TLDR Index

Search the archive

Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.

Results

16,678 matches

TAXY AI

Taxy uses GPT-4 to control your browser and perform repetitive actions on your behalf.

2023.03.29
github.comGitHub repositoryTLDR AI

CHARACTER AI'S NEW MODEL AND NEW FUNDING

The start up that encourages you to talk with an AI chatbot has a boost in resources with their latest round of funding. True dialogue agents are a powerful and dangerous technology. They're hoping to build them out safely for everyone.

2023.03.27
blog.character.ai4 min readTLDR AI

GENERATIVE RETRIEVAL RECOMMENDER SYSTEM

Most deep recommendation systems work by embedding products and then doing some sort of nearest neighbor retrieval. There are pros and cons here. This new work uses a sequence to sequence model and a discrete product embedding to recommend the next product in a sequence of customer interactions. It improves as much as 17% over some strong baselines.

2023.03.27
shashankrajput.github.io25 min readTLDR AI

GENERATED TEXT IS VERY HARD TO DETECT

With the rise of convincing text generation systems, we as a community are scrambling to build detectors. It turns out this may not even be possible, given that the LLMs are trained to mimic human language. There is some hope though by using retrieval for semantically similar generations hosted by the model providers.

2023.03.27
arxiv.org34 min readTLDR AI

NANOPALM

Inspired by NanoGPT and NanoT5, we now have NanoPaLM. PaLM is the pathways model from Google and it has some novel architectural changes that make it interesting to study. A personal favorite is the use of multi-query attention which reduces computational costs and allows for much longer context lengths.

2023.03.27
github.comGitHub repositoryTLDR AI

MODELS

The article discusses a new technique called Saliency-aware Noise Blending (SNB) that blends the predicted noises of two diffusion models to achieve more controllable image generation. The proposed method is training-free, and experiments show that it is highly effective in various applications.

2023.03.27
magicfusion.github.io6 min readTLDR AI

AND SUCCEED

Brian O'Malley, managing partner at Forerunner, emphasizes the potential of AI to reshape industries, but also highlights the current stage's uncertainties and comparisons to the buzz around Web3. To make AI an essential part of daily life, he believes it needs to understand nuances in intent and emotion, similar to how GPS and camera advancements transformed mobile technology. AI should be leveraged to serve customers, with the best AI companies being those where customers don't even know AI is involved. Focusing on reliable and repeatable AI experiences will help build enduring businesses, ultimately bringing humanity back into technology-driven experiences.

2023.03.27
archive.ph2 min readTLDR AI

SHELL GPT

Shell GPT is a command-line productivity tool powered by GPT-3.5

2023.03.27
github.comGitHub repositoryTLDR AI

CHATGPT PLUGINS

Plugins are tools designed specifically for language models with safety as a core principle, and help ChatGPT access up-to-date information, run computations, or use third-party services. The first plugins have been created by Expedia, FiscalNote, Instacart, KAYAK, Klarna, Milo, OpenTable, Shopify, Slack, Speak, Wolfram, and Zapier. Plugins could also increase safety challenges by taking harmful or unintended actions, so OpenAI is implementing several safeguards. For example, developers can add information to their robots.txt file to prevent ChatGPT from accessing their website.

2023.03.24
openai.com8 min readTLDR AI

EARLY EXPERIMENTS WITH GPT-4

GPT-4, the latest model developed by OpenAI, exhibits more general intelligence than previous AI models, as it can solve novel and difficult tasks spanning various domains, without needing special prompting, and its performance is strikingly close to human-level. This paper is a thorough study of GPT4 capabilities and shortcomings.

2023.03.24
arxiv.org30 min readTLDR AI

REPOCODER: STREAMLINING REPOSITORY-LEVEL CODE COMPLETION

RepoCoder is a simple and effective framework for repository-level code completion that utilizes a similarity-based retriever and a pre-trained code language model, and an iterative retrieval-generation paradigm. It significantly improves the zero-shot code completion baseline and outperforms the vanilla retrieval-augmented code completion approach. The proposed benchmark, RepoEval, provides valuable insights for future research.

2023.03.24
arxiv.org12 min readTLDR AI

GENERATIVE MODELS FOR MEDICAL

A good step towards evaluation, creation, and curation of generative models for a wide variety of tasks with a specific focus on medical. This project aims to assist by building tools for generative medical tasks and comes with metrics and tutorials to help you get started.

2023.03.24
github.comGitHub repositoryTLDR AI

MODELSCOPE

ModelScope is a web-based platform that provides interactive visualization and analysis of machine learning models' performance and interpretability, enabling users to explore and compare various models' results.

2023.03.24
github.comGitHub repositoryTLDR AI

OPENAI RESEARCHER ACCESS PROGRAM

Researchers can get subsidized access to models and continued access to codex. The form asks a set of questions about you and your research direction. It’s not clear what sorts of programs they’re looking to support, but application seems necessary for some research directions to continue.

2023.03.24
openai.com2 min readTLDR AI

DATA DEDUPLICATION FOR THE REST OF US

There have been many papers arguing both sides of the debate around the usefulness of data deduplication. It seems, in any case, that training on the same data point 300,000+ times is wasteful. A simple deduplication technique here is to embed your data points, cluster into k (e.g., 10k clusters), then remove a subset that has really high cosine similarity with other data points. This technique seems robust to the choice of embedding and the number of clusters, as long as the clusters are high enough. It also works across modalities.

2023.03.23
arxiv.org30 min readTLDR AI

TEXT TO ROOM

There has never been a community more interested in interior design than the generative AI community currently is. This work fits in well with the current trend. There are a lot of novel tweaks they needed to do to make the entire 3D space align with the prompt. In all, very interesting work for 3D asset generation.

2023.03.23
lukashoel.github.io4 min readTLDR AI

FREEDOM: A TRAINING-FREE CONDITIONAL DIFFUSION MODEL

The article presents FreeDoM, a training-free conditional diffusion model that leverages off-the-shelf pretrained networks to construct time-independent energy functions, enabling effective generation for various conditions. The proposed method offers advantages in simplicity, effectiveness, and low cost and has a broader range of applications than existing training-free methods, as demonstrated by experiments on diverse data domains, including image and latent code domains.

2023.03.23
arxiv.org15 min readTLDR AI

CHATLLAMA

A nice wrapper interface around the instruction tuned Alpaca model. Only the 7B model is available, with others coming. While the quality doesn’t match ChatGPT it is a fun open alternative with linked code.

2023.03.23
chatllama.baseten.coToolTLDR AI

GENERATIVE MODELS FOR MEDICAL

A good step towards evaluation, creation, and curation of generative models for a wide variety of tasks with a specific focus on medical. This project aims to assist by building tools for generative medical tasks and comes with metrics and tutorials to help you get started.

2023.03.23
github.comGitHub repositoryTLDR AI

NEXT.JS CHATGPT

Next.js ChatGPT is a responsible chat application powered by GPT-4.

2023.03.23
github.comGitHub repositoryTLDR AI

GOOGLE LAUNCHES BARD

It's here! Google’s long anticipated entry into the language models as a service game. Bard is a lightweight version of Lambda, which was a 70B parameter conversational agent trained a few years ago. They have spent a lot of effort making things safe and grounded in truthfulness, we’ll see when people start getting added to the waitlist if it lives up to the promises!

2023.03.22
blog.google4 min readTLDR AI