GPT-4’s hot new competitors

Plus: Humanoid robots are coming

Happy Friday! This morning, OpenAI made headlines…thanks to a sticky situation with its ex-cofounder, Elon Musk. Musk is suing OpenAI and its CEO Sam Altman for “abandoning the startup's original mission” to develop open-source AI for the benefit of humanity—not for profit. This comes after an already rough week of legal action: Just yesterday, OpenAI was sued by three more news organizations, while the company is still battling the New York Times in court.

At this point, I wouldn’t be surprised to see Sam Altman studying for the Bar in his free time. 🙃 

Mistral and Phind’s New GPT-4 Alternatives

Shutterstock

OpenAI’s GPT-4 has been leading the charge of large multimodal models—but this week saw two new model releases that are upping the competition.

First up: Mistral. On Monday, the French AI startup unveiled its new flagship model, Mistral Large. Designed to closely compete with OpenAI’s GPT-4 model, Mistral Large stands out for several reasons:

  • It’s natively multilingual (i.e. trained in English, Spanish, French, German, and Italian).

  • It has a smaller context window than GPT-4 Turbo (32k vs. GPT-4’s 128k), but its API usage pricing is 20% cheaper.

  • On standard reasoning and knowledge benchmarks, Mistral Large is second only to GPT-4—beating Gemini, LLaMA, and Claude.

Unlike Mistral’s previous LLMs, Mistral Large isn’t open-source—it’s available commercially through Mistral’s API or Microsoft Azure (thanks to a new partnership with Microsoft). 

Next up: Phind. Phind just launched an impressive new AI model for developers and coders: Phind-70B. Phind-70B beats GPT-4 Turbo on coding benchmarks…while running 4x faster.

Why it matters: The performance gap between OpenAI’s models and competitor LLMs is shrinking by the day. If this week was any indicator, there’s plenty of potential for young AI startups to seriously challenge the industry leader. 

Google’s Genie Dreams Up Video Games

Google DeepMind

What if you could whisper a gaming wish to an AI genie and poof—out pops a fully playable video game? Thanks to Google DeepMind, this might be a reality. 

The details: This week, Google DeepMind introduced Genie, a generative AI model that creates interactive, 2D platformer games (think Super Mario or Contra) from text prompts or image prompts—even hand-drawn sketches.

But the coolest thing about Genie? It was trained on 200,000 hours of internet video game footage—without human supervision. This means that Genie learned to understand and create interactive worlds without explicit instructions, identifying patterns and rules of game design on its own.

Google hasn't announced when—or even if—Genie will transition from a project to a commercial product. For now, you can check out some of Genie’s cool game generations here

Why it matters: Genie highlights the untapped potential of unsupervised AI training. With the right algorithms, AI can understand and recreate the complex physics that governs the real world—which could be the key to training autonomous AI agents of the future.

Your All-in-One Branding Solution

Ready to bring your brand vision to life, but struggling with design? Zoviz has got you covered.

Zoviz uses powerful AI to craft a cohesive brand identity from scratch. Simply enter your business name, and Zoviz instantly outputs everything you need:

  • A unique, high-res logo in 35 logo types

  • A full brand book with insights on font and color usage

  • Email signatures and letterheads

  • Favicons and social media covers

Whether you’re a freelancer, entrepreneur, or established business, Zoviz helps bring your brand vision to life with innovative AI.

Exclusive Offer: Use code zoviz-logo-cxi8f2 for a 20% discount on your premium logo package—only for the Future Tools audience!

Figure AI Is Bringing Humanoid Robots to Market

Get ready: Humanoid robots are coming. Robotics startup Figure AI officially has the cash and big-name support to bring robots to the public.

The details: Yesterday, Figure AI announced a casual $675 million funding round and partnership with OpenAI. The funds come from a star-studded list of investors—including Jeff Bezos, Nvidia, and Intel.

So what’s the hype all about? Figure AI developed a general-purpose robot—Figure 01—that looks and moves like a human. The company plans to use these robots to perform dangerous jobs and address labor shortages.

Why does it matter? AI hardware has considerably lagged behind AI software advancements. But with this influx of funding and powerful partnerships, Figure AI has the potential to shape the future of humanoid robotics.

AI Video Generation Is Just Getting Started

Pika Labs

Think you’ve seen the peak of AI video? Think again. This week saw even more mind-blowing features elevating AI video.

Spoken dialogue: Pika Labs, the AI startup celebrated for its cutting-edge text-to-video technology, dropped a groundbreaking new feature called Lip Sync. 

  • Lip Sync allows the AI-generated characters in Pika videos to “speak” with realistic lip movements synchronized to AI audio from ElevenLabs.

  • AI analyzes voice recordings and aligns the characters' mouth movements with the spoken words, making the videos more lifelike and engaging.

All-in-one platforms: Later this week, two AI filmmaking platforms came to the scene:

  • LTX Studio (by Lightricks, the company behind Facetune) and Morph Studio (in partnership with Stability AI) both announced platforms that can turn simple video generation into complete film projects.

  • These platforms generate characters, scenes, storyboards, and entire movies from text, images, or existing videos.

  • Get on the waitlist to be the first to try Morph Studio and LTX Studio when they launch! (Quick disclaimer: I’m partnering with the team over at Lightricks to help launch LTX Studio).

Why does it matter? 2024 is shaping up to be the year of AI video generation. With Sora’s major improvements in video quality, Pika’s innovative audio capabilities, and Lightricks’ and Morph Studio’s all-in-one platforms, AI will soon give professional film studios a run for their money.

  • OpenAI argues that the New York Times “hacked” ChatGPT in its copyright lawsuit. 

  • WordPress and Tumblr are planning to sell user content to OpenAI. 

  • Alibaba released a new AI model for audio-to-video face portraits. 

  • Scientists at Princeton University found a way to use AI to solve a major roadblock to generating fusion energy.

  • Huawei’s new AI-enhanced smartphone lets you control a car by just staring at it.

More important AI news: Dive deeper into this week’s hottest AI news stories (because yes, there are even more) in my latest YouTube video:

Is coding dead? Check out my thoughts on what AI means for the future of coding in this latest video:

And there you have it! I’ll be back in your inbox next Wednesday with a fresh roundup of new AI tools for you to try out. Have a great weekend!

—Matt (FutureTools.io)

P.S. This newsletter is 100% written by a human. Okay, maybe 96%.