Welcome back!

2024 is the year of Video AI, and Haiper has just announced its latest release allowing for the creation of 8-second videos from text, image, and video prompts, enhanced to 1080p HD with free access. Let’s unpack...

Today’s Summary:

• 🎥 Haiper launches 1.5 AI video generator

• 📊 Microsoft creates SpreadsheetLLM

• 🎬 Google tests AI video presentation creator

• 🚀 Microsoft CTO on AI progress

• 🧠 Meta advances System 2 LLMs

• 💰 Breakdown of OpenAI revenue

• 🛠️ 2 new tools

TOP STORY

Haiper 1.5 AI Video Generator

The Summary: Haiper, an AI startup founded by ex-Google DeepMind researchers, has launched Haiper 1.5. This update allows generation of videos up to 8 seconds with an HD upscaler. The company is aiming to compete with Runway and Sora.

Key details:

• Generates 8-second videos from text, image, and video prompts

• Upscaler enhances video quality to 1080p HD

• Platform has over 1.5M users since March launch of version 1.0

Why it matters: Haiper's progress shows the fast-paced innovation in 2024 for AI-generated video. As competition continues to heat up, we may see more advanced AI understanding of visual elements like motion and texture, pushing the boundaries of AI in creative tools.

Source: Haiper AI

RESEARCH

Microsoft Creates SpreadsheetLLM

The Summary: Microsoft researchers have created SpreadsheetLLM, an AI model designed to understand and work with spreadsheets. This new system uses encoding methods to compress spreadsheet data, making it easier for AI to process. SpreadsheetLLM shows promise in tasks like table detection and question answering.

Key details:

• SpreadsheetLLM uses a special encoding framework called SHEETCOMPRESSOR

• The system achieves a 96% reduction in token usage for spreadsheet encoding

• It outperforms previous methods in spreadsheet table detection by 12.3%

• The model introduces "Chain of Spreadsheet" technique for complex reasoning tasks

Why it matters: As spreadsheets remain a cornerstone of business operations, SpreadsheetLLM could change how companies extract insights from their data and we might see Excel and similar tools evolve from passive data containers to active partners in analysis.

Read the full paper

TOOLS

Google Tests AI Video Presentation Creator for Work

The Summary: Google has launched a beta version of Google Vids, an AI-powered video presentation tool, to select Workspace users. The app uses Gemini AI to generate customizable presentations from user prompts, including text, audio, and video elements. Google Vids is designed to make video creation as simple as making a slide, targeting business and enterprise accounts.

Key details:

• Available to "trusted testers" in Google Workspace Labs

• Uses Gemini and Vertex AI for content generation

• Offers royalty-free stock videos, images, voices, and music

• Allows collaboration on shared projects

Why it matters: This move shows Google's push to integrate AI into everyday work tools. As businesses seek efficient ways to create engaging content, Google Vids could become a standard tool to help teams communicate and present ideas. However, early user feedback suggests the tool may still need refinement to meet real-world needs.

Get started with Google Vids | Demo

QUICK NEWS

Quick news

• Meta researchers distill System 2 thinking into LLMs

• Microsoft CTO Kevin Scott thinks AI progress not slowing down, it's just warming up

• Estimated breakdown of OpenAI revenue

TOOLS

🥇 New tools

• Claude by Anthropic - New Android version

• Jobright - AI job search copilot