The first Open-Source txt2video 1.7 billion parameter diffusion model has been released and you can play with it now at HuggingFace.

Runway announces Gen-2, a multimodal AI system that can generate realistic videos from text.

Gen2 let's you go from:
text → video
image → video
text +image → video

A new note on "MEV Resilient Ethereum"

1. Defines MEV Resiliency framework
2. Reviews current state of MEV ecosystem & solutions being considered
3. Proposes two related MEV research projects

Microsoft speech recognition subsidiary Nuance debuts DAX Express, which can generate drafts of clinical notes after patient visits and is partly built on GPT-4

Coinbase compared 20 smart contract risk scores between ChatGPT and a manual security review

ChatGPT produced the same result as the manual review 12 times. Of the 8 misses, 5 of them were cases where ChatGPT incorrectly labeled a high risk asset as low risk.

While ChatGPT shows promise for its ability to quickly assess smart contract risks, it does not meet the accuracy requirements to be integrated into Coinbase security review processes.

OpenAI and Microsoft Are Partners, Until They Vie for the Same Customers

OpenAI’s ChatGPT has enraptured the business world since its November release and OpenAI is signing up customers eager to pay to use its artificial intelligence models in their own products. But the Microsoft-backed startup faces a surprising rival: Microsoft itself.
As part of its multibillion-dollar investment in OpenAI, Microsoft has the rights to sell the startup’s software through its Azure cloud business, even as OpenAI licenses its own software directly to customers. Microsoft also gets a share of OpenAI’s profits.

The offerings cost the same, a fraction of a cent per query. Meanwhile, all of OpenAI’s technology runs on Microsoft’s Azure cloud infrastructure rent free.

The dual offerings mean the companies are at times pitching the same customers on nearly identical products, putting salespeople at Microsoft in the uneasy position of trying to lure customers away from OpenAI while touting its technology.

Jensen Huang announced NVIDIA AI Foundations:

- Foundation Model as a Service is coming to enterprise, customized for your proprietary data.

- Multimodal from day 1: text LLM is just one part. Bring your images, videos, even 3D data! Let's build *custom* multimodal LLM and generative models for your use case.

- Biology is a unique type of LLM we support: this is the first large-scale AlphaFold API/finetuning service!

Powers up your drug discovery and research pipelines.

2023 is an inflection point. NVIDIA is going beyond a pure hardware provider and becoming an enterprise-first AI provider.

Generative AI has taken the world by storm. The FTC has published guidance for developers and deployers to help ensure that the technology complies with applicable law and operates responsibly.

This guidance should be factored into the "ethics and compliance by design" governance framework that extends throughout the product lifecycle.

Chip giants Nvidia, TSMC, ASML, Synopsys will work together on next generation chip manufacturing as making transistors tinier runs into the limits of physics, media report, noting TSMC's 2nm plant, Fab 20 in Hsinchu, will begin risk production in 2024.

The race has publicly begun:

- GPT still blows everyone else out of the water wrt coding abilities, with Claude coming a close second.
- Bard can't do coding yet (according to their FAQ)
- LLaMa / Alpaca, etc. is having a huge community-driven moment.

GPT-5 is probably being worked upon, while everyone still needs to catch up to GPT-3 level of programming ability and GPT-4 level of non-programming output.

Whether OpenAI remains the leader depends on how quickly others can catch up (if at all).

One of the most interesting advantages OpenAI has is their unwavering focus as an organization -- as opposed to the short-term promotion-driven culture at Google (eg. their Messaging story).

In Blockdata’s research, 40 of the top 200 companies made announcements regarding their blockchain initiatives in 2022.

Most of these companies are in the financial sector (19 companies), followed by consumer and retail and energy and utilities (6 companies each). Of the companies profiled, 30% were domiciled in the US (12 companies), followed by Japan (6 companies), and then by Germany and Spain (4 companies each), and 60% of these companies were in the production stage, 22.5% in the development stage, and 12.5% in the pilot phase.
We will take a look at announcements about blockchain and crypto and conduct a sector-wide analysis of these 40 companies.

Opera just announced GPT-based AI prompts in browser.

You can now interact with ChatGPT from anywhere to:

- Shorten or explain articles.
- Generate tweets, and content.
- Find relevant info based on text.

They also added ChatSonic which gives users capability to create AI-generated images.

We think Opera is already working on their own AI browser engine. Could be big.

Sparks of Artificial General Intelligence: Early experiments with GPT-4

Reports on their investigation of an early version of GPT-4, when it was still in active development by OpenAI.

DRPC developers say Ethereum remains dependent on a few major centralized RPC players, which undermines the blockchain's sustainability and security.

Canva announces a suite of new AI tools.

This follows just after Adobe Firefly yesterday. What's with the timing!?

Here's what's new for Canva's 125 million users:

First, Canva already had basic AI features:

- Text to Image
- Magic Write, an AI copywriting assistant

Nasdaq eyes crypto custody launch by end of second quarter — link

A recently developed machine learning model has given scientists a new schema for predicting the scent of individual molecules: metabolism.

Sequoia Capital's newest market map! The AI revolution is the most consequential computational trend in decades.

Mira Murati (CTO of OpenAI) + Sam Altman (CEO of OpenAI) + Greg Brockman (President of OpenAI) + Ilya Sutskever (Chief Scientist of OpenAI)

Can LLMs like GPT learn from their mistakes and self-reflect just like humans?

The answer is yes and the method resulted in 33-35% accuracy improvement on HotPotQA!

Paper
Code

Google introduced a multilingual dataset for parsing realistic task-oriented dialogues, which includes about half a million realistic conversations in six languages, many of which exhibit common phenomena like disfluencies, user revisions and code-mixing.