Google Unveils Powerful Yet Light Gemma 3

Welcome to another edition of AI PlanetX.

Google launches Gemma 3, a powerful compact AI; OpenAI unveils business tools for AI agents; Meta advances AI chip research.

• Hottest AI News

• Top AI & SaaS Tools: ElevenLabs F-R-E-E Speech to Text!

• Top AI & Tech News

• Interesting Uses of AI: Using AI to Improve Writing

• AI Video Tutorial

• Complimentary AI Course of the Day: Manus AI Agent Course

Hottest AI News

DeepMind

Google Unveils Gemma 3, Their Most Powerful Compact AI Yet

Google has unveiled Gemma 3, a major upgrade to its "open" AI model, and the most powerful system able to run on a single GPU. Building on last year's foundation, it expands capabilities while maintaining ease of use for developers.

Details:

• Processes text, images, and short videos in 35+ languages with an upgraded encoder for high-resolution, non-square images, outperforming Llama, DeepSeek, and OpenAI on single-GPU

• Optimized for Nvidia GPUs and AI hardware, with the ShieldGemma 2 image classifier filtering explicit, dangerous, or violent content

• Supports developers with cloud credits, including $10,000 for researchers via the Gemma 3 Academic program, while sparking debates on its "open source" licensing

Try it:

• Visit AI Studio and sign in using your Google account

• Once logged in, navigate to the Create Prompt section on the left

• On the right, open the Model dropdown menu, scroll all the way down and select Gemma 3 27B

• At the bottom, enter your query in the text box and press Run to send it

Modernize your marketing with AdQuick

AdQuick unlocks the benefits of Out Of Home (OOH) advertising in a way no one else has. Approaching the problem with eyes to performance, created for marketers with the engineering excellence you’ve come to expect for the internet.

Marketers agree OOH is one of the best ways for building brand awareness, reaching new customers, and reinforcing your brand message. It’s just been difficult to scale. But with AdQuick, you can easily plan, deploy and measure campaigns just as easily as digital ads, making them a no-brainer to add to your team’s toolbox.

Learn more now.

OpenAI

OpenAI Debuts Business Tools for AI Agent Creation

OpenAI has launched tools to help businesses create autonomous AI agents, making it easier for developers to build systems that perform complex tasks without ongoing human supervision.

Details:

• OpenAI's Responses API replaces Assistants API (sunsetting in 2026), enabling developers to create AI agents that can search the web, scan files, and navigate websites automatically, similar to Operator

• Developers can use GPT-4o search models, which outperform GPT-4.5 on factual questions with accuracy scores of 90% and 88% on OpenAI's SimpleQA benchmark

• It also includes the CUA model in research preview, allowing AI to automate tasks via mouse and keyboard actions, plus an open-source Agents SDK for integration, safeguards, and monitoring

OpenAI acknowledges limitations like the CUA model’s reliability and 10% error rate in search models. As Product Head Olivier Godement said, "Demoing an agent is easy; scaling and usage are hard."

Meta

Meta Takes Step Forward With AI Chip Testing

Meta is testing its first RISC-V-based AI chip for training workloads, potentially one of the industry's first for this purpose, marking a shift in its strategy to reduce reliance on Nvidia's GPUs.

Details:

• Meta has taped out its first AI training accelerator with TSMC and Broadcom. The chip has moved past initial testing, with samples in limited deployment to assess performance

• The processor likely includes HBM3 or HBM3E memory and a systolic array architecture for matrix computations. Meta has optimized the instruction set and data formats for better performance, power efficiency, and die size

• It is part of Meta Training and Inference Accelerator (MTIA) program, which has had setbacks. Despite relying on Nvidia GPUs for AI tasks, Meta plans to use its custom chips for AI training by 2026

Meta's RISC-V training accelerators aim for hardware independence and cost savings with open-source architecture, potentially matching Nvidia's performance-per-watt while giving Meta more control over its AI hardware.

Top AI & SaaS Tools

• Zemith (Life-time Deal): Use top AI models (GPT-4o, o1, DeepSeek, Claude, Gemini) for chat, search, note-taking, document analysis, and image generation

• Scribe: ElevenLabs has made its most accurate Speech to Text model available at no cost via the UI until April 9th [F-R-E-E]

• Dora: Turn ideas into stunning 3D-animated websites without any coding [F-R-E-E]

• Wispr Flow: Use your voice to write 3x faster in every application: AI commands, auto-edits, 100+ languages [F-R-E-E]

• Perplexity Windows App: Take advantage of voice dictation, keyboard shortcuts, and latest AI models with Perplexity app [F-R-E-E]

Top AI & Tech News

• OpenAI may soon release two models: o3-mini-pro and another, described by Sam, as skilled in creative writing, with an example being a metafictional short story

• Anthropic’s Claude boosted annualized revenue from $1B to $1.4B, partly due to powering Manus, a popular Chinese AI agent

• Reka open-sourced Flash 3, a 21B parameter reasoning model that competes with OpenAI's o1-mini, featuring a 32k context length

• Luma Labs unveils Inductive Moment Matching, a pre-training method delivering 10x more efficient, high-quality image generation

• Manus AI has partnered with Alibaba's Qwen team to advance its autonomous general AI agent beyond traditional chatbots

• Google DeepMind unveils Gemini Robotics and Gemini Robotics-ER, two AI models designed to improve robots' real-world task performance

Interesting Uses of AI

AI Art Spotlight

Model: Midjourney

Prompt:

Mother (black, green eyes) gently strokes black kitten (yellow eyes) with paw. Kitten looks relaxed, eyes closed. Siblings next to them. Alley background, light rain. Emotions emphasized. --ar 9:16 --v 6.1

AI Prompt Tip of the Day

Top AI Video Tutorial

How To Create Your Own CUSTOM ChatGPT For Any Use Case

Complimentary AI Course of the Day

Manus AI Agent FULL Course

In this course, the instructor teaches how to use Manus AI agents for automation and building various applications. The course covers everything from basic to advanced techniques, showcasing the powerful capabilities of Manus AI.