- Not So Artificial
- Posts
- š§ Meta AI Gets Too Personal
š§ Meta AI Gets Too Personal
Plus: Check these AI Tools
Hey there, š
Welcome to another edition of Not So Artificial, where we break down the latest in AI, tech, and everything. Buckle upātodayās lineup is packed with innovation, competition, and a splash of controversy.
š DeepSeek's Janus-Pro: Redefining AI Image Models
Chinese startup DeepSeek has launched Janus-Pro, an open-source AI model thatās shaking up the image generation space. Think DALL-E 3 and Stable Diffusionābut sharper, smarter, and ready to rumble.
The Details:
Janus-Pro generates high-quality images from text descriptions, with 1B and 7B parameter models to choose from.
The model scored higher than competitors in industry benchmarks like GenEval and DPG-Bench.
Released under an MIT license, developers can freely use or modify it for commercial projects.
Why It Matters: The release comes hot on the heels of DeepSeek's viral R1 model, which disrupted markets by delivering top-tier performance at a fraction of typical costs. As the U.S. grapples with the implications, DeepSeek is proving that the race for AI dominance is far from one-sided.
Hire an AI BDR to Automate Your LinkedIn Outreach
Sales reps are wasting time on manual LinkedIn outreach. Our AI BDR Ava fully automates personalized LinkedIn outreach using your teamās profilesāgetting you leads on autopilot.
She operates within the Artisan platform, which consolidates every tool you need for outbound:
300M+ High-Quality B2B Prospects
Automated Lead Enrichment With 10+ Data Sources Included
Full Email Deliverability Management
Personalization Waterfall using LinkedIn, Twitter, Web Scraping & More
š§ Meta AI Learns About You
Meta just rolled out personalization upgrades for its assistant, making it smarter (and maybe a little creepier) than ever.
The Details:
The AI assistant now remembers chat details like dietary preferences and interests to offer tailored responses.
It can leverage Facebook locations, Instagram viewing history, and profile data to deliver personalized recommendations.
These features are live in the U.S. and Canadaābut thereās no opt-out for users (though specific memories can be deleted).
Why It Matters: While personalization could boost user experience, Meta's approach raises privacy concerns. With a history of shaky trust in handling user data, the lack of an opt-out feels like a risky move in a hyper-sensitive privacy landscape.
š Qwenās AI Takes Control
Alibaba's Qwen team has unveiled Qwen2.5-VL, a vision-language AI model that doesn't just understandāit acts. From controlling apps to parsing complex documents, Qwen might just be your next digital operator.
The Details:
The 72B parameter flagship outperforms GPT-4o and Claude 3.5 Sonnet in document parsing and video analysis.
It can extract key moments from hour-long videos and handle tricky documents like invoices.
New agentic features let it control smartphones and computers for tasks like airfare booking and image editing.
Why It Matters: As the AI arms race heats up, Qwenās latest release highlights how quickly Chinaās tech is narrowing the gap with Western giants like OpenAI.
šļø Zuckerbergās $65B AI Gambit
Meta CEO Mark Zuckerberg is doubling down on AI with plans to invest a staggering $60-65B in infrastructure by 2025.
The Details:
Meta is building 1GW of compute power, enough to cover a chunk of Manhattan.
The company is on track to deploy 1.3M GPUs by year-end.
Spending is up 70% from 2024 as Meta pushes for Llama 4 to become the next industry standard.
Why It Matters: With U.S. players like Meta and OpenAI pouring billions into datacenters, the AI infrastructure race is scaling up. But DeepSeekās disruptive breakthroughs are proof that money alone isnāt the only factor in winning the AI game.
š„ Metaās Panic Mode: Reverse-Engineering DeepSeekās Magic
DeepSeekās R1 is giving Meta nightmares. Zuck assembled four war rooms to crack how the Chinese lab built a GPT-4 rival for pennies. Priorities:
Decode R1ās cost-slashing sorcery.
Steal its data recipe.
Remodel Llama to mimic its mojo.
Plot twist: This comes days after Meta pledged $65B to outspend⦠everyone.
ā” Other News You Should Know
𧬠LinkedIn co-founder Reid Hoffman raised $24.6M for Manas AI, a cancer-fighting drug discovery platform.
š± DeepSeekās R1 app topped Appleās App Store with 2.6M downloads, but cyberattacks forced it to pause new sign-ups outside China.
š¤ xAIās Grok-3 briefly went live, teasing advanced reasoning abilities ahead of its full release this week.
š„ Pika Labs launched v2.1 of its AI video generator, adding realistic physics and motion controls.
š Appleās iOS 18.3 now enables Apple Intelligence by default, but AI summaries remain disabled.
š«š· The French governmentās AI chatbot Lucie was suspended days after launch due to widespread factual errors.
š¬ OpenAI CEO Sam Altman says AI will require societal restructuring, with changes to the social contract and governance models.
š» Mastery tools of the day
WePost - Generate on-brand marketing with ease
UI-TARS - Control your computer using natural languag
Citations - Allow Claude to ground answers in source documents and provide references
Perplexity Assistant - Android agentic assistant capable of controlling phone apps and performing complex tasks
š”What else are we reading and seeing?
DeepSeek FAQ
Why AI reminds me of cloud computing
The secret sauce of Chinese social media apps
The Past, Present, and Future of UI at GitHub
Oracle and Microsoft are reportedly in talks to take over TikTok
Learn how to make AI work for you
AI wonāt take your job, but a person using AI might. Thatās why 1,000,000+ professionals read The Rundown AI ā the free newsletter that keeps you updated on the latest AI news and teaches you how to use it in just 5 minutes a day.
Thanks for Reading!
Thatās all for today! Feeling inspired, curious, or just a tad overwhelmed by all the AI news? Us too. Hit reply with your thoughts or questionsāwe love hearing from you.
Catch you tomorrow!
ā Not So Artificial š§