Search the archive

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

Results

14,500 matches

Sponsored

2023.02.20
linearb.ioSponsorTLDR

2023.02.20
theverge.com6 min readTLDR

Sponsored

2023.02.20
semanticlayersummit.comSponsorTLDR

EFFICIENCY’

Meta managers have reportedly given 'subpar' performance reviews to thousands of employees, raising fears of more layoffs. Around 10% of workers received poor reviews. The company's metaverse division is seemingly safe from cuts. Reality Labs was responsible for $13.7 billion in losses last year. Despite this, Mark Zuckerberg remains committed to the division.

2023.02.20
archive.ph3 min readTLDR

IT.

2023.02.20
mashable.com4 min readTLDR

THREADLESS PROCESS INJECTION

Process injection is a technique used for executing code from within the address space of another process, commonly used to mask activity within legitimate processes.

2023.02.20
github.comGitHub repositoryTLDR

1B MULTIMODAL MODEL OUT PERFORMS 175B COUNTERPART

Narrow models can still, unsurprisingly, beat larger general purpose models. This paper is another example, the authors use a small multimodal model to beat GPT3 by 16% on the visual / text benchmark ScienceQA. It's only one benchmark, but it's good to see improvements on more reasonably sized models.

2023.02.20
arxiv.org33 min readTLDR AI

THE STABLE ENTROPY HYPOTHESIS

This paper submits an hypothesis that “human-like” text generation lies in a narrow region of low entropy. Degenerate text which has repeated words, incoherent sentences, and grammatical errors lie away from this region. They provide supporting evidence and present an entropy aware decoding scheme.

2023.02.20
arxiv.org45 min readTLDR AI

MORAL SELF-CORRECTION IN LARGE LANGUAGE MODELS

Large language models (LLMs) exhibit harmful biases that can get worse with size. Reinforcement learning from human feedback (RLHF) helps, but not always enough. The authors show that simple prompting approaches can help LLMs trained with RLHF produce less harmful outputs.

2023.02.20
arxiv.org20 min readTLDR AI

LANGCHAIN FOR THE WEB

Similar to other recently released libraries, Langchain now has a JavaScript library. Use this to build LM powered web apps. Importantly, interfacing with the python version is a huge focus specifically around serialization of objects.

2023.02.20
github.comGitHub repositoryTLDR AI

TALK-TO-CHATGPT

Talk-To-ChatGPT is a Google Chrome extension that allows users to talk with ChatGPT using their voice and listen to the AI’s answer with a voice.

2023.02.20
github.comGitHub repositoryTLDR AI

SUMMARY OF CREATIVE AI RESEARCH AT SONY

A list of papers from the Sony creative AI research group. They focus on deep generative modeling, music, and cinematic AI. Lots of great links here to put on your reading list if you want to understand the state of the art for music.

2023.02.20
sony.github.io3 min readTLDR AI

PROBE

Nishad Singh is expected to plead guilty to US criminal charges related to a multiyear fraud at crypto exchange FTX. The Commodity Futures Trading Commission and the Securities and Exchange Commission are planning to sue Singh over his alleged role in the scheme. Singh had a close relationship with Bankman-Fried. He helped write the software the exchange was built on. The deal, which could involve cooperating with authorities, has yet to be finalized.

2023.02.20
archive.ph4 min readTLDR Crypto

2023.02.17
arstechnica.com2 min readTLDR

DOWN

2023.02.17
arstechnica.com3 min readTLDR

INQUERY

Inquery is a utility that triggers webhooks in response to changes made in a Postgres table. It works by sending low-latency websocket messages to a Go application, which then calls the configured webhook(s) with a JSON payload that includes values from the database row. Use cases include sending notifications via Slack, Email, Text Message or Push Notification, calling serverless functions, triggering analytics events, and streaming data real-time.

2023.02.17
github.comGitHub repositoryTLDR