AI news stories I thought were cool + American Primeval

Dr. Ryan Ries here. Today I want to dive into four AI developments that caught my attention this week. 

Let's take a look at how AI is reshaping everything from weather forecasting to personal image generation. Read through until the end for a random fun (and kind of sad) fact.

Quick announcement before I start talking about the articles. I’m hosting another Gen AI Ask Me Anything session on February 25th. Jonathan LaCour, Mission CTO, and I will be answering all your questions about AI. Register here and join us!

The New AI Weather Wars

Google DeepMind just dropped something incredible - GenCast, an AI weather forecasting system that's outperforming traditional models across the board. 

What makes this particularly interesting is how it handles uncertainty: instead of making a single prediction, it generates multiple possible scenarios, helping meteorologists better understand the range of possible outcomes.

This tech is huge for disaster planning, saving lives, and protecting communities. 

Take the recent California wildfires, for example. 

With systems like GenCast's ability to predict extreme weather patterns, fire departments could better anticipate dangerous wind conditions that often fuel these fires. The model's 15-day forecast capability could give emergency services crucial extra days to prepare and position resources. Even more promising is GenCast's specific strength in predicting wind patterns - a critical factor in understanding how wildfires might spread.

Another New Model to Compete with the Rest

I feel like every week I’m seeing something about another new model. Companies are really cranking these out.

Alibaba chose the Lunar New Year holiday to release their Qwen 2.5-Max model, claiming it outperforms both DeepSeek and GPT-4. But, plot twist, while Chinese companies like DeepSeek made headlines with claims of extreme efficiency (like training their model for just under $6M), the reality is more complex.

A new report from SemiAnalysis reveals that DeepSeek actually has about 50,000 Nvidia Hopper GPUs and has spent around $1.6 billion on infrastructure buildouts. 

Chinese companies are being pushed to innovate differently due to resource constraints like infrastructure limitations and data restrictions. 

DeepSeek, for example, has focused heavily on algorithmic improvements and runs its own data centers rather than relying on cloud providers. 

They've also taken an interesting approach to talent, exclusively hiring from within mainland China and offering competitive compensation - some AI researchers there reportedly earn over $1.3 million annually.

These more efficient models, even if they require more resources than initially claimed, are still putting pricing pressure on Western companies and challenging the assumption that you need unlimited resources to compete in AI.

You can read more about my thoughts on DeepSeek in last week’s Matrix.

AI and Copyright: The Human Element

The U.S. Copyright Office just dropped Part 2 of their AI report, and it's a big one. 

Their stance on AI-generated works is quite nuanced. 

They say copyright protection is available only in 2 specific scenarios:

1. When a human-authored work is perceptible in the AI output
2. When humans make creative arrangements or modifications to the AI-generated content

Just typing prompts into ChatGPT won't cut it, no matter how detailed or creative those prompts may be. You need to demonstrate real creative contribution. 

This clarity from the Copyright Office is pretty significant for businesses. 

For example, if your marketing team is using AI to generate content, they'll need to ensure there's substantial human creative input in arranging, modifying, or incorporating that AI content into larger human-created works to maintain copyright protection. 

It’s still not totally clear to me how they’ll deem things as being “human-edited enough.” 

But, the Copyright Office mentioned they’ll be updating registration guidance and practice manuals to include more specific details on evaluating works. So maybe we’ll know soon enough!

Random Fun Fact

The average social media user scrolls through about 300 feet of content every single day - that's roughly the same height as the Statue of Liberty! 

With 5 billion social media users worldwide, that means collectively we're scrolling the equivalent distance of going to the moon and back multiple times each day. 

So, if you’re working on doom scrolling less in 2025, visualizing climbing Lady Liberty as you scroll through one post at a time may help you put the phone down!

Until next time,
Ryan Ries

NNow, time for this week’s image and the prompt I used to create it. I just watched the first episode of American Primeval so I figured it might be fun to see how muppets would’ve looked in the Wild Wild West. Let me know if you’re also watching this show (and thinking about how scary the Wild West must’ve been)!

"Create an image of muppets in American Primeval."

Sign up for Ryan's weekly newsletter to get early access to his content every Wednesday.