AI video is taking off

AI Video’s ChatGPT Moment?

This Wednesday, Pika Labs unveiled its impressive new AI video platform Pika 1.0—and many in the AI space are calling it “AI video’s ChatGPT moment.”

ChatGPT completely changed the way we create text…and Pika 1.0 is expected to do the same with video creation. Here’s why: 

• Pika 1.0 generates videos across a wide range of styles—including 3D animation, anime, and live-action content—from simple text prompts.

• It can seamlessly convert photos to videos (like bringing a meme to life) or transform videos into different styles (like “live action” to “cartoon”).

• Users can edit content elements (like changing a character’s clothing), camera angles, and movements using simple text prompts.

• Pika 1.0 can extend video lengths by predicting content beyond the original video.

Pika Labs’ first version is already used by over half a million people on Discord—and users are seriously loyal, spending up to 10 hours a day on the platform. Now, Pika 1.0 will be available on an easy-to-use web interface…and the waitlist is already growing. 

Moving forward: At the current pace of development, it’s only a matter of time until we have full text-to-movie tools. Soon, anyone with an idea will be able to create a high-quality film.

Let Our AI Chatbots Take Care of Your Customer Support

• ✅ Fully automated AI chatbots to help you or your customers 24/7/365.

• 🎓 Train the Chatbot on your company using your website, PDFs, Word Docs, CSVs, XLS files, or even YouTube videos.

• 📋 Have the Chatbot automatically collect leads and email them to you or your team.

• 🚀 Send prompts to your chatbot via any App on Zapier.

• 💰 Fully white label our chatbots for reselling to your own clients.

• 🎤 Tune the Chatbot to match your brand voice.

For a limited time, we're offering the chance to claim a FREE AI-powered chatbot. Sign up and claim your free Chatbot now (no credit card required).

AI’s Material Science Breakthrough

Google DeepMind researchers developed a new AI tool that revolutionizes material science—and it could have massive implications across industries from renewable energy to computing.

The details: The tool, Graph Networks for Materials Exploration (GNoME), predicted the structures of over 2.2 million new materials—381,000 of which are predicted to be stable.

Here’s why that matters: 

• Stable materials are substances or compounds that maintain their structural properties over time—and they’re incredibly important in the production of everything from solar panels to computer chips.

• The DeepMind team said the model’s findings are the equivalent of nearly 800 years’ worth of knowledge—AKA this tool has the potential to seriously supercharge material development.

If you’re curious how GNoME works, check out Google DeepMind’s paper published in Nature. 

Big picture: GNoME is significantly accelerating materials discovery, demonstrating AI’s potential to completely transform research. AI’s next challenge? Actually synthesizing materials to bring them to market. And I have a feeling that’s not so far away…

Amazon Is Joining the AI Race

Amazon dropped some major AI announcements at its re:Invent conference this week. Let’s dig into two heavy hitters, both available to AWS customers:

1) Titan is Amazon’s new text-to-image model, specifically made for businesses:

• Unlike other AI image generators, Titan doesn’t have its own website or app; it’s a tool for developers and businesses to integrate into their own platforms.

• Images generated with Titan will automatically include invisible watermarks. According to AWS, the watermarks “will not interfere with the visual, have no latency, and cannot be cropped out.”

2) Amazon Q is a chatbot and productivity assistant for businesses:

• Amazon Q can be connected to business software tools like Slack, Salesforce, and Microsoft 365, so users can ask questions about their company’s private data.

• Unlike ChatGPT, which only runs on GPT-4, Amazon Q isn’t restricted to a specific foundation model. Because it runs on AWS’ repository of AI models (including Meta’s Llama 2 and Anthropic’s Claude 2), companies can choose which model works best for them.

Zoom out: Amazon has been racing to catch up with its Big Tech competitors in AI. Its launch of Amazon Q follows Microsoft’s Copilot and Google’s Duet AI. Meanwhile, Titan’s watermarking feature follows similar efforts from Google and Adobe to label AI-generated images.

Bonus Story:

LLMs just got way better at logical reasoning. Meta researchers unlocked a groundbreaking training technique called S2A. If you’re an AI builder—or just interested in the backend of LLMs—you should check out this paper.

• 18 countries signed the first coordinated international deal that outlines safe AI development principles. 

• Inflection AI claims its new AI model is the second-most powerful AI behind GPT-4.

• Amazon is using AI to improve the holiday experience. 

• Sports Illustrated is under fire for publishing AI-generated articles under fake author personas. 

• ElevenLabs launched a grant program for AI startups and solopreneurs to put AI voice synthesis to work. 

DIY: Turns out making high-quality movie trailers using AI is easy (and pretty fun). Here’s my step-by-step tutorial:

Deep dive: Want to understand how large language models actually work? Here’s a comprehensive introduction to LLMs: