170 | 🤖🎥 OpenAI Broke the Internet Last Night

In today’s Brainyacts:

1. OpenAI broke the internet with Sora

2. Google debuts Gemini Pro 1.5 with massive context

3. OpenAI to take on internet search and other AI model news

4. Author copyright claims not holding and other AI-related content

👋 to new subscribers!

To read previous editions, click here.

Lead Memo

🤖🎥 OpenAI Broke the Internet Last Night

The internet blew up yesterday afternoon. Why? Sam Altman, CEO of OpenAI, was taking prompt suggestions for testing out Sora. Sora is a text-to-video capability that OpenAI is currently testing and Sam and team were playing with in the public. They got prompt suggestions such as:

Two golden retrievers podcasting on top of a mountain

a wizard wearing a pointed hat and a blue robe with white stars casting a spell that shoots lightning from his hand and holding an old tome in his other hand

A bicycle race on ocean with different animals as athletes riding the bicycles with drone camera view

Details of Sora:

Sora will be able to understand the nuances of the prompt as well as how various objects behave in the physical world.

Sora also generates an entire video at once, rather than creating it frame by frame. That helps avoid what has been a continuity challenge with other approaches — ensuring a subject stays the same even when it goes out of view temporarily.

OpenAI stressed that it doesn't plan to make Sora broadly available any time soon as it continues to work on a range of safety issues, including efforts to reduce misinformation, hateful content and bias — and also to clearly label the output as generated by AI.

OpenAI said in its blog post that it is making Sora available to some early testers, including both "red teamers" who will look to poke at potential security concerns, as well as several visual artists, designers, and filmmakers, to get feedback on how the model could help them with their work.

What some are saying: 

Reactions to Sora were a mixed bag of emotions:

• AI creator and ex-Googler Bilawal Sidhu said: "OpenAI just shattered the 'Visual' Turing Test, 'Is it real or is it fake?' isn't just for photos now. Videos too have fallen, and I couldn't be more excited."

• Filmmaker Justine Bateman, who has been highly critical of the technology, posted: "Every nanosecond of this #AI garbage is trained on stolen work by real artists. Repulsive."

Big questions remain:

While this is mind-blowing, it highlights the big questions of how can we prevent deep fakes AND what, if any, rights are there in the underlying training data, AND are any of these works protected in any manner?

Sorry folks, I don’t have the answers and feel bad that I keep raising these questions. This is just an unprecedented time and until there are answers, we have to keep asking the questions.

Spotlight

♾️ 🫙 Google creating the ability for Huge Context Windows (or, soon AI models will have immense memories)

Google has just released Gemini Pro 1.5. Recall Google has three Gemini models - Ultra (most capable), Pro, and Nano (think on-device models).

Pro 1.5 is said to perform as well as Ultra 1.0 - which is significant because it is a smaller model yet can do what the larger (more expensive model does). While Pro currently offers a context window of 128k tokens, a larger 1 million token version is being released to a select group of developers. This 1 million context window allows for active conversations with up to 1 hour of video, 11 hours of audio, or about 700,000 words (which is over 100k more words than Tolstoy’s War and Peace).

An AI model’s “context window” is made up of tokens, which are the building blocks used for processing information. Tokens can be entire parts or subsections of words, images, videos, audio, or code. The bigger a model’s context window, the more information it can take in and process in a given prompt — making its output more consistent, relevant, and useful.

But there is more . . . in testing, they have actually far exceeded the 1M token context window - keeping extremely high success rates at the 10M token level. This is a mind-blowing performance. 👇

Why does this matter to you?

1. Improved Coherence: AI can manage longer texts seamlessly, ensuring relevance and continuity in conversations and documents.

2. Enhanced Memory: Better understanding and recall of previous interactions, offering more personalized and context-aware responses.

3. Richer Contextual Analysis: Deeper understanding for nuanced tasks like legal analysis or sophisticated content creation.

4. Complex Problem Solving: Ability to tackle intricate problems by synthesizing information from extensive sources.

5. Versatile Applications: Expanded utility across diverse tasks, from creative writing to comprehensive academic or legal research.

6. Educational Support: Assists in learning by integrating and summarizing knowledge from vast materials.

7. Interdisciplinary Research: Facilitates research across fields by handling complex, multi-source information.

For more details, check out the research paper HERE.

AI Model Notables

► OpenAI may take on Google with its own internet search tool

► V-JEPA, one of Meta’s AI models, learns by watching videos.

►  X is adding AI into its Explore tab, with trending topics set to get more context via Grok-generated summaries.

► Google to set up new AI hub in France

► Gemini’s mobile app will start rolling out outside the US

► Apple buys domain suggesting generative AI additions may come to iWork

News You Can Use:

➭ Judge rejects most ChatGPT copyright claims from book authors

➭ Air Canada's chatbot gave a B.C. man the wrong information. Now, the airline has to pay for the mistake

➭ Australia announced new artificial intelligence expert group

➭ Japan's ruling party pushes for AI legislation within 2024

➭ Peru advances with EU-inspired artificial intelligence regulation

➭ EU provides final seal of approval for world's first AI laws

➭ Spanish artist Alicia Framis set to become first woman to marry ai-generated hologram

➭ Microsoft and OpenAI: hackers are using ChatGPT for Cyberattacks

Who is the author, Josh Kubicki?

Some of you know me. Others do not. Here is a short intro. I am a lawyer, entrepreneur, and teacher. I have transformed legal practices and built multi-million dollar businesses. Not a theorist, I am an applied researcher and former Chief Strategy Officer, recognized by Fast Company and Bloomberg Law for my unique work. Through this newsletter, I offer you pragmatic insights into leveraging AI to inform and improve your daily life in legal services.