- The AI Pulse
- Posts
- 🤖 Google DeepMind’s New “Veo 2” > OpenAI’s “Sora”
🤖 Google DeepMind’s New “Veo 2” > OpenAI’s “Sora”
PLUS: Eighth Day of “12 Days of OpenAI” Event, Teaching a Robot Its Limits With an LLM
Welcome back AI enthusiasts!
In today’s Daily Report:
📽️Google DeepMind’s New “Veo 2” > OpenAI’s “Sora”
🎄Eighth Day of “12 Days of OpenAI” Event
🦾Teaching a Robot Its Limits With an LLM
🛠Trending Tools
💰Funding Frontlines
💼Who’s Hiring?
Read Time: 3 minutes
🗞RECENT NEWS
GOOGLE DEEPMIND
📽️Google DeepMind’s New “Veo 2” > OpenAI’s “Sora”
Image Source: Google DeepMind’s “Veo 2”/Prompt: Golden maple syrup pours over a stack of fluffy pancakes as crispy bacon sizzles./Screenshot
Google DeepMind announced “Veo 2,” a new State-of-the-Art (SotA) video generation tool that creates cinematic clips from text or images.
Key Details:
“Veo 2”:
“Veo 2” can generate 8-second cinematic clips at 4K resolution while understanding camera controls in prompts such as POV, Wide Shot, and Drone Shot.
“Veo 2” showcases impressive detail, realism, and Artifact Reduction: minimizing unwanted distortions like blurriness, blockiness, or pixelation.
It outperformed OpenAI’s “Sora” in head-to-head comparisons with 1,003 human raters who preferred the prompt accuracy of “Veo 2.”
It’s gradually rolling out through the VideoFX Waitlist in Google Labs, with plans to integrate “Veo 2” into YouTube Shorts in 2025.
“Imagen 3”:
“Imagen 3” is a text-to-image AI model that generates lifelike images. Now, it generates brighter, better-composed images.
“Imagen 3” renders richer details and textures across diverse styles, from Photorealism to Impressionism.
It also follows prompts more faithfully, ensuring greater precision in translating text to images.
Why It’s Important:
“Veo 2” captures the Physics of cinematic scenes while comprehending specific use cases of Cause and Effect. For example, if a person bites an apple, the apple displays a bite mark.
Integrating “Veo 2” into YouTube Shorts could democratize video creation by allowing anyone to generate short-form cinematic clips in seconds.
🩺 PULSE CHECK
Are you on Team “Veo 2” or Team “Sora”?Vote Below to View Live Results |
OPENAI
🎄Eighth Day of “12 Days of OpenAI” Event
Image Source: OpenAI/YouTube/“Search-12 Days of OpenAI: Day 8”/Screenshot
OpenAI announced three major updates to ChatGPT Search during the eighth day of the “12 Days of OpenAI” event.
Key Details:
The “12 Days of OpenAI” event involves 12 livestreams across 12 days of “a bunch of new things, big and small.”
ChatGPT Search quickly and directly responds to prompts with up-to-date information from the internet with links to relevant sources.
Here’s a summary of the the major updates:
Use Your Voice: You can use Advanced Voice Mode to interact with ChatGPT Search through your voice.
Everyone Has Access: ChatGPT Search is now available to ChatGPT Free, Plus, Enterprise, and Pro plans.
Default Search Engine: You can set ChatGPT Search as your default Search Engine to write prompts directly into the browser bar.
OpenAI also integrated Maps into ChatGPT’s mobile app, so you can chat about local restaurants or find nearby businesses.
Why It’s Important:
Google search has been the dominant Search Engine for decades. For context, Google Search processes over 8.5 billion search queries daily. However, ChatGPT Search is threatening this dominance.
It directly provides concise answers with citations to relevant sources. In contrast, Google Search lists websites that may contain information relevant to your question.
AI RESEARCH
🦾Teaching a Robot Its Limits With an LLM
Image Source: MIT CSAIL/YouTube/“Teaching a robot its limits for open-ended chores”/Screenshot
When you’re at the gym, your friend might say, “You’re really pushing yourself hard today. Remember, it’s important to know your limits to avoid injury.” To a robot, the idea of “know your limits” refers to the robot’s ability to recognize its capabilities and limitations within environments when performing tasks.
Imagine asking a robot to clean your kitchen when it doesn’t understand the Physics of its surroundings. How can the robot generate a plan to ensure the kitchen is spotless? Large Language Models (LLMs) can help, but if the AI model is only trained on text, it’ll overlook the robot’s limitations like size, reach, and mobility relative to the kitchen. So, how can we guide the robot to clean the kitchen flawlessly?
Researchers at the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) created “ProC3S,” a method that equips LLMs with the ability to generate, test, and refine plans for robots.
Here’s a concise breakdown of how “ProC3S” works:
Task Description: The LLM is given a plan like “make breakfast by gathering ingredients, preparing the cooking area, and plating food.”
Plan Generation: The LLM breaks down the plan into smaller sub-tasks like “check the pantry for eggs, bread, and cereal.”
Simulation: The sub-tasks are simulated in a virtual environment, allowing the LLM to identify constraints or limitations.
Constraints: The LLM modifies the plan based on the robot’s limitations. For instance, the LLM suggests using a stool if the robot can’t reach a shelf in the pantry.
Refinement: The LLM continues to refine the plan until the robot can safely, efficiently, and successfully execute it.
🛠TRENDING TOOLS
📈Alpha builds a watchlist of stocks you care about.
🧠Constella quickly captures and recalls your thoughts.
🔍Reddit Answers is an AI-powered search engine for Reddit.
⚙️DepthAI deploys AI assistants that deeply understand your codebase.
✏️Paperguide enables you to discover, write, and manage your research.
🔮Browse our always Up-To-Date AI Tools Database.
💰FUNDING FRONTLINES
Wald.ai closes a $4M Seed Fund to protect data in conversations with AI assistants.
Pin raises a $3M Seed Round to build an AI-based recruitment tool that automates tedious hiring tasks.
TruVideo secures a $40M Growth Investment to build AI-driven communication technology for the transportation industry.
💼WHO’S HIRING?
Invesco (New York, NY): Client Research Intern, Summer 2025
Zoox (San Carlos, CA): Test Infrastructure Intern, Summer 2025
Block (San Francisco, CA): AI Research Intern, Cash App, Summer 2025
Astranis (San Francisco, CA): DevOps Engineer Intern, Flight Software, Summer 2025
NVIDIA (Santa Clara, CA): Developer Technology Engineer, Public Sector, New College Grad 2025
🤖PROMPT OF THE DAY
REVENUE
💵Managing Deferred Revenue
Develop a comprehensive way to manage Deferred Revenue for [Small Business] with [Product or Service] in [Industry] with [Target Audience]. Analyze how it impacts cash flow and complies with revenue recognition standards in [Specific Sector].
Small Business = [Insert Here]
Product or Service = [Insert Here]
Industry = [Insert Here]
Target Audience = [Insert Here]
Specific Sector = [Insert Here]
📒FINAL NOTE
FEEDBACK
How would you rate today’s email?It helps us improve the content for you! |
❤️TAIP Review of The Day
“Thanks for being a window to the AI world for me! :)”
REFER & EARN
🎉Your Friends Learn, You Earn!
You currently have 0 referrals, only 1 away from receiving ⚙️Ultimate Prompt Engineering Guide.
Refer 3 friends to learn how to 👷♀️Build Custom Versions of OpenAI’s ChatGPT.
Copy and paste this link to friends: https://theaipulse.beehiiv.com/subscribe?ref=PLACEHOLDER
Reply