Top stories today:

1. OpenAI launches Sora, text-to-video AI model, in research preview
2. Gemini 1.5 Pro launches with 1M-token context window
3. TikTok launches app for Vision Pro
4. Coinbase returns to profitability as revenue +50% YoY to $905M
5. Mr. Beast's empire at ~$700M in revenue in 2023

0. Data and calendar

1. OpenAI launches Sora, text-to-video AI model, in research preview

• <=1-minute videos can be generated with Sora, vs. 3 seconds for some of the already available options like Pika or Runway.

• Available only to red teamers and “a number of visual artists, designers, and filmmakers.”

Our view: Sora is a leap in quality above other text-to-video AIs, and it will only get better from here

• Detecting that these videos are AI-generated can be hard, especially if one isn't focused on doing so.

• These models will increasingly threaten stock video and low-quality videographers, just as ChatGPT/DALL-E has been doing for text/images.

• Google Deepmind is rumored to launch its “Gemini Video” soon, which will be exciting to watch.

• Relatedly, the FTC is seeking public comment on a rule that would punish AI impersonation.

2. Gemini 1.5 Pro launches with 1M-token context window

• 1M token is approximately 100K lines of code/text or 1 hour of video.

• Similar in quality to Gemini 1.0 Ultra, according to Google, but we'll wait on 3rd-party benchmarks.

• Only available now to business users and developers, via Google's Vertex AI and AI Studio.

• In testing: 10M-token context window.

Gemini 1.5 Pro is at the level of Gemini 1.0 Ultra and beats GPT-4 in information retrieval, according to Google

4. TikTok launches app for Vision Pro