🏁Ndea Joins AGI Race

AI researcher François Chollet partners with Zapier co-founder to develop AGI

Key Points 

• François Chollet and Mike Knoop's new startup, Ndea, aims to develop AGI using program synthesis and innovative AI techniques.

• Ndea will focus on creating AI that learns efficiently like humans and continuously improves without bottlenecks.

☕News - François Chollet, a leading AI researcher, is launching a new startup called Ndea. The company is focused on developing frontier AI systems with the goal of creating artificial general intelligence (AGI). Ndea will serve as both an AI research lab and a science hub for innovation.  

🧐What's their plan? Chollet outlined the company’s vision in a series of posts on X. He shared that the team is taking a different approach to build AI capable of true invention, adaptation, and innovation. They believe they have a genuine chance to create AI that learns as efficiently as humans and can keep improving over time without any significant bottlenecks.  

🤔How so? Ndea plans to use program synthesis, a technique that allows AI to generalize problems it has never encountered before using only a few examples. 

Chollet believes that program synthesis, combined with other technical methods, can help overcome some of the most difficult challenges in AI research. Although program synthesis is typically computationally intense, he is confident the team can find ways to overcome this limitation and accelerate progress toward AGI.  

🥸See also - Chollet co-founded Ndea with Mike Knoop, co-founder of Zapier and head of AI. Knoop mentioned that their initial focus will be on deep learning-guided program synthesis to develop AGI that can invent, adapt, and innovate.

Read the full story here. 

Chinese AI firm MiniMax launches new models that challenge U.S. giants

Key Points 

• MiniMax has launched three new AI models: MiniMax-Text-01, MiniMax-VL-01, and T2A-01-HD, targeting text, multimodal, and audio tasks. 

• MiniMax-Text-01 is claimed to outperform Google’s Gemini 2.0 Flash, while MiniMax-VL-01 is said to outperform Anthropic’s Claude 3.5 Sonnet.

🚀News - Chinese firms are rapidly releasing AI models that challenge the capabilities of systems developed by OpenAI and other U.S.-based companies. 

This week, MiniMax, a startup backed by Alibaba and Tencent, unveiled three new models: MiniMax-Text-01 (a text-only model), MiniMax-VL-01 (which understands both images and text), and T2A-01-HD (an audio generator for speech). 

✨How they compare - MiniMax-Text-01, with its 456 billion parameters, is said to outperform models like Google’s Gemini 2.0 Flash in complex text-based tasks, such as answering math problems and factual questions. Its larger context window—about 31 times bigger than that of GPT-4 and Llama 3.1—gives it an edge in processing larger chunks of information.

MiniMax-VL-01, which excels in multimodal tasks, outperforms Anthropic’s Claude 3.5 Sonnet, particularly in tasks that require understanding both images and text, such as answering questions about graphs and diagrams. 

The T2A-01-HD model generates speech with customizable tone and cadence and can clone a voice from just 10 seconds of audio, supporting around 17 languages, including English and Chinese.

💻How to access them - While T2A-01-HD is only available through MiniMax’s API and the Hailuo AI platform, the other models are downloadable from GitHub and the AI development platform Hugging Face.

Read the full story here. 

In last Wednesday’s Workflow Wednesday, we shared AI workflows that tackled the questions our readers had been asking. Have AI workflow questions you need answered? Join here.

Q&A: Sales Hacks Edition

Q: What’s the fastest way to create a sales deck that actually looks good?
A: Gamma.app, hands down. In less time than it takes to brew your morning coffee, Gamma transforms your sales strategy into a professional presentation. Bye-bye, endless tweaking. Hello, polished perfection. Read our guide to creating a powerful sales deck with Gamma here.

Q: I hate making videos—but I need one for a campaign. Help?
A: Introducing Pictory, your campaign lifesaver. It takes your script and turns it into a professional sales video. No editing headaches, no wasted hours. Read our guide to creating high-converting videos with Pictory here.

Q: Writing product descriptions makes me fall asleep. Any tips?
A: You need to be using Writesonic. With just a few product details, it’ll whip up engaging, sales-optimized descriptions in seconds. Read our guide to mastering product descriptions with Writesonic here.

Want more time-saving AI workflows delivered straight to your inbox? Join Workflow Wednesday, our premium newsletter that helps you work smarter, not harder.