#155 - ChatGPT memory, Altman seeks trillions, Califonia AI regulation, art gen lawsuit
Feb 16, 2024
auto_awesome
ChatGPT gets 'memory' feature for personalized experiences. NVIDIA introduces local custom chatbot for RTX AI PCs. Sam Altman seeks trillions to reshape chip and AI business. Waymo car destroyed by crowd in San Francisco. OpenAI developing AI agents for automated work processes. Cohere for AI launches open source LLM for 101 languages. California bill aims to regulate AI companies.
Stable Cascade offers a stage-based architecture for text-to-image generation, improving performance and versatility.
OpenAI is developing two AI agents to automate work processes, enhancing efficiency across industries.
Go Here For AI's AYA language model supports 101 languages, providing a valuable resource for natural language processing tasks.
EverClip 18B is an open-source text-to-image model with 18 billion parameters, offering improved results and fostering collaboration in the field.
Deep dives
Scalable Cascade for text-to-image generation is introduced.
Stable AI has released Stable Cascade, a text-to-image model that utilizes a stage-based architecture for improved performance. This model outperforms other methods by offering better results and lower inference times. It allows the composition of reasoning modules to solve specific classes of problems, making it more versatile and effective. Stable Cascade represents a notable advancement in text-to-image generation and offers an open-source solution for researchers and developers.
OpenAI unveils AI agents for automated work processes.
OpenAI is developing two AI agents, one for device-based tasks and the other for web-based tasks. These agents allow for automated completion of various work processes, such as data transfers and travel itinerary creation. By leveraging AI technology, these agents streamline and improve efficiency in diverse industries. OpenAI's initiative in developing AI agents marks an important step towards automation and optimization of complex work tasks.
Go Here For AI launches open-source LLM for 101 languages.
Go Here For AI, a nonprofit research lab, has unveiled AYA, an open-source language model that supports 101 languages. AYA offers a dataset with over 513 million annotations, making this project a significant milestone in language modeling. This open-source model is optimized for languages beyond English, providing developers and researchers with a valuable resource for natural language processing tasks.
EverClip 18B scales Clip to 18 billion parameters.
EverClip 18B is an open-source text-to-image model that has been scaled to 18 billion parameters, making it the largest open-source model of its kind. This model builds upon the success of previous Clip models and offers improved results due to its larger size and enhanced architecture. The open-source nature of EverClip 18B allows for collaboration and further advancements in text-to-image generation.
AI Safeguards Can Easily be Broken
The UK's AI Safety Institute has found that AI technology can produce biased outcomes, deceive human users, and lack safeguards against harmful information. The institute's research focuses on large language models and demonstrates how these models can be bypassed or hacked to provide biased or dangerous outputs. This highlights the current challenge in guaranteeing the behavior of AI systems and reinforces the need for further research and development of robust safeguards.
Class Action Lawsuit Against AI Art Companies
A class action copyright lawsuit has been filed by artists against companies like Stability, Mid-Journey, and Runway that provide AI image and video generation services. The companies have filed various motions in response, including introducing new evidence and seeking to dismiss the case entirely. Their counter-argument is that their AI models do not copy existing artwork, but instead reference it to create entirely new products. However, legal proceedings are ongoing, and it remains to be seen how the case will be resolved.
US Judge Rejects Free Speech Defense in AI Art Lawsuit
A US district judge has rejected the argument put forth by AI companies that they are entitled to First Amendment protections for free speech, claiming that the case is in the public interest. This means that the case will not be dismissed on First Amendment grounds. The legal implications and potential outcomes of the lawsuit against AI art companies remain to be seen.
Our 155th episode with a summary and discussion of last week's big AI news!
Correction: Andrey said CLIP came out with DALL-E 2; it came out alongside the first DALL-E.
Check out our sponsor, the SuperDataScience podcast. You can listen to SDS across all major podcasting platforms (e.g., Spotify, Apple Podcasts, Google Podcasts) plus there’s a video version on YouTube.