AI That Thinks Like Us🧠💭

This week in the world of artificial intelligence, major breakthroughs are reshaping how we understand human cognition, generate multimedia content, and supercharge productivity tools. At the forefront is Helmholtz Munich’s Centaur AI model, a revolutionary system trained on over ten million human decisions that can simulate and predict human behavior with remarkable precision—ushering in a new era of psychological research and ethical AI design. Meanwhile, Baidu has unveiled MuseSteamer, a powerful image-to-video generator for enterprise users, alongside significant upgrades to its AI-driven search platform to enhance user interaction and personalization. On the productivity front, Perplexity launches its premium tier, Perplexity Max, offering unlimited access to AI-powered tools, early feature releases, and top-tier models like OpenAI’s o3-pro and Anthropic’s Claude Opus 4—tailored for professionals and researchers alike. These developments underscore AI’s accelerating influence across science, business, and everyday life. Let’s explore.

Centaur AI Model Accurately Simulates Human Decision-Making

Researchers at Helmholtz Munich have developed a groundbreaking AI model called Centaur, which can simulate and predict human decision-making with remarkable accuracy across a vast range of psychological experiments. Trained on the Psych-101 dataset—containing over ten million decisions from more than 60,000 participants in 160 behavioral experiments—Centaur not only matches but often outperforms traditional, specialized cognitive models in predicting human behavior. Unlike previous models, Centaur adapts to new scenarios and can generalize its predictions to unfamiliar tasks, even capturing reaction times and decision-making patterns similar to those observed in real people. The model’s internal mechanisms have been found to align closely with human brain activity, suggesting it has developed representations that mirror human cognition simply by learning from behavioral data. Centaur’s open-source nature and its ability to serve as a virtual laboratory open new possibilities for research in psychology, health, and neuroscience, while also raising important ethical considerations regarding privacy and responsible use.

Baidu Launches MuseSteamer AI Video Generator and Upgrades Search

Baidu has launched MuseSteamer, an AI-powered image-to-video generator aimed at business users, capable of transforming static images into short videos of up to 10 seconds and available in three versions—Turbo, Pro, and Lite. Unlike consumer-focused rivals such as OpenAI’s Sora, MuseSteamer is designed specifically for enterprise applications, targeting sectors like marketing, gaming, and e-commerce to streamline content creation and reduce production costs. Alongside this, Baidu has rolled out a major upgrade to its search engine, introducing a redesigned search box that supports longer, more detailed queries as well as voice and image-based searches, all powered by advanced AI to deliver more accurate and personalized results. These moves come as Baidu faces growing competition from domestic tech giants like ByteDance, Tencent, and Alibaba, and reflect the company’s strategy to strengthen its position in China’s rapidly evolving AI and digital services market.

Perplexity Launches Max: Unlimited AI Power and Early Access

Perplexity has introduced Perplexity Max, its most advanced subscription tier, designed for users seeking unlimited AI productivity and early access to the latest features. With Perplexity Max, subscribers get unlimited usage of Labs—allowing them to create dashboards, spreadsheets, presentations, and web applications without restrictions. Max users are the first to try new products like the Comet browser and receive exclusive benefits, including access to top-tier AI models such as OpenAI o3-pro and Claude Opus 4, plus priority customer support. Tailored for professionals, content creators, business strategists, and academic researchers who need comprehensive analysis and research tools, Perplexity Max is now available on web and iOS, with upgrades accessible through user settings. An Enterprise version of Max with unlimited Labs queries is also planned for organizational use.

Hand Picked Video

In this video, we'll look at how Olly’s Smart Post Borders help you instantly spot high-engagement posts on LinkedIn, X, and Instagram using intelligent color-coded borders and real-time engagement scores.

Get Your Free Trial Today and See How Instantly You Can Identify Top-Performing Posts with Olly.social Smart Post Borders.

Top AI Products from this week

• LLM Gateway - Route, manage, and analyze your LLM requests across multiple providers with a unified API interface.

• OpenMemory Chrome Extension - OpenMemory Chrome Extension is your shared memory across ChatGPT, Claude, Perplexity, Grok & Gemini. It captures useful context like preferences, goals & tasks from conversations and syncs it across tools, so you never have to repeat yourself.

• OpenDia - Connect your browser to AI models. No browser switching needed—works seamlessly with any Chromium browser, including Chrome & Arc.

• a0.dev - a0.dev (YC W25) helps developers build native mobile apps in seconds instead of weeks by generating full React Native apps with AI.

• Memolect - Memolect is your AI notetaker for dev teams - joins virtual meetings, transcribes, and suggests Jira, Linear & Confluence updates you can execute with a single click. Includes AI summaries and ChatGPT-style search to recall decisions instantly.

• Wagoo - Wagoo is your private, invisible AI desktop assistant. Get intelligent help anywhere with automatic screen context detection, voice transcription, and offline private AI capabilities. Instantly answer queries with latest models. Use local LLMs for privacy.

This week in AI

• Meta Forms Superintelligence Labs - Zuckerberg launches Meta Superintelligence Labs, led by Alexandr Wang, hires top AI talent, and commits billions to AI superintelligence research.

• DeepSeek-TNG R1T2 Released - DeepSeek-TNG R1T2 Chimera, a faster and smarter Tri-Mind AI model, launched with improved token consistency; available on Hugging Face under MIT license.

• Google Veo 3 Global Launch - Google Veo 3 now lets Gemini AI Pro users in 159+ countries create 8-sec videos with sound from text or images, limited to 3 videos daily.

• iFLYTEK AINOTE Air 2 Launch - iFLYTEK unveils AINOTE Air 2, a 230g, 8.2" E Ink AI tablet with multilingual transcription, Wacom stylus, and advanced productivity tools for $479–$499.

• DeepFleet AI Boosts Robot Fleet Efficiency - Amazon’s new DeepFleet AI foundation model optimizes robot routes, cutting travel time by 10% and enabling faster, more efficient deliveries across its global network.