We introduce Composer's Assistant, a system for interactive human-computer composition in the REAPER digital audio workstation. We consider the task of multi-track MIDI infilling when arbitrary track-measures have been deleted from a contiguous slice of measures from a MIDI file, and we train a T5-like model to accomplish this task. Composer's Assistant consists of this model together with scripts that enable interaction with the model in REAPER. We conduct objective and subjective tests of our model. We release our complete system, consisting of source code, pretrained models, and REAPER scripts. Our models were trained only on permissively-licensed MIDI files.

Developers often have questions about semantic aspects of code they are working on, e.g., "Is there a class whose parent classes declare a conflicting attribute?". Answering them requires understanding code semantics such as attributes and inheritance relation of classes. An answer to such a question should identify code spans constituting the answer (e.g., the declaration of the subclass) as well as supporting facts (e.g., the definitions of the conflicting attributes). The existing work on question-answering over code has considered yes/no questions or method-level context. We contribute a labeled dataset, called CodeQueries, of semantic queries over Python code. Compared to the existing datasets, in CodeQueries, the queries are about code semantics, the context is file level and the answers are code spans. We curate the dataset based on queries supported by a widely-used static analysis tool, CodeQL, and include both positive and negative examples, and queries requiring single-hop and multi-hop reasoning. To assess the value of our dataset, we evaluate baseline neural approaches. We study a large language model (GPT3.5-Turbo) in zero-shot and few-shot settings on a subset of CodeQueries. We also evaluate a BERT style model (CuBERT) with fine-tuning. We find that these models achieve limited success on CodeQueries. CodeQueries is thus a challenging dataset to test the ability of neural models, to understand code semantics, in the extractive question-answering setting.

The machines may one day rule over humanity -- but in the near term AI-based chatbots like ChatGPT could help struggling workers hold their own in writing tasks. Those are the findings of a new study, at least, which tracked the writing performance of 453 'experienced, college-educated professionals' divided equally into two groups: one taught to use ChatGPT and another forced to complete their writing assignments on their own. The researchers found that the chatbot made their test subjects 40-percent more productive and improved the quality of their work by 18 percent. In fact, a significant number of the study participants, who had been encouraged to use ChatGPT, stuck with the software for an added boost in the real world. After two weeks, 34 percent reported using ChatGPT in a professional setting.

This spring, Congress hosted OpenAI's chief executive Sam Altman for a hearing, in which he admitted the technology could be a sousce of errors. He called for regulations to be crafted for the emerging industry and recommended that a new agency be formed to tackle it. He said he expected the technology to have a significant impact as its uses become clear, including on jobs.

The Associated Press (AP) and ChatGPT parent company OpenAI have reached a news-sharing agreement, but not for the reasons you may think. It doesn't involve AI chatbots quickly churning out content, but rather a way for OpenAI to train its algorithmic models, as reported by Axios. The two-year deal gives OpenAI access to select news content and technology from the AP archives, dating back to 1985. All of this sweet, sweet data will be used to improve the efficacy of future iterations of ChatGPT and related tools. This is one of the first high-profile partnerships between a major news organization and an artificial intelligence company.

Now, a rising group of authors, musicians, news organizations and social media companies has been pushing back, arguing that the use of their content to train AI is a massive shift in the way the internet works, especially since some of the AI tools being trained on human-made content are already being used to replace human workers. A wave of lawsuits has washed over the industry in the past two weeks alleging improper data use, including class-action suits against OpenAI and Google, and lawsuits against OpenAI from the comedian Sarah Silverman and two prominent fiction authors.

Shakked Noy and Whitney Zhang recruited 453 marketers, data analysts, and college-educated professionals and got each of them to complete two kinds of tasks they'd normally undertake as part of their jobs, such as writing press releases, short reports, or analysis plans. Half were given the option of using ChatGPT to help them complete the second of the two tasks. Each piece of work was evaluated by three people working in the same professions, hired through the research platform Prolific. The writers who chose to use ChatGPT took 40% less time to complete their tasks, and produced work that the assessors scored 18% higher in quality than that of the participants who didn't use it. The writers who were already skilled at writing were able to reduce the amount of time they spent on their work, while those who were assessed as being weaker writers produced higher-quality work once they gained access to the chatbot.

American regulators now appear to be clamping down on generative AI in earnest. The Washington Post has learned the Federal Trade Commission (FTC) has launched an investigation into OpenAI, the creator of ChatGPT and DALL-E. Officials have requested documents showing how the company tackles risks stemming from its large language AI models. The FTC is concerned the company may be violating consumer protection laws through "unfair or deceptive" practices that could hurt the public's privacy, security or reputation. The Commission is particularly interested in information linked to a bug that leaked ChatGPT users' sensitive data, including payments and chat histories.

Stealing someone else's joke is one of the highest crimes in comedy. With new AI tools like ChatGPT, some comedians are now worried about getting ripped off. After comedian Sarah Silverman joined a lawsuit against OpenAI and Meta for allegedly using her content to train their bots without permission, one comic told Fox News ChatGPT does not pose a threat to him. "In terms of how ChatGPT affects comedy, yes, I think we're going to enter the golden age of in-print comedians, meaning people who can type things on the internet," said Jimmy Failla, comedian and host of "Fox Across America" on Fox News Radio and Fox Nation. "But where true performers and people with actual charisma and comedic wherewithal will always flourish is no one's going to show up to a comedy club and buy a two-drink minimum to stare at a laptop, typing out words, or even saying those words through some Bluetooth audio," he continued.

The US's top competition watchdog has opened an investigation into OpenAI, the maker of ChatGPT, on claims it has run afoul of consumer protection laws by putting personal reputations and data at risk. The move marks the strongest regulatory threat to the Microsoft-backed startup that kicked off the frenzy in generative artificial intelligence, enthralling consumers and businesses while raising concerns about its potential risks. The Federal Trade Commission (FTC) this week sent a 20-page demand for records about how OpenAI addresses risks related to its AI models. The agency is investigating whether the company engaged in unfair or deceptive practices that resulted in "reputational harm" to consumers. One of the questions has to do with steps OpenAI has taken to address the potential for its products to "generate statements about real individuals that are false, misleading, or disparaging".