Meta Just Revolutionize the AI Space With This..

Anthropic's Constitutional Classifiers, OpenAI’s Deep Research and o3-mini, & more

February 05, 2025

In partnership with

AI has been picking up the last few weeks and it seems like we are only getting started…

Here are the top stories this week:

  • Japan's Bold AI Leap: SoftBank and OpenAI Drive a $3B Annual Investment

  • Anthropic's Constitutional Classifiers: Raising the Bar on AI Security

  • OpenAI’s Deep Research: Elevating ChatGPT for In-Depth Analysis

  • OpenAI’s o3-mini: Fast, Cost-Efficient STEM Reasoning

  • Meta just dropped VideoJam and redefines physics

SoftBank commits $3 billion annually to integrate OpenAI technology across subsidiaries like Arm and PayPay, while launching “Cristal Intelligence” to deliver a tailored version of ChatGPT and OpenAI’s API exclusively for Japanese businesses.

This move positions SoftBank as a major player in Japan’s AI race, complementing its role in the "Stargate" data center project with OpenAI and Oracle, and challenging traditional consulting services with secure, customized enterprise solutions.

Source: SoftBank

Anthropic has unveiled its new AI safety system, Constitutional Classifiers, designed to thwart diverse jailbreak attempts by generating training data in multiple languages and styles. In tests against 10,000 advanced attacks, the system blocked 95.6% compared to just 14% for unprotected models, following over 3,000 hours of rigorous bug bounty trials by 183 security experts.

This breakthrough underscores the growing importance of robust AI safeguards as models become more powerful. Anthropic is now inviting public testing until February 10, demonstrating its commitment to collaboratively advancing AI safety measures.

Source: Insider’s Edge via Ideogram

OpenAI has introduced Deep Research, a new ChatGPT feature that conducts extensive web research on complex topics, delivering detailed, citation-rich reports in under 30 minutes. Leveraging a specialized version of o3, the system analyzes text, images, and PDFs from multiple sources, guiding users with clarifying questions and timely notifications throughout the research process.

Initially available to Pro subscribers at $200/month (with 100 queries per month), Deep Research has already demonstrated its prowess by significantly outperforming competitors on challenging exams. With plans to expand access to Plus and Team users pending stable safety metrics, this launch marks a pivotal step towards more autonomous, long-form research capabilities in AI.

Source: OpenAI

OpenAI has launched o3-mini, a new reasoning model that delivers advanced STEM capabilities—especially in math and coding—at 24% faster response times. Free users now gain access to their first AI-driven reasoning experience, while paid subscribers enjoy expanded limits of up to 150 messages daily.

Developers can fine-tune the model's "reasoning effort" across low, medium, and high settings, all at 63% lower operational costs—just $1.10 per million input tokens. With the full o3 rollout coming soon, o3-mini sets the stage for broader access to high-performance AI reasoning.

Source: OpenAI

In Other News

US stock futures slipped as escalating U.S.-China trade tensions rattled investors. After President Trump imposed 10% tariffs on Chinese goods and China countered with tariffs on 80 US products, S&P 500 and Nasdaq futures edged down by 0.2% and 0.3%, respectively, while Dow Jones futures remained roughly flat.

On the earnings front, Alphabet faced a double hit with an antitrust probe and disappointing Q4 results, sending shares down 7% post-bell. Similarly, AMD and Chipotle experienced after-hours declines following cautious guidance, setting the stage for Disney’s upcoming earnings report amid persistent market uncertainty.

Insider’s Edge via Ideogram

An entirely new way to present ideas

Gamma’s AI creates beautiful presentations, websites, and more. No design or coding skills required. Try it free today.

Meta just dropped VideoJam and here's why this is a major breakthrough.
Every video model out there struggles with physics. This is why you see deformed humans and characters specially when they are in movement. What differentiates this model is that while other AI video models guesses frame by frame, VideoJam solves this by doing a completely different approach.

You first start out by typing your prompt, then AI predicts the movement first, after, it creates the video around it and this is how it is able to generate perfect physics every time.

Cool tools of the week from insidersedge.io

Web 3 Summary - Simplify complex crypto projects using ChatGPT and AI.

Ghola - Get personalized conversations wit this AI chatbot.

SaaS Prompts - AI & ChatGPT prompts for SaaS entrepreneurs.

Ocoya - Create & schedule posts faster on social media.

Jobs & Opportunities

Senior Software Engineer AI - Ripple

Applied AI ML Engineer - Heretic

AI Communication Designer - ZAUBAR

Thanks for tuning into today’s edition!

How's your vibe with our newsletter?

Be brutally honest. DM me or email me back with any suggestions!