- Superpower Daily
- OpenAI responds to New York Times lawsuit
OpenAI responds to New York Times lawsuit
Volkswagen is bringing ChatGPT into its cars and SUVs
OpenAI, creators of ChatGPT, is facing a lawsuit filed by The New York Times in December, accusing them and Microsoft of copyright infringement for using the Times' journalistic content in ChatGPT's training data. OpenAI, in a statement, expressed disagreement and surprise over the lawsuit, stating they believed their training practices constituted fair use and offered an opt-out option. They also noted efforts to resolve content regurgitation issues in ChatGPT.
Collaboration with News Organizations: The company highlights its efforts to support the news industry through partnerships and AI tools to assist journalists.
Training and Fair Use: It claims that using internet content for AI training is fair use, but offers an opt-out for publishers.
Addressing "Regurgitation" Issues: The company acknowledges occasional content regurgitation by its AI models as a rare bug and is working to minimize it.
Dispute with The New York Times: The company is surprised by The New York Times' lawsuit, suggesting discussions were initially positive and hinting that the newspaper might have manipulated prompts to produce specific results.
Ongoing Commitment: Despite the lawsuit, the company remains committed to enhancing journalism with AI and is open to partnerships with news organizations, including The New York Times.
Additionally, the lawsuit by The New York Times is part of a broader trend of legal challenges against generative AI companies. Notably, in September, prominent authors like Jonathan Franzen and George R.R. Martin sued OpenAI for using their works in ChatGPT training. Two authors filed a similar lawsuit in July. On the image generation front, Getty Images sued Stability AI for allegedly using its images for training data. Stability AI, Midjourney, and DeviantArt faced another lawsuit over their AI image generators. Lastly, Microsoft, GitHub, and OpenAI are involved in a lawsuit over allegations of scraping licensed code for training their AI code generators.
This new version of Ballie, about the size of a bowling ball, features several AI upgrades. It has a spatial lidar sensor for navigation, a 1080p projector with two lenses for projecting movies and video calls, and the ability to act as a second PC monitor. Ballie can adjust projections based on wall distance and lighting, and it can detect people's posture and facial angle to optimize the viewing angle.
Ballie can be controlled via voice commands or text messages, and it responds through a chatbot to confirm requests. It's capable of integrating with smart home devices, turning on smart lights, and controlling non-smart devices like air conditioners and older TVs via an infrared transmitter. The robot can also map a home's floor plan to locate smart devices.
Additional promised features include automatic reminders for tasks like watering plants, access to remote medical services, and personalized responses based on user recognition. These features, along with Ballie's availability and pricing, are still under development.
The question is would these features be enough to make Ballie a success in the home robot market, considering past challenges faced by other companies in this sector? The success of Samsung's Ballie in the market remains to be seen.
AutoRT, Google's new autonomous wheeled robot, integrates a visual language model (VLM) and a large language model (LLM) to understand its environment, adapt to new settings, and select appropriate tasks. It's guided by a "Robot Constitution" based on Isaac Asimov’s "Three Laws of Robotics," which includes safety-focused prompts to avoid tasks involving humans, animals, sharp objects, and electrical appliances.
To enhance safety, the robots automatically stop if joint force exceeds a set threshold, and they are equipped with a physical kill switch. Google tested 53 AutoRT robots in four office buildings over seven months, conducting over 77,000 trials. These robots, which have a camera, robot arm, and mobile base, use the VLM to recognize their environment and objects, while the LLM suggests and decides on tasks, like placing a napkin on a countertop or opening a bag of chips.
DeepMind also introduced SARA-RT, a neural network architecture for improving the Robotic Transformer RT-2's accuracy and speed, and RT-Trajectory, which adds 2D outlines to assist robots in physical tasks like wiping tables. The article concludes that while fully autonomous robots for household tasks like serving drinks and fluffing pillows are still far off, technologies like AutoRT are laying the groundwork for their development.
Volkswagen announced at CES 2024 plans to integrate an AI-powered chatbot, based on OpenAI's large language model and Cerence’s Chat Pro software, into its vehicles equipped with the IDA voice assistant. This feature will initially be available in Europe from the second quarter, in models including Volkswagen's EVs (ID.7, ID.4, ID.5, ID.3) and other vehicles like Tiguan, Passat, and Golf. The U.S. release is still under consideration.
The chatbot aims to enhance driver experience by reading researched content aloud, answering vehicle-specific questions, and engaging in more intuitive, conversational interactions. It will work with the existing "Hello IDA" wake word or a steering wheel button, forwarding complex queries to the AI chatbot.
Volkswagen is setting content limits for the chatbot, avoiding sensitive topics and brand-specific comparisons. This initiative, separate from VW Group's software arm Cariad's projects, marks a significant step in integrating conversational AI into automotive technology, following Mercedes-Benz, which integrated a similar AI bot into its system earlier.
An “AI Breakthrough” on Systematic Generalization in Language?
Getty and Nvidia bring generative AI to stock photos
GPT Checkup - This website measures how GPT-4 with Vision performs across a range of experiments.
How to Build a Thinking AI
The Great Sobering Of 2024: A Look Ahead For AI Startups
Visualizing novels using Midjourney v6 and GPT-4
Figure released a new video of Figure-01, their first humanoid robot that has learned to make coffee after watching humans make coffee
Amazon turns to AI to help customers find clothes that fit when shopping online
Can ChatGPT get into Harvard? We tested its admissions essay.
ChatGPT Tips and Tricks - Adjusting Conversation Width, Disabling Auto-Scroll, and more...
How did you like today’s newsletter?
Hope you enjoyed today's newsletter
Did you know you can add Superpower Daily to your RSS feed https://rss.beehiiv.com/feeds/GcFiF2T4I5.xml
⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.