Baby-AGI is taking the world by storm

Here's an implementation within the LangChain framework, allowing you to easily substitute in other vectorstores and other LLMs.

DeepSpeed Chat. Impressive open-source effort by Microsoft!

DeepSpeed Chat offers an end-to-end RLHF pipeline to train ChatGPT-like models. This is the missing piece from other efforts like Alpaca and Vicuna. The RLHF pipeline is replicated from the InstructGPT paper.

The other challenges are cost and efficiency. DeepSpeed Chat aims to make this process more accessible and affordable through a unified hybrid engine (DeepSpeed-HE) for RLHF. For example, DeepSpeed-HE can train an OPT-66B model in 2.1 days for $1620.

With further scalability (e.g., using multi-node multi-GPU systems), DeepSpeed-HE can train an OPT-13B model in 1.25 hours for $320 and an OPT-175B model in under a day for $5120. That's a big deal!

Elon Musk Makes Big Investment in Generative-AI Project at Twitter

He recently purchased roughly 10,000 GPUs for the platform

Two AI engineers from Alphabet's DeepMind joined Twitter last month.

Intel plans to make graphics processors (GPUs) that meet US rules for export to China, media report, adding that Intel CEO Pat Gelsinger is in China now meeting government officials.

Current architecture of research agent enables learning through Google Search, Pinecone, multiple agent feedback, and incremental draft improvement.

Based on babyAGI

Dolly is the first *commercially viable*, open source, instruction-following LLM.

Dolly 2.0 is available for commercial applications without having to pay for API access or sharing data with 3rd parties.

Google AI researchers have developed a new aging clock that can predict a person's biological age with greater accuracy than previous methods.

The clock, which is based on deep learning models of retinal images, was trained on a dataset of over 800,000 people.

This is an exciting development in the field of aging research. The aging clock could have a major impact on our understanding of aging and our ability to prevent and treat age-related diseases.

The Startup Industry Map for Pre-Seed Fundraising

It’s based on US startups mostly in SF, California and New York.
Data is from January 1, 2023 to March 14, 2023 (so incredibly recent).

What are the key findings?

• Biotech companies raise the most money at the highest valuation caps - but not a lot of them actually get funded.

• SaaS is middle of the pack in valuation cap and total raised, but by far the biggest category.

• Investors seem less enthused about food startups lately.

New stablecoin bill draft introduced in US House of Representatives

Members of the U.S. House of Representatives are taking another crack at creating a comprehensive regulatory framework for stablecoins, like USDC and Tether, digital assets intended for payments whose price is supposed to remain the same at all times. 

The House Financial Services Committee released a new discussion draft bill with no official notice on Saturday.

Here are highlights of what the draft bill would do if it becomes law: 

1. Put the Federal Reserve in charge of nonbank stablecoins

The U.S. central bank would approve and regulate non-bank companies like Circle and Tether that currently issue or want to issue their own stablecoins in the U.S. Credit unions and banks that want to issue their own stablecoins could do so with approval from the main financial regulator they fall under, the National Credit Union Administration, Federal Deposit Insurance Corp. or Office of the Comptroller of the Currency. Failure to register would be punishable by up to five years in prison and a $1 million fine. Any issuer that wants to do business in the U.S., regardless of where the company is based, would need to register. 

2. A (temporary) ban on new stablecoins without fiat backing
The new draft bill includes a two-year ban on stablecoins that aren't backed by a hard asset, and it directs the Treasury Department to lead a study on the topic of such "endogenously backed" stablecoins. Tokens already in existence before the bill passes into law would be grandfathered in.

3. Allow the government to set interoperability standards

Banking regulators and the National Institute of Standards and Technology would have the ability to set standards for interoperability between stablecoins to allow for ease of use, including mandatory technical and legal specifications to allow users to clear and settle across different payment systems without buying native stablecoins for each. 

4. Direct the Fed to study a digital dollar

If the bill becomes law, Congress and the president would direct the Federal Reserve to study the effects of a digital dollar issued by the central bank. The Fed has already begun studying whether to issue a digital dollar, but the bill would mandate certain areas of focus, like the potential impacts on monetary policy, financial stability and privacy for individuals.

3 CAMEL datasets released: Physics, Chemistry and Biology. Each dataset contains 20K problem-solution pairs, consisting of 25 topics, 25 subtopics and 32 problems for each "topic, subtopic" pair generated and solved by GPT4.

Web LLM is insane.

1) Go download the latest Chrome beta, which shipped WebGPU support: google.com/chrome/beta/

2) Now use a 7B-param LLM in your browser! mlc.ai/web-llm/

3) Marvel at the "How" section on their GitHub: https://github.com/mlc-ai/web-llm

NY financial regulator adopts virtual currency assessment rule

The regulation will require companies to meet rigorous standards for capitalization, cybersecurity protection, and anti-money laundering protocols, NYDFS said in a statement Monday.

“As the first prudential regulator of virtual currency in the nation, New York has created a framework that sets the highest standards for safety, soundness, and consumer protection while fostering responsible growth,” said NYDFS Superintendent Adrienne Harris. “This regulation provides the Department with additional tools and resources to regulate the virtual currency industry now and in the future as innovators create new products and use cases for digital assets.”

Today OpenAssistant released, we got a contender for open source chat!

The future of AI development depends heavily on high quality datasets and models being made publicly available, and that’s exactly what this project does.

https://huggingface.co/OpenAssistant

Meta AI announced DINOv2

The new Self-supervised Vision Transformer Model can be used as a backbone for almost all your CV tasks. No fine-tunning needed.

• Train CV models without the need for large amounts of labeled data.

• Multipurpose backbone: image classification, segmentation, image retrieval, and depth estimation.

• Learn features directly from images without relying on text descriptions, which can lead to a better understanding of local information.

• Can learn from any collection of images.

• The pretrained version of DINOv2 is already available and competes with CLIP and OpenCLIP on a wide array of tasks.

Code https://github.com/facebookresearch/dinov2

Demo https://dinov2.metademolab.com/

Huge announcements in the world of AI today from Adobe, Elon Musk, the NBA, and Meta.

Here's a rundown on everything you need to know:

1. Adobe adds Firefly to video editing

Edit videos using simple text commands, plus:

-AI color corrections/enhancements
-Animated text and motion graphics generation
-Matching B-roll footage
-Music and sound effects
-Auto-generated storyboards

2. Elon Musk reveals TruthGPT

In a recent interview, Elon Musk said he will start a maximum truth-seeking AI called ‘TruthGPT.’

The AI will try to understand the nature of the universe.

3. ChatGPT got featured in a new NBA commercial with Jack Harlow

ChatGPT has gone mainstream, fast.

Everyone is talking about ChatGPT and Generative AI.

Beyond the hype and novelty applications, its important to start thinking about how it impacts our economy and our businesses.

This is a good way to get your thinking started

This roadmap is a good place to start.

UK’s new technology department to tackle country’s metaverse, Web3 strategy

The U.K.'s newly formed Department of Science, Innovation and Technology will advance the country's metaverse and Web3 strategy.

The country’s 2023 Spring Budget, released in March, said the government wanted to “lead on the future of web technology, sometimes known as Web3 or the metaverse.”

The new tech department, set up in February, will lead the charge in this work.

The department’s work will not focus on specific technologies like blockchain and virtual reality but will instead look at potential economic growth opportunities, investment and business models associated with concepts including the Metaverse and Web3, along with implications for regulation, according to the source.

NVIDIA just released a very impressive text-to-video paper

Video Latent Diffusion Models (Video LDMs) use a diffusion model in a compressed latent space to generate high-resolution videos.

Here's a brief overview of how it works:

1. Pre-train image LDM on a dataset of images.

2. Turn the image LDM into a Video LDM by adding temporal layers to model video frames.

3. Fine-tune the Video LDM on encoded video sequences to create a video generator.

4. Temporally align diffusion model upsamplers to generate high-resolution videos.

5. Validate Video LDM on real driving videos of 512x1024 resolution, achieving state-of-the-art performance.

6. Apply the approach in creative content creation with text-to-video modeling.

Paper: https://arxiv.org/abs/2304.08818

New smart glasses tell u what to say on dates using GPT-4

A team of crafty student researchers at Stanford University have come up with a pair of smart glasses that can display the output of OpenAI's GPT-4 large language model — potentially giving you a leg up during the next job interview, or even coaching you during your next date.

The device, dubbed rizzGPT, offers its wearer "real-time Charisma as a Service" (CaaS) and "listens to your conversation and tells you exactly what to say next," as Stanford student Bryan Hau-Ping Chiang explained in a recent tweet.

The glasses were made using a monocle-like device that can be snapped onto practically any glasses, built and donated by Brilliant Labs. It features a camera, microphone, and a high-resolution display that can output text generated by GPT-4.

OpenAI's speech recognition software, Whisper, allows the glasses to feed speech directly to the chatbot, which can generate answers in a matter of seconds to its wearer.