- AI Fire
- Posts
- 🎭 Deepfakes Are 10x More Out of Control
In partnership with
Plus: The Biggest AI App Store
Read time: 5 minutes
OmniHuman-1 is here, and it’s straight-up wild. TikTok’s parent company just introduced an AI that can clone anyone into a video—from just one photo and an audio clip.
Yeah, let that sink in. It’s eerily real. And now the big question: Is the world actually ready for this?
What are on FIRE 🔥
PRESENTED BY CREATOPY
AI-powered on-brand ad generation
Resize ads across multiple sizes with one click
Effortlessly impress clients with on-brand visuals
Engage with professional looking ads and increase ROI
AI INSIGHTS
🤖 The Next AI Giants Might Not Come From the U.S.
Sure, the usual suspects (Silicon Valley, Beijing) are running the AI arms race. But Europe? It’s carving out its own space in healthcare, climate tech, and infrastructure AI.
Take France, for example.
750 AI startups
35,000 jobs created
2,000 researchers + 600 PhD students
The country is quickly becoming a powerhouse for AI talent—and U.S. investors are taking notes. Half of the money going into late-stage European AI startups (Series C+) is coming from the U.S.
The reason? Europe’s AI boom isn’t about building the next ChatGPT—it’s about building highly specialized AI that solves real problems in health, energy, and automation.
AI in Europe Is Looking Different—And That’s a Good Thing
Many European AI startups are not focused on general AI models but solving industry-specific problems.
✅ Healthcare AI → Drug discovery, medical imaging, treatment optimization
✅ Climate AI → Carbon tracking, energy efficiency, agtech
✅ AI Infrastructure → Data pipelines, inference optimization, automation
Take Owkin and Bioptimus—leaders in AI-powered drug discovery. Or Altrove, which is experimenting with AI-driven new materials.
These aren’t just cool projects. They’re funded, fast-growing, and solving trillion-dollar problems.
Why it matters: While the U.S. and China dominate general AI, Europe is carving out a niche in AI applications for healthcare, climate, and automation. With strong early-stage funding and growing U.S. investment. In the next few years, the next AI giant? It just might come from across the Atlantic.
IN PARTNERSHIP WITH HUBSPOT
HubSpot's Starter Bundle for Startups
Unlock the potential of your growing startup with HubSpot's Starter Customer Platform. For a discounted price of just $20/month, unlock access to the Starter edition of HubSpot’s six core products for marketing, sales, and customer service – powered by HubSpot’s Smart CRM – all for the price of one. Built for startups like yours, HubSpot Starter has all the essential tools you need to scale.
TODAY IN AI
AI HIGHLIGHTS
🚀 Hugging Face is building Open Deep Research, an open-source version of OpenAI’s Deep Research. In 24 hours, they used CodeAgent to boost GAIA benchmark scores to 55.15%.
🎭 Researchers from TikTok owner ByteDance introduced OmniHuman-1, which creates ultra-realistic deepfake videos from just one image and audio. It handles any style, edits existing footage, and was trained on 19,000 hours of video.
Our POV: OmniHuman-1 acts like a universal translator for images and motion, making photos move with sound. China leads AI video tech with Hunyuan, Kling, and MiniMax. Startups like Synthesia ($2.1B) & HeyGen ($60M) are making ultra-realistic avatars, but with deepfakes rising, AI is moving from looking unnatural to becoming a money-making machine.
📲 Hugging Face is Now the Biggest AI App Store! With 400,000+ AI apps and 2,000 new ones daily, Hugging Face gets 2.5M visits weekly! Now, you can search AI apps easily by category or AI. The AI future is decentralized
🗣️ Amazon is set to introduce a major AI upgrade for Alexa on Feb 26 after delays. The new Alexa will hold conversations, take actions, and remember preferences. It may cost $5–$10/month but will first launch for free to limited users.
🤖 Figure has ended its partnership with OpenAI and hinted at a major advancement in robot AI. A big reveal is coming within 30 days.
🔍 ChatGPT search is now available to everyone on http://chatgpt.com — no sign up required.
🎨 OpenAI rebranded with a new logo, typeface (OpenAI Sans), and colors to feel more human and organic.
DAILY AI FUNDRAISING
StackAdapt secured $235M funding for its AI-powered advertising platform. The Canadian company helps businesses automate and optimize their digital ad placements across various media channels.
AI SOURCES FROM AI FIRE
NEW EMPOWERED AI TOOLS
🌐 Chatbase is AI Agents for magical customer experiences
☎️ Sendbird is AI for delightful customer service
📊 ZylerAI is your AI agent for Google Analytics
📱 Replit brings AI-powered app development directly to your iPhone.
🚀 Omniflow guides complete product journeys with intelligent development automation.
AI QUICK HITS
📜 Oracle Launches AI Vector Search Pro Certification & Course (Link)
📱 Snap Introduced AI Text-to-Image Model for Mobile Devices (Link)
📝 Adobe Acrobat’s AI Chatbot Now Explains Contracts Clearly (Link)
🎙️ LiveKit’s New AI Voice Model Cuts Interruptions by 85% (Link)
🗺️ AI-Powered Abdomen Map Could Detect Cancer Earlier (Link)
AI CHART
Google’s Gemini 2.0 Just Leveled Up—And It’s Open for Developers
The Gemini 2.0 family is expanding fast, and Google just made its latest models widely available. The key updates?
✅ 2.0 Flash → The high-speed, high-efficiency model is now open to all developers in Google AI Studio & Vertex AI.
1M token context window (reads massive amounts of data).
Better multimodal reasoning (understands text, images, and more).
Coming soon: Image generation + text-to-speech.
✅ 2.0 Pro Experimental → Google’s best coding model yet, featuring 2M tokens of context and improved reasoning.
2 million token context window → Handles massive amounts of information.
Better factual accuracy, reasoning, and world knowledge.
Calls external tools like Google Search & code execution.
But here’s the reality—Gemini 2.0 Pro underperforms compared to O3, O1, and R1. Pre-training seems to have hit a wall, and all the gains are coming from scaling inference rather than true model improvement. Surprisingly, Gemini 2.0 Pro even falls behind the much smaller Gemini 2.0 Flash Thinking version.
And for coding? It’s still behind Sonnet 3.5.
That said, Flash 2.0 being generally available is a win. It’s shaping up to be a solid replacement for Haiku and GPT-4o mini—fast, efficient, and capable.
✅ 2.0 Flash-Lite → A low-cost AI model that outperforms its predecessor at the same speed & price.
Google is doubling down on affordability with 2.0 Flash-Lite—a budget-friendly model that’s faster and more accurate than 1.5 Flash.
💡 Example? It can generate captions for 40,000 images for less than $1.
📌 1M token context window → Handles big workloads without breaking the bank.
And here’s the kicker: Multimodal inputs (text, images, video, audio) are coming soon.
AI JOBS
We read your emails, comments, and poll replies daily
How would you rate today’s newsletter?Your feedback helps us create the best newsletter possible |
Hit reply and say Hello – we'd love to hear from you!
Like what you're reading? Forward it to friends, and they can sign up here.
Cheers,
The AI Fire Team
Reply