The Daily Index — Wednesday, March 13, 2024 · TLDR Index

RESEARCH & INNOVATION

TLDR AI

TLDR AI

STEALING PART OF A PRODUCTION LANGUAGE MODEL

You can steal portions of closed language models such as the embeddings layer just by using their public APIs. This can be done for a modest budget of less than $2,000.

arxiv.org · 24 min

TLDR AI

MAMBA BASED DNA SEQUENCE MODEL

A sequence prediction model for DNA built on the Transformer competitor Mamba. It is extremely efficient and powerful for a small model.

caduceus-dna.github.io · 7 min

TLDR AI

3D GENERATION WITH VIDEO DIFFUSION MODELS

This project introduces a novel approach that enhances 3D object generation, producing high-quality, detailed objects from a single image.

heheyas.github.io · 2 min

TLDR AI

RESEARCHERS CREATE AI WORMS THAT CAN SPREAD FROM ONE SYSTEM TO ANOTHER

Researchers have created a generative AI worm called Morris II that can attack AI systems like ChatGPT, spreading autonomously while potentially stealing data. The worm uses “adversarial self-replicating prompts” to perpetuate and compromise AI email assistants, highlighting new cyberattack risks within the AI ecosystem. Security experts urge AI developers to take potential AI-driven threats seriously as AI applications become more autonomous and interconnected.

arstechnica.com · 6 min

TLDR AI

THE FUTURE OF POETRY

38 AI experts and 39 English experts were asked to rate and guess whether poems were written by an AI or a human. The human came in 1st place, while Bard, ChatGPT-4, and Claude came in 2nd, 3rd, and 4th places respectively, both in writing quality and their ability to fool respondents into believing their poems were authored by a human. English experts were far better at discerning which poems were written by AI, which points to a need for them to play a greater role in helping shape future versions of AI technology.

medium.com · 14 min

TLDR AI

BARACK OBAMA ON AI, FREE SPEECH, AND THE FUTURE OF THE INTERNET

In a discussion about the need for AI regulation and transparent development practices with tech companies, Former President Barack Obama highlighted AI's potential risks and rewards and urged tech experts to engage in government roles to help shape thoughtful AI policies. The conversation also tackled First Amendment challenges and the necessity of a multi-faceted, adaptive regulatory approach for AI.

theverge.com · 27 min

TLDR AI

PERPLEXITY BRINGS YELP DATA TO ITS CHATBOT

Perplexity has integrated Yelp's data to enhance chatbot responses with maps, reviews, and detailed information on restaurants to provide users with direct and rich local search results.

theverge.com · 1 min

TLDR AI

GEMINI NOW LETS YOU TUNE AND MODIFY RESPONSES WITH A PROMPT

Google is rolling out a new feature for the Gemini web app that allows users to fine-tune AI-generated responses using options like Regenerate, Shorter, Longer, and Remove, with the goal of providing greater control over their creative content.

9to5google.com · 2 min

TLDR AI

MICROSOFT’S NEURAL VOICE TOOL FOR PEOPLE WITH SPEECH DISABILITIES ARRIVES LATER THIS YEAR

Microsoft announced updates focused on accessibility, like new Azure AI features for developers with disabilities, a preview of the “Speak For Me” feature for those with speech disabilities, Copilot's new skills, and a Mental Health toolkit, at its 14th Ability Summit.

engadget.com · 3 min