Search the archive

VISIONOS | APPLE DEVELOPER DOCUMENTATION

Apple has released their developer documentation for VisionOS, the operating system for the Apple Vision Pro.

EXPLORING GITHUB WITH THE REDESIGNED NAVIGATION, NOW IN PUBLIC BETA

GitHub’s redesigned navigation is enabled for all users as a public beta. The navigation has been designed to make it easier to find your way around the site and quickly access your most-used features and functionality.

USING TYPESCRIPT GENERATORS

Normal functions run from top to bottom and then exit. Generator functions also run from top to bottom, but they can be paused during execution and resume later from the same point. This article will teach you how to use generator functions in TypeScript and cover a few different examples and use cases.

Nigerian engineering students are increasingly turning to Indian YouTubers for supplementary academic support. Though the Indian educational channels weren’t created to cater to Nigerian students, they’re now trying to make their lessons accessible to all.

INSIDE THE WHITE HOUSE’S URGENT PUSH TO REGULATE AI

PIXAR USED AI TO STOKE ELEMENTAL'S FLAME

Pixar turned to AI to solve a unique animation challenge for its new movie 'Elemental', using AI to animate a character made of fire. Paul Kanyuk, a technical supervisor at Pixar, applied Neural Style Transfer (NST) technology to animate the fiery character, Ember. Despite the hefty computational demand, the team achieved a 20X speed-up by leveraging GPUs, heralding a new intersection of art and technology in animation.

CHINA'S UNDERGROUND MARKET FOR NVIDIA AI CHIPS

SYNTHETIC DATA AND NEXT LEVEL CURATION GIVE BIG HUMANEVAL BOOST

Textbooks may be all you need for extreme AI performance. By using a highly curated set of data and synthetic textbooks with exercises, a 1.3B parameter model achieves 50+ on HumanEval, a challenging code benchmark. For context, this rivals ChatGPT and far surpasses the original 12B Codex model. This paper follows the Tiny Stories line of work and has lots of nice tidbits.

IN LANGUAGE MODELS

Language models often hallucinate/confabulate information. This is likely because they are trained to imitate expert sequences which in turn causes them to treat their own (potentially wrong) generations as ground truth. Therefore, as sequence lengths grow, models are more likely to make stuff up. However, if you treat it as a Reinforcement Learning problem and find the next most probable action - while including a backspace action - you can dramatically reduce these hallucination issues.

HIGH THROUGHPUT AND MEMORY EFFICIENT TRANSFORMERS

As Transformers have started to gain more and more traction, they are being served far more often than trained. This can be quite slow and expensive, but with some clever optimizations around memory layout on GPU, they get many multiples faster.

WANDA: A NEW PRUNING METHOD FOR LARGE LANGUAGE MODELS

The repo presents Wanda, a unique method for reducing the complexity of large language models by selectively dropping less significant network weights. Unlike other methods, Wanda doesn't require retraining or intensive computation, providing effective model pruning while maintaining performance.

STANDARDIZING SPATIO-TEMPORAL PREDICTIVE LEARNING

This repo introduces OpenSTL, a benchmark for models that predict future patterns based on past frames. It categorizes approaches and evaluates them on various datasets.

AI IS A LOT OF WORK

ANT IS DEVELOPING AN LLM

Jack Ma-backed Ant Group Co. is developing large-language model technology that will power ChatGPT-style services.

FRESHWORKS UNVEILS NEW GENERATIVE AI ENHANCEMENTS

SECRET INVASION’S OPENING CREDITS ARE AI-MADE

SCHUMER UNVEILS STRATEGY TO REGULATE AI

OPENAI PLANS APP STORE FOR AI SOFTWARE, THE INFORMATION REPORTS

OpenAI plans to launch a marketplace for AI models built on top of its technology. Enterprise customers often tailor OpenAI's technology to their specific use cases. The marketplace would allow these customers to make their models available to other businesses. It will compete with other app stores, including one by Microsoft, and help OpenAI's technology reach more people.

RIVIAN IS THE NEXT AUTOMAKER TO ADOPT TESLA'S CHARGING PLUGS

The Combined Changing System (CCS) standard's connector and cable are big and unwieldy, unlike Tesla's more elegant alternative, the North American Charging Standard (NACS). Rivian will start building NACS ports into its vehicles in 2025. It will make adaptors available for its customers. Tesla charging sites will be added to Rivian's mobile and vehicle navigation apps.

DEEPMIND'S CO-FOUNDER SUGGESTED TESTING AN AI CHATBOT'S ABILITY TO TURN $100,000 INTO $1 MILLION TO MEASURE HUMAN-LIKE INTELLIGENCE

Mustafa Suleyman, formerly head of applied AI at DeepMind and now CEO and co-founder of Inflection AI, believes that the Turing test is an outdated method to test AI intelligence. The Turing test doesn't reveal information about what systems understand, whether they have established complex internal monologues, or if they engage in planning over abstract time horizons. Suleyman proposes a concept called artificial capable intelligence (ACI), where bots are able to complete short-term goals and tasks with little human input. Bots can achieve ACI by passing a modern Turing test that involves turning a $100,000 seed investment into $1 million. Suleyman believes that AI will be able to achieve that milestone in the next two years.

ROBOCAT: A SELF-IMPROVING ROBOTIC AGENT

RoboCat is a self-improving AI agent for robotics. It learns to perform tasks across different arms and then self-generates new training data to improve its technique. RoboCat can solve and adapt to multiple tasks across different robots. It learns much faster than other state-of-the-art models, picking up new tasks with as few as 100 demonstrations. A video demonstrating RoboCat's abilities is available in the article.

NOVEL

Novel is a Notion-style editor with AI-powered autocompletions. It can be deployed to Vercel with one click. A 30-second video demo is available.

HOW TO USE GITHUB COPILOT: PROMPTS, TIPS, AND USE CASES

This guide shares examples and best practices for communicating with GitHub Copilot. Generative AI coding tools are transforming the way developers work, but just like with any emerging tech, there is a learning curve. Learning how GitHub Copilot processes information can help developers adjust how they communicate with it.

NOT THAT KIND OF OPEN

Some Mastodon/Fediverse administrators have pledged to preemptively block Meta's upcoming ActivityPub service. ActivityPub makes social networking truly open. Blocking Meta from participating in the protocol seems to go against that ideal. There are billions of people who use Meta's services and there is no reason to cut them off from the ActivityPub social world.

EVERYONE HAS A SIDE HUSTLE NOW. IS THAT A GOOD THING?

Almost everyone has a side hustle. Even many leaders have significant side hustles. Remote work means that people no longer have to hide their other gigs from their co-workers. Many founders and CEOs don't like this. Side hustles are a distraction to workers' full-time jobs. While people can be good at doing more than one thing, being truly great requires full commitment.

EMERGING ARCHITECTURES FOR LLM APPLICATIONS

This post shares a reference architecture for the emerging LLM tech stack that shows the most common systems, tools, and design patterns used by tech companies and startups.

ARGUABLY

arguably turns functions into command line interfaces, automatically generating help messages using docstrings.

IOS 17 AND MACOS SONOMA AUTOMATICALLY GENERATES APPLE ID PASSKEYS

iOS 17, iPadOS 17, and macOS Sonoma users are automatically assigned a passkey to their Apple IDs for use on iCloud and Apple sites.

PG_EASY_REPLICATE

pg_easy_replicate is a command-line orchestrator tool that makes setting up logical replication between two PostgreSQL databases simple.

AI IS A LOT OF WORK

This article tells the story of the people that sort and tag the data that AI learns from.

RELIABLE

Commands or functions trapped in EXIT can ensure that bash scripts always perform necessary cleanup operations even when something unexpected goes wrong.

OPENAI PLANS AN APP STORE

OpenAI plans to launch a marketplace that will allow developers to sell their AI models built on top of its own AI technology.

EMERGING ARCHITECTURES FOR LLM APPLICATIONS

OPENAI LOBBIED EU TO WATER DOWN REGULATION

DEEPMIND'S NEW ROBOTICS CONTROLLER

A NOVEL APPROACH TO COLLABORATIVE IMAGE RECOGNITION

PhotoBook is an interactive game in which two players find common images from their own private sets, pushing machines to learn human-like communication around visual context. The approach, based on the DeBERTa model and CLIPScore features, bypasses the need for reference chains, directly predicting whether an image is shared.

LOMO: EFFICIENT LLM TRAINING

This research introduces LOw-Memory Optimization (LOMO), an optimizer designed to dramatically decrease the resources needed for Large Language Models (LLMs) training, a process that currently requires extensive GPU resources. LOMO's innovative approach allows the fine-tuning of a 65 billion parameter model using just a single machine with 8 RTX 3090 GPUs, each with 24GB memory, reducing memory usage to just 10.8% of the standard method.

CLIP2PROTECT: A METHOD FOR PROTECTING FACIAL PRIVACY

The Clip2Protect project introduces an innovative two-step approach to ensure facial privacy, especially in an age of powerful face recognition systems. It creates 'makeup' faces that look natural, yet fool facial recognition software, showing significantly better performance than previous privacy-enhancing methods.

AUTOLABEL

Autolabel is a Python library to label, clean and enrich text datasets with any Large Language Models (LLM) of your choice.

ASSET MANAGERS PRESSURE TECH COMPANIES OVER AI MISUSE

Institutional investors are increasing pressure on technology companies to take responsibility for the potential misuse of artificial intelligence as they become concerned about the liability for human rights issues linked to the software.

NEED TO DEDUP LOTS OF DATA?

The quality of the training mix for AI models is key to getting good performance. In code models, deduplication of data tends to matter quite a bit. It's not clear if it matters as much for text. However, in any case, this blog presents a great description of the techniques used to deduplicate large data sets of text-like data.

WHAT I LEARNT AFTER WRITING 25+ ARTICLES BY USING AI

AI should be used as a tool to enhance productivity, imagination, and creativity, and not as a means of automation. This article provides insights on how to use AI without sacrificing quality.

ENCODEC IN HUGGINGFACE TRANSFORMERS

OPERA PREPARES AI-INFUSED BROWSER

QUIVR

Dump all your files and thoughts into your GenerativeAI Second Brain and chat with it.

OPERATIONS

EDX Markets, a cryptocurrency exchange backed by Citadel Securities, Fidelity Investments, and Charles Schwab, recently launched, quietly beginning to execute trades in the past few weeks. The platform aims to differentiate itself by operating as a noncustodial exchange, thus not directly handling customers' assets. Instead, it serves as a marketplace for firms to negotiate prices and settle trades independently. EDX intends to target brokers and institutional investors rather than individual ones. As of now, the exchange plans to list only four cryptocurrencies - BTC, ETH, LTC, and BCH, aiming to avoid potential issues with the SEC.

BLACKROCK, THEN BITWISE — HOW THE SPOT BITCOIN ETF FILINGS DIFFER

Bitwise has revived its proposal for a spot Bitcoin ETF following a similar move by BlackRock and rumors of Fidelity pursuing one as well, even though the SEC has never approved such an offering. The Bitwise filing, like BlackRock's, emphasizes surveillance of Bitcoin via the CME futures markets. Analysts suggest Bitwise's move is a direct response to BlackRock's bid, and both filings cite the SEC's approval of a Bitcoin futures fund by Teucrium as a precedent. The outcome of Grayscale Investments' lawsuit against the SEC could further influence the prospects for approval of a spot Bitcoin ETF.