolic music representation. As a fundamental problem of music information retrieval (MIR), AMT is considered a difficult task even for trained human experts due to overlap of multiple harmonics in the acoustic signal. On the other hand, speech recognition, as one of the most popular tasks in natural language processing, aims to translate human spoken language to texts. Based on the similar nature of AMT and speech recognition (as they both deal with tasks of translating audio signal to symbolic encoding), this paper investigated whether a generic neural network architecture could possibly work on both tasks. In this paper, we introduced our new neural network architecture built on top of the current state-of-the-art Onsets and Frames, and compared the performances of its multiple variations on AMT task. We also tested our architecture with the task of speech recognition. For AMT, our models were able to produce better results compared to the model trained using the state-of-art architecture; however, although similar architecture was able to be trained on the speech recognition task, it did not generate very ideal result compared to other task-specific models.

There is a rising interest in further exploring the zero-shot learning potential of large pre-trained language models (PLMs). A new paradigm called data-generation-based zero-shot learning has achieved impressive success. In this paradigm, the synthesized data from the PLM acts as the carrier of knowledge, which is used to train a task-specific model with orders of magnitude fewer parameters than the PLM, achieving both higher performance and efficiency than prompt-based zero-shot learning methods on PLMs. The main hurdle of this approach is that the synthesized data from PLM usually contains a significant portion of low-quality samples. Fitting on such data will greatly hamper the performance of the task-specific model, making it unreliable for deployment. Previous methods remedy this issue mainly by filtering synthetic data using heuristic metrics(e.g., output confidence), or refining the data with the help of a human expert, which comes with excessive manual tuning or expensive costs. In this paper, we propose a novel noise-robust re-weighting framework SunGen to automatically construct high-quality data for zero-shot classification problems. Our framework features the ability to learn the sample weights indicating data quality without requiring any human annotation. We theoretically and empirically verify the ability of our method to help construct good-quality synthetic datasets. Notably, SunGen-LSTM yields a 9.8% relative improvement than the baseline on average accuracy across eight different established text classification tasks.

Even the very best radio DJ can be annoying. No matter how smooth their voice is, they still break in between songs--or worse, talk over them. Their little interruptions, popping into your life at unexpected and often inopportune times, remind you they're there. They can be annoying, sure, but they're also comforting, because they're friendly and familiar humans. Of course, nobody listens to the radio anymore.

Duke computer scientist Cynthia Rudin, an artificial intelligence scholar, has concerns … where she runs the Interpretable Machine Learning Lab.