TLDR AI 2023-10-26

When incredibly valuable assets come up for sale, it's typically the wealthiest that end up taking home an amazing investment. But not always...

One platform is taking on the billionaires at their own game, fractionalizing prized blue-chip artworks for its investors. In just the last few years, its investors have realized annualized net returns of 17.8%, 21.5%, 35% and more.

It's called Masterworks. When Masterworks sells a painting – like the 16 it's already sold – investors reap their portion of the profits.

Offerings can sell out in minutes, but TLDR AI readers can skip the waitlist to join.

Past performance is not indicative of future returns, Investing involves risk. See disclosures masterworks.com/cd

AI search startup Perplexity has secured a new round of funding led by IVP that reportedly values the one-year-old company at $500m. Perplexity uses artificial intelligence to challenge Google’s dominance in online search with an AI-powered search engine and a new iOS app. It reported a $3m annual recurring revenue as of this month.

Anthropic, in partnership with the Collective Intelligence Project, developed a new AI system using a constitution sourced from the public to explore democratic influences on AI. It gathered input from ~1,000 Americans via the Polis platform where participants could vote on or suggest principles for the AI chatbot. The resulting "Public" model, which showed a greater emphasis on accessibility and impartiality, performed comparably to Anthropic's standard model in understanding tasks and bias evaluations.

Google now offers two indemnities for generative AI products: one covering the use of training data, protecting customers from third-party intellectual property infringement claims, and a second indemnifying clients for generated outputs from a range of Google Cloud services given responsible AI practices. These measures aim to reassure customers using Google's generative AI services against potential copyright infringement claims.

Retrieving information from structured text documents is a well-studied problem that gets much harder when the documents are PDFs. Adobe Research has introduced a tool that allows you to condition either structure or text for retrieval. It also released a benchmark dataset for further study.

This project introduced RLMRec, a framework that combines LLMs with recommender systems, capturing the deeper meanings of user behaviors and preferences and enhancing the quality of recommendations.

VidChapters-7M is a big collection of videos with 7 million user-made chapter marks. This new dataset helps in the creation and testing of methods for splitting videos into chapters and naming them, making it simpler for viewers to find what they're looking for.

AI-to-AI testing for LLM-based applications.

ReactAgent is an experimental autonomous agent that uses GPT-4 to generate and compose React components from user stories.

An all-in-one managed backend for AI apps.

The dust is settling now that the AI hype has started to subside. Digging in, where are the real opportunities in AI?

Peak Data explores how an internet increasingly filled with internet data might lead to model collapse, its effect on apps like ChatGPT, and possible solutions that foundation model vendors might adopt to overcome model collapse.

A distillation of the current leading open source text to image model. The researchers claim a minimal reduction in image quality with a dramatic speed-up at inference time.

Humane’s AI assistant pin attaches magnetically to your clothing and is powered by GPT-4.

Create visual prompt chains with various models.

Create perpetual chatbots with self-editing memory.