Need to Know News - April 3rd, 2025
In this week's Need to Know News edition:
🤖 Runway's New Gen-4 keeps AI characters consistent across scenes — creators are going wild...
🤖 How Amazon's new "Buy for Me" AI agent shops other websites — brands are rushing to partner up...
🤖 Elon's $33B X-AI merger — the real reason he's combining 600M users with his "Colossus" supercomputer...
And whole lot more!
Runway Unveils Gen-4: AI That Creates Consistent Characters Across Multiple Scenes
Runway's new Gen-4 AI model solves one of the biggest problems in AI-generated content—keeping characters and objects consistent across different scenes.
With just one reference image, Gen-4 maintains character identity. And across various lighting conditions and environments. No more characters that look different in every frame.
The technology also lets creators:
- View the same elements from multiple angles
- Place objects consistently in any setting
- Generate realistic physics in videos
- Create "GVFX" effects that blend with live action
Gen-4 works right out of the box—no fine-tuning or extra training needed.
Elon Musk Merges His AI Company with X in $33B Deal
Elon Musk's AI company xAI just bought his social media platform X for $33 billion in an all-stock deal. The merger values xAI at $80 billion. Musk says combining xAI's AI technology with X's 600 million users will create "smarter, more meaningful experiences."
The deal gives xAI's chatbot Grok (which competes with ChatGPT) a massive distribution platform. X users already get Grok with premium subscriptions.This move comes as xAI ramps up its computing power. Their "Colossus" supercomputer in Memphis is reportedly the world's largest. The merger also gives xAI access to X's real-time user data—valuable for training AI models.
Chinese AI Companies Scramble to Adapt After DeepSeek's Breakthrough
DeepSeek's powerful new AI model has forced China's top AI startups to completely change their business plans.
- 01.ai stopped building its own AI models and now sells solutions using DeepSeek's technology
- Zhipu is rushing toward an IPO despite losing $41 million while making only $5.7 million in sales
- Moonshot cut marketing for its Kimi chatbot to focus on improving its tech
- Baichuan abandoned its financial AI business to focus solely on healthcare
DeepSeek's success comes from using a "mixture of experts" approach—combining smaller specialized models instead of one giant model. This method requires less computing power but delivers better results. Chinese companies now face a tough choice: compete directly with DeepSeek or use their technology and find smaller niches.
Papa Johns Bets Big on Google AI to Boost Sales and Cut Costs
Papa Johns is teaming up with Google Cloud to use AI across its business. The pizza chain has created a special "PJX" team to lead five key AI projects.
They'll use AI to predict what customers want to order. Then, send personalized push notifications based on your preferences. The system will offer real-time personalized discounts based on your order history and location. AI-powered chatbots will handle customer questions, with human backup for complex issues.
The technology will also optimize delivery routes and automate restaurant tasks. Papa Johns expects this tech to increase order size and frequency while cutting customer service costs.
Tinder Launches AI Flirting Coach to Help Dating App Users Practice Their Game
Tinder just released "The Game Game," an AI-powered flirting simulator built with OpenAI's GPT-4o technology. Users can practice their pickup lines by talking to AI voices in funny dating scenarios like airport luggage mix-ups.
The AI makes the first move. Then users respond by voice in real-time and get feedback on their flirting skills. Each session lasts just three minutes, with a limit of five plays daily.
Tinder created this after finding 64% of young singles don't mind a little cringe if it leads to real connections. The company hopes this low-pressure practice will help combat dating app burnout, which affects 79% of Gen Z users.
AI Therapy Chatbot Achieves Comparable Results to ‘Gold-Standard Cognitive Therapy’
A new AI therapy chatbot called Therabot is showing it can deliver mental health care as effectively as human therapists. In a Dartmouth College study, people using the chatbot for 8 weeks saw significantly better improvements than the control group.
The results were striking:
- 65% better symptom scores for anxiety and eating disorders
- 47% better scores for depression
- Patients trusted the AI at levels similar to human therapists
Users spent about 6 hours with the chatbot—equal to 8 therapy sessions. Most weren't taking medication or seeing other therapists.
Anthropic Scientists Peek Inside AI's Brain to See How It "Thinks"
Anthropic researchers have created an "AI microscope" that reveals how large language models actually think—and it's nothing like humans.
When adding 36 and 59, AI doesn't use the step-by-step math we learned in school. Instead, it estimates the answer range and final digit separately, then combines these probabilities.
The team discovered AI thinks in universal concepts that work across languages. When Claude writes poetry, it plans ahead—thinking about rhyming with "rabbit" before finishing the previous line ending with "grab it."
This research helps scientists understand AI's reasoning patterns, which could lead to better safety controls. As researcher Joshua Batson notes, while experience with AI builds trust, he'd be "a heck of a lot more comfortable if we also understand what's going on inside."
Amazon Tests AI Shopping Agent That Buys Products From Other Websites
Amazon is rolling out "Buy for Me," a new AI agent that lets customers purchase products from other brands without leaving the Amazon app.
Here's how it works:
- When searching for branded items, customers see results labeled "Shop brand sites directly"
- Users can ask Amazon's AI to complete the purchase for them
- Order confirmations come from the brand, but tracking happens in the Amazon app
This agentic AI feature is currently in beta testing with select US customers and limited brand partners. Amazon's shopping director Oliver Messenger says it helps customers "quickly and easily find and buy products" not sold on Amazon while giving brands "increased exposure and seamless conversion."
Thanks for reading.
Until next time!
The AI Marketers
P.S. Help shape the future of this newsletter – take a short 2-minute survey so we can deliver even better AI marketing insights, prompts, and tools.
[Take Survey Here]
