Gemini 1m context, Sora text-to-video, & $7 trillion for chips
AI went through another sea change last week. It started when Google launched Gemini 1.5 with a 1 million token context limit ??
That's ...
?1 hour of video
?11 hours of audio
?30,000 lines of code
?700,000 words... in a single prompt.
And, it maintains near-perfect “needle” recall (>99.7%) up to 10 million token context lengths.
??3 hours of video(2.8M tokens)
??22 hours audio (2M tokens)
??7 million words (10M tokens)
Not to be outdone, OpenAI introduced Sora – their new text-to-video model that creates realistic 60-second movies from text prompts.
Here's one of the videos everyone is talking about:
(They even got the hands right.)
AI makes things that seemed impossible happen every week and is actively shaping geopolitics. While Sam Altman seeks up to $7 trillion in funding for AI chip infrastructure, Nvidia's CEO is selling world leaders on the importance of sovereign AI.
If you're trying to stay ahead of the curve as a business (or a professional) here are four questions to ask yourself.
Other Useful News
(In a different week, each of these would be headlines.)
PS
We hope these newsletters help you stay informed. We want to know what you think is useful and you can tell us below.
GenAI Growth @Vellum
1 年thanks for writing these! Super useful