How well does AI know you?

Small AI models, big changes

Welcome back! Pokémon Go got you outside. Now Niantic wants to map the whole world. Using data from its games, Niantic is building large-scale geospatial models—think of it as creating a virtual twin of our planet. This tech could completely change how we interact with AR, but here’s the catch: All that data raises big questions about privacy. 

Should we be excited—or worried?

Adobe puts AI power in your pocket

Via Notebookcheck

Adobe is putting the power of AI in your hands—literally. With its new SlimLM, you can run advanced language models directly on your smartphone, no cloud required. Tap into AI to draft emails, summarize reports, or even analyze data, all while keeping your information private and working offline.

What it means for you: For businesses and consumers, this is a huge shift. It’s smarter, faster, and respects your privacy in ways cloud-based tools can’t.

  • For businesses: Teams can rely on AI tools on the go, even in areas with poor connectivity, while keeping sensitive data secure.

  • For consumers: Mobile apps powered by SlimLM will be faster, more responsive, and won’t send personal information to the cloud.

Mobile AI is trending: Google’s Gemini Nano and Meta’s LLaMA 3.2 are also bringing advanced AI models to devices, marking a major shift from cloud-reliant systems to more localized solutions.

What’s driving this shift? Smaller models consume less energy while still delivering high performance. As smartphones become more powerful, they’re becoming the new frontier for AI innovation.

Is smaller the future? SlimLM shows that compact AI models aren’t just practical—they’re powerful. They use fewer resources, work offline, and open up new opportunities in areas with limited connectivity. While LLMs dominate headlines, smaller, more agile models like SlimLM might quietly redefine the future of AI.

​​Could analog chips make AI greener?

Via Medium

Sagence, a bold new player in AI hardware, is betting on analog chips to revolutionize processing. These chips promise faster speeds and dramatically lower energy use than today’s digital chips. If they succeed, Sagence could reshape the way AI systems are powered.

Why analog chips matter: Unlike digital chips, which process data in binary (1s and 0s), analog chips work with continuous signals. This makes them faster and more energy efficient for AI tasks like image recognition or natural language processing. They’re ideal for applications where speed and efficiency matter more than exact precision.

The best of both worlds: Sagence doesn’t see analog chips as a replacement for digital ones. Instead, the company envisions a hybrid future where analog chips handle rapid, energy-saving tasks while digital chips take care of precision-heavy computations. It’s a complementary approach that could significantly reduce the environmental impact of AI.

The big picture: With the energy needed to run large-scale AI models, analog chips offer a glimpse at a more sustainable future. They’re faster, cheaper, and greener—a rare combination in AI hardware.

​​One Prompt, One Complete Video

Invideo AI V3 has landed, bringing true end-to-end video generation to your creative toolkit.

Unlike other AI tools that only generate clips, Invideo AI V3 delivers complete, publication-ready videos from simple text prompts.

With features like voice cloning, 50+ language translations, and text-based command editing, you can create professional videos without any technical expertise. Simply type commands like "add upbeat music" or "translate to French," and watch your vision come to life.

Start with free access, or unlock full generative capabilities at $96/month.

Google’s chatbot just got personal

Via YouTube

Google’s Gemini chatbot just leveled up: With its new memory feature, Gemini can recall your preferences and interests.

What it means for you: Imagine asking Gemini for restaurant recommendations, and it remembers you love spicy food. Instead of generic suggestions, it filters the list to match your tastes. AI is actively learning and adapting to you, making it feel more like a real assistant.

The privacy question: Google promises that personal data saved in Gemini’s memory won’t be used to train its models. Users can also view, delete, or manage their saved preferences anytime. It’s a step toward transparency, but privacy concerns still loom large.

A glimpse into the future: Memory is not just a feature but a signal of where AI is headed. As chatbots become more personalized, they’ll likely integrate deeper into our daily lives. With personalization comes responsibility, and how companies like Google handle this will shape the future of AI trust.

  • The HP OmniBook Ultra is a new AI laptop model that helps you cut through tedious work with features like live presentation style feedback, AI assisted work summaries, and more.*

  • OpenAI teams up with educators on free AI course.

  • Want more AI in your inbox? Mindstream covers daily news, tool recs, and AI training. Become even more of an AI expert by signing up here (and get free access to their Ultimate ChatGPT at Work bundle).

  • AI startup launches Runner H for agentic applications.

  • Messenger adds AI backgrounds and HD video calling.

  • Quantum computing errors tackled by AlphaQubit’s AI decoder.

  • DeepSeek’s R1-Lite-Preview boosts AI reasoning power.

  • Alexa will soon focus on smarter APIs over more AI.

*This is sponsored advertising content.

Will 2025 be the year AI takes over? OpenAI CEO Sam Altman predicts we may reach AGI in the next few years…watch along as I break down the latest news in AI. This is one you don’t want to miss.

2024 AI wrapped: Im’m ranking the most popular AI tools of the year ➡️ 

That’s a wrap for this week! Niantic’s vision of a virtual world isn’t just a fun concept—it’s evolving how AR could fit into our lives. But does the potential outweigh the risks? 

I’d love to hear your thoughts—hit reply and let me know what excites (or concerns) you most.

—Matt (FutureTools.io)

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