Good morning. Yesterday, AI image generator Midjourney made its website freely accessible to everyone. At the same time, Ideogram launched 2.0, its “most advanced text-to-image model” to date, simultaneously making the service available on the IOS app store.

Yeah, this sounds like a good idea.

— Ian Krietzberg, Editor-in-Chief, The Deep View

In today’s newsletter:

AI for Good: Chasing storms and extreme weather

In the midst of a hurricane season NOAA promised would “rank among the busiest on record,” Nvidia has published new research that positions artificial intelligence as a valuable storm chaser. 

The details: Nvidia this week unveiled a new generative AI model called Storm Cast, which can emulate “high-fidelity atmospheric dynamics.” 

Though StormCast is only one step toward more advanced weather forecasting, researchers are already excited about its potential. Imme Ebert-Uphoff, machine learning lead at Colorado State University’s Cooperative Institute for Research in the Atmosphere, said that the development of high-resolution weather models requires AI models to resolve convection, something he called a “huge challenge.” 

But StormCast represents a “significant step toward the development of future AI models for high-resolution weather prediction.”

Accurate, Explainable, and Relevant GenAI Results

Vector search alone doesn’t get precise GenAI results. Combine knowledge graphs and RAG into GraphRAG for accuracy, explainability, and relevance.

Read this blog for a walkthrough of how to:

Washington Post publishes story based on AI-led investigation 

On Sunday, the Washington Post published an article investigating TV advertising related to U.S. immigration and the border. The investigation involved the analysis of 745 campaign ads that ran between January and June of 2024. 

And though the story was put together by a long list of Post journalists and editors, the analysis was built on the back of a large language model. 

The details: The Post developed an in-house tool called Haystacker that extracts stills from videos, then labels the objects present in each still. Post reporters then analyzed the resulting text and visual information. 

My view: This kind of video analysis is definitely a pain point in journalism. My only concern with applying LLMs to it is in developing an over-reliance on these tools; if I was using Haystacker, I would constantly be wondering what the model isn’t showing me. It creates a degree of separation that, as a journalist, I’d rather not have. 

But utilized as the backbone of an analysis that’s later confirmed by journalists certainly seems like the best option. 

Save Time & Scale Video Creation with AI

Say goodbye to video production bottlenecks. PlayPlay's AI Video Assistant removes the hassle of video editing — helping teams create professional-looking videos in seconds.

Simply describe your video needs in a sentence and watch the Assistant create your video using the perfect template, text, images, and music.

Trusted by 3,000+ brands like Dell, CVS Health and L'Oréal.

Create engaging videos now — start your 14-day free trial.

If you want to get in front of an audience of 200,000+ developers, business leaders and tech enthusiasts, get in touch with us here.

ElevenLabs launches program to help ALS victims reclaim their voice

ElevenLabs, the audio-genAI company, this week launched an Impact Program “to help one million people reclaim their voice.” 

The details: To start, the firm is partnering with nonprofits to offer free pro licenses to patients with ALS and MND, which would allow them to clone and preserve their voice with ElevenLabs’ AI tech. 

Creative AI expert Alex Champandard said that, while the offer itself is good, the context of offering a free “license” is a red flag. The licenses — which lock people into the platform — could end at any time, he said, which would leave the people they’re trying to help in a worse state than before. 

The context: In terms of the rest of their business, I’ve written about ElevenLabs a few times, first, when they were identified as the platform that was used to generate that fake robocall of President Joe Biden in January, and more recently, in the context of digital necromancy. 

Exclusive interview: CEO of AI engineering startup Monumo talks LEMs

We’ve talked often about that one area where artificial intelligence far surpasses human capabilities: deriving insights from truly enormous piles of data. This is prevalent in everything from medical applications of AI, in which AI is used to find new drug candidates, to applications of AI in environmental conservation, where AI is used to track and analyze everything from pollution levels to the movements of endangered species. 

British deep tech startup Monumo is applying the tech — specifically 3D generative AI (large engineering models, or LEMs) — to engineering, with the simple idea of revolutionizing and optimizing the creation of electric motors to create cheaper, more sustainable and more capable designs.  

Monumo, which was founded in 2021, came out of stealth with £10.5 million in funding in February of 2024.

I sat down with Dominic Vergine, Monumo’s founder and CEO. 

The details: The reason Monumo decided to start with the electric motor is two-fold: first, it is at the heart of clean energy and so plays a big role in decarbonization efforts, and second, it’s a “nicely constrained multi-physics problem. It’s neither too big nor too small.”

“Super” human isn’t about intelligence: Vergine said that the system “is already finding things that are arguably beyond human in terms of some of the shapes and some of the waveforms that are being produced.”

The reality of this, though, isn’t about intelligence, according to Vergine. It’s about capability. In both the corporate and academic worlds, an engineer has to have some evidence that a “route you’re going to take is going to yield results.” 

But the AI-driven simulation approach allows engineers to “go down avenues that no one in, in either a business setting or in an academic setting, would ever normally explore, because it takes too long and it costs too much.”

“You have to start somewhere, but you can grow this out into being, ultimately, a system that can design the next aircraft, or the next car, or the next wind turbine in its entirety,” Vergine said. “You explore every conceivable route. Some of them are dead ends. Some of them are not, but ultimately, you come up with an apex design,” he said, referring to a design that uses the fewest materials and delivers the maximum result for its given task. 

“We're very, very far off apex designs of any kind of motor or airplane or wind turbine at the moment,” Vergine added. “I think we will get there in the next 50 years or so.”

A poll before you go

Thanks for reading today’s edition of The Deep View!

We’ll see you in the next one.

Here’s your view on electing an AI to public office:

More than half of you — I’m with you guys on this — said “are you serious, absolutely not.” 10% said you would and 18% said you’re open-minded; it depends who the competition is.

Something else:

Absolutely not: