large language models Archives

White House Takes Action to Combat Deepfakes and AI Scams

December 13, 2023December 13, 2023 Business

The White House is taking action to combat deepfakes and AI scams by setting an example for content authentication and security. Through President Biden’s Executive Order on AI, federal agencies will partner with the Department of Commerce to develop tools for authenticating AI-generated content. Watermarking is one aspect of this initiative, but the details are yet to be revealed. The EU has also released its own AI regulations in the form of the Artificial Intelligence Act. The White House aims to protect privacy and advance AI ethics through the development of cryptographic tools and the evaluation of privacy techniques. These efforts reflect the commitment of the White House to address the risks and consequences associated with AI technology.

Safeguarding against the Security Risks of Large Language Models

November 20, 2023November 27, 2023 Business

Large language models (LLMs) offer significant advantages but also present security risks. Lasso Security is a newly launched company that aims to address these concerns by intercepting and monitoring LLM interactions. By capturing data sent to and received from LLMs and using advanced threat detection measures, Lasso Security provides end-to-end protection against threats such as prompt injection, jailbreaking, data poisoning, and model denial of service. Their platform also ensures that organizations can safely leverage LLMs by enforcing security policies and gaining complete control over LLM-related interactions.

AI’s Roadblock: Transformers Aren’t Good at Generalizing

November 7, 2023November 22, 2023 Business

Researchers from Google have revealed that transformers, a technology powering AI tools like ChatGPT, struggle with generalizing and performing tasks beyond their training data. This poses a challenge for achieving artificial general intelligence (AGI), as AI currently lacks the capacity to transfer skills across domains like humans. The study has prompted a reevaluation of the capabilities of large language models (LLMs) and the need to temper expectations of imminent AGI. Despite the limitations, some experts remain optimistic about the future role of AI models.

Reka Unveils Yasa-1: A Multimodal AI Assistant with Advanced Capabilities

October 5, 2023November 11, 2023 Business

AI startup Reka has introduced Yasa-1, a multimodal AI assistant that can understand images, short videos, and audio. Yasa-1 is highly customizable, supports multiple languages, and provides context-based answers from the internet. It also boasts features like processing long documents and executing code. Reka plans to iron out limitations and expand Yasa-1’s accessibility in the coming weeks. While Reka is relatively new in the AI industry, its team combined with substantial funding positions it as a competitor against established players like OpenAI and Anthropic.

Protection for Writers in AI-powered Entertainment Industry

September 30, 2023November 11, 2023 Technology

The Writers Guild of America (WGA) contract provides protections against the displacement of writers by AI in the entertainment industry. The contract states that AI cannot be used to write or rewrite literary material, and AI-generated content will not be considered as source material. Writers retain sole credit for their creative works.