The Southern DIstrict of New York has dismissed a data scraping copyright case brought against OpenAI by Raw Story Media and AlterNet Media. The dismissal was based on plaintiffs’ lack of standing as they could not specify concrete harm to them resulting from OpenAI’s actions. OpenAI and Microsoft still face a copyright suit brought The New York Times on slightly different grounds.
AI may be intelligent but it’s not always smart. Researchers at Cornell have shown that AI agents that understand visual interfaces (like a computer screen) can easily be fooled by adversarial pop-ups into clicking buttons or filling forms. Pop-ups that would look very suspicious to humans just don’t trigger the same response from naive AI.
AWS will allow organizations and their partners to generate data and insights in AWS clean rooms using their own custom machine learning models. Advertisers and publishers will be able to optimize campaigns while keeping their data and machine learning models private. AWS Clean Room’s own machine learning offers lookalike modeling that it says improves 36% on industry baselines.
ChatGPT is hoping to grab another slice of your attention with Pulse, a now-in-limited-preview feature that will proactively deliver daily updates on information it thinks you’ll find useful. The new service is positioned as a personal assistant that will select contents “based on your chats, feedback, and connected apps like your calendar.” Forget about privacy; the real question is whether this becomes a news source that gives ChatGPT the ability to shape your opinions.