In what would have probably had steam coming out of Sam Altman’s ears this week, a group called "Sora PR Puppets" leaked access to OpenAI's video generator Sora on the open-source platform Hugging Face.

They did it to protest what they see as OpenAI exploiting artists. The group created a project that allowed users to generate 10-second videos using Sora, which is still in early access mode. However, OpenAI has suspended this access temporarily. 

The group claims the Altman-led company is pressuring artists to promote Sora without fair pay and restricting their ability to share their work. OpenAI responded by saying that artists’ participation is voluntary, reported TechCrunch. 

Despite technical challenges and Sora’s late launch into the market, it faces stiff competition from other video generation tools like Kling AI and Luma’s DreamMachine, which we have spoken about at length in my previous newsletter.

AI backpack cameras: Coming soon

Odyssey, a startup founded by former self-driving pioneers, is developing AI to help creators generate cinematic virtual worlds. Using a 25-pound camera system (which sounds heavy) with multiple sensors, the company collects high-resolution 3D data from real-world environments like forests, beaches, and cities.

This data feeds into AI models that create detailed, dynamic virtual scenes. The look of the equipment kind of reminds us of Doctor Octopus from the Spiderman movie series, doesn’t it?

Despite current limitations in generative world-building, Odyssey raised $18 million to expand its data collection in California and eventually other locations. The company wants to enable more realistic virtual environments for films, games, and creative projects.

However, I’m skeptical about Odyssey’s approach—25 pounds is a hefty load. Anyone who’s hiked or trekked knows that even an extra pound can turn a steady climb into a frequent struggle, with breaks every minute instead of every five.

OpenAI files trademark for new model

OpenAI has applied to trademark its new AI model, "o1," as part of protecting its intellectual property. The trademark application, filed with the US Patent and Trademark Office (USPTO), follows a similar filing in Jamaica months before the model’s announcement. The USPTO is still reviewing the request.

The "o1" model is designed for advanced reasoning, allowing it to fact-check itself and handle complex tasks better than typical AI models. The said model is already embedded with ChatGPT, as OpenAI plans to expand it into a series of reasoning-focused models.

OpenAI has filed around 30 trademarks, including "ChatGPT" and "DALL-E," but failed to trademark "GPT" earlier this year because it was deemed too generic. 

The company has mostly avoided aggressive trademark enforcement, except for a legal battle over the name "Open AI," where a court recently sided with OpenAI against entrepreneur Guy Ravine.

CEOs of OpenAI and Anthropic predict AGI

OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei have both expressed excitement about the arrival of AGI, with Amodei predicting it could happen by 2026 or 2027, and Altman having a more optimistic estimate of next year.

Their comments come amid rapid advancements in AI capabilities and the booming global AI market, which is projected to grow from $184 billion in 2024 to $826 billion by 2030.

Are things about to get wild? Read about it more in my column.

Amazon invests an additional $4 billion in Anthropic

Things are definitely going to get wilder for Anthropic. Amazon is investing another $4 billion into the company led by Amodei, making it total investment to $8 billion.

This will strengthen Amazon’s role as a minority investor and primary cloud and training partner. Anthropic will train its AI models, like chatbot Claude, using Amazon’s Trainium and Inferentia chips.

AMD lays off 4% of workforce

Nvidia's competitor, AMD, is in the process of laying off 4% of its workforce, which is about 1,000 employees. The layoffs follow mixed quarter-3 results, with strong revenue but a 69% drop in gaming division sales and underwhelming AI chip progress compared to Nvidia. 

Because Nvidia is currently dominating the market, AMD is planning on releasing competitive AI chips in 2025, but faces challenges. AMD's stock was also 4% down this year, reported WCCFTech.

On the other hand, Nvidia is showing no signs of stopping. The company has reported a 94% revenue increase and a 106% profit jump for the October quarter, with $35.08 billion in revenue and $19.04 billion in net income.

It projected a 70% revenue growth for the next quarter, driven by strong demand for its new AI chip, Blackwell. However, supply constraints on the chip caused its shares to dip 3% in after-hours trading, reported the New York Times.

OpenAI releases blueprint for US AI infrastructure

OpenAI has released a blueprint for US AI infrastructure, emphasizing government support for energy systems, data centers, and policies to boost AI development. 

The plan is to compete with China. The blueprint proposes AI-focused economic zones, a national transmission initiative, and a North American AI alliance.

OpenAI compared AI technology to electricity, and said that AI will give similar access and benefits to users.

“AI presents an unmissable opportunity to reindustrialize the US and through that, generate the kind of broad-based economic growth that will revitalize the American Dream,” says the blueprint.

The plan also highlights the Navy’s nuclear resources and calls for workforce training in AI-related roles. This aligns with ongoing government efforts to support advanced technology but also pushes policies favoring infrastructure benefiting companies like OpenAI.

Did a friend forward this e-mail to you?

IE₊ SUPPORT INTERESTING ENGINEERING
_{Invest In Science And Engineering}

Enjoy exclusive access to the forefront of AI content, highlighting trends and news that shape the future. Join a community passionate about AI, delve into the latest AI breakthroughs, and be informed with our AI-focused weekly premium newsletters. With IE+, AI reporting goes beyond the ordinary - and it is Ad-Free.

NEWS

• China launches world’s first ‘self-driving’ satellites for military surveillance
  The two new satellites are part of a larger project to create a network of 28 satellites for comprehensive Earth observation.

• White House power? Elon Musk’s legal strike targets OpenAI’s shift to profit
  Musk’s legal team accuses OpenAI and Microsoft of violating federal racketeering and antitrust laws.

• Fighter jet-inspired washing machine offers 15-minute AI-enhanced cleanse for humans
  The human washing machine aligns with broader trends toward health technology, smart devices, and the increasing integration of AI into our daily routines.

• Watch: Tiny robot ‘kidnaps’ 12 big Chinese bots from a Shanghai showroom, shocks world
  The video of the bizarre incident got quite a lot of attention online after being posted on Douyin.

MUST-READ

China’s cat-like robot dog perfects mid-air jumps to conquer asteroid surfaces

Chinese researchers have created an AI-powered robot dog capable of leaping across the rugged, low-gravity surfaces of asteroids.

Inspired by a cat’s ability to twist and land on its paws, researchers from the Harbin Institute of Technology developed a robot using reinforcement learning.

AI PICTURE OF THE MONTH

This month’s AI image of the month is more like ‘AI GIF of the month’. 

Runway released a new AI image generator last week, which is one of the most realistic models one has seen to date. It can create cartoon, abstract and alien-y images, which isn’t exactly novel in the world of AI image generators. But what caught my attention when I was browsing through its capabilities was its uncanny precision in perfectly channeling the vibes of 1960s and 70s art.

Here is a snippet of the images created by Frames.

AI TOOL OF THE MONTH

We have to talk about Google’s NotebookLM. It’s what everyone’s been talking about, with podcasters indulging in a hushed chorus of criticism.

To people who write blogs (or want to start writing one), hear me out. There is an AI tool that can help you streamline your process or get things started.

While I don’t exactly condone the help of generative AI to write original content, its advantage in making text concise and helping things move faster cannot be undermined.

The tools allows one to draft SEO-optimized blogs within minutes from YouTube videos, podcasts, web links, or uploaded files (audio, video, or documents).

Here is a tutorial on how to get yourself a blog:

1. Logging in

• Go to the Blogify website and sign up.

• You can try the product for free for 3 days without any charge. However, you’ll need to provide your card details, which might feel a bit inconvenient. Still, it's a chance to test the product before committing.

2. Navigating the Dashboard

• Once you log in to your account, you will land on a page called the ‘Dashboard.’

• You’ll be able to see here how many blogs you have created, the credits that are available to you, and your earnings (which you earn by way of Blogify inserting their affiliate links into your blog. They make money, and you make money.)

• Down below, under ‘Blogs Activity,’ ’you can see how your blogs are performing.

3. Creating a blog using a video/audio URL

• Click on ‘create a blog’ and add the URL for the source video and audio you want to use to create the blog

• Next, you have blog settings like the blog size (word count), tone, language of the source material, and the language you want your blog to be in.

• You can also choose whether you want to create the blog in the first, second, or third person. Pretty cool, no?

• Next, you choose a blog cover image and some content images for within the blog post.

4. Publishing

• There is also the option of automatically publishing your blog on to your WordPress blog, LinkedIn, Twitter, Facebook, etc.

• Click on ‘Generate Blog’ and voila!

However, Blogify offers limited customization options for content style and format, and requires a stable internet connection, limiting offline use. Additionally, the subscription model may be costly for users who need the service only occasionally.