Welcome to the [AI Daily] column! Here is your daily guide to exploring the world of artificial intelligence. Every day, we bring you the hottest content in the AI field, focusing on developers, helping you to understand technical trends and learn about innovative AI product applications.

Fresh AI Products Click to Learn: https://top.aibase.com/

Major PR Update! Sora Joins Adobe Suite, Video Editing Gets AI Magic

Adobe Premiere Pro has recently undergone a groundbreaking update, introducing the AI tool Firefly, which greatly simplifies the video editing process and enhances editing efficiency. New features include object addition and removal, generative AI capabilities, etc., providing users with more creative flexibility. This move heralds a significant transformation in the video editing field.

image.png

[AiBase Summary:]

⭐ Adobe Premiere Pro is introducing AI tools to simplify the video editing process and improve editing efficiency.

⭐ Currently launched features include automatic generation of transcripts, AI audio classification tags, speech-to-text, enhanced speech, automatic scene cutting, automatic color grading, etc.

⭐ The introduction of third-party AI models, including Sora, Runway, Pika, etc., allows users to click and drag to extend videos.

⭐ Users can use AI tools to remove, add, extend videos, transcribe videos, etc.

Product Experience Address: https://top.aibase.com/tool/adobe-premiere-pro

Three Important Messages from Baidu Create AI Developer Conference

[AiBase Summary:]

🌟 Baidu officially released the Wenxin Large Model 4.0 tool version, with code generation, interpretation, and optimization reaching international advanced levels.

🌟 Baidu released the Wanyuan Intelligent Computing Operating System, supporting 77 large models.

🌟 Baidu introduced three development tools: AgentBuilder, AppBuilder, ModelBuilder, to lower the technical threshold for development.

🌟 Baidu's Wenxin Yiyan has over 200 million users, with API daily call volume reaching 200 million.

GPT-4Turbo Defeats Claude3, Reclaiming the Title of "Best AI Model"

OpenAI's latest update has propelled GPT-4Turbo to surpass 82 LLM models in Chatbot Arena, demonstrating outstanding performance and reclaiming the top spot. Users can compare the performance of different models in Chatbot Arena.

image.png

Highlights Summary:

1、⭐ GPT-4Turbo surpasses 82 LLM models in Chatbot Arena.

2、⭐ Users can compare model responses, with GPT-4Turbo performing exceptionally.

3、⭐ Visit the Chatbot Arena website for comparison and testing.

Quora's Poe Introduces Multi-Robot Chat Functionality

Poe introduces multi-robot chat functionality, aiming to become a conversational AI app store, offering monetization tools and enterprise services. Facing challenges and uncertainties, such as the variability of chatbot experiences, quality control, cost pressures, etc.

image.png

Highlights Summary:

1、⭐ Poe introduces multi-robot chat functionality, integrating users with multiple AI models in a single conversation.

2、⭐ Poe plans to become a conversational AI app store, providing monetization tools and enterprise services for creators.

3、⭐ Challenges and uncertainties: Variability in chatbot experiences, quality control, cost pressures, etc., need to be addressed.

Product Entry: https://top.aibase.com/tool/poe

World's First AI Beauty Pageant, the Most Popular AI Beauty to Win a $13,000 Prize

image.png

[AiBase Summary:]

⭐ Fanvue is hosting the world's first "Miss AI" beauty pageant, with a total prize pool of over $20,000.

🤖 Contestants are AI-generated models, with judges including AI influencers and beauty pageant historians.

💰 The champion will receive a total prize of $13,000, but the actual winners are the creators of these robots.

Detailed Content: https://www.chinaz.com/2024/0416/1610520.shtml

2txt Image-to-Text Tool Can Recognize Text in Images and Convert It to Editable Text

1.png

[AiBase Summary:]

⭐ 2txt, based on AI, can recognize image text and convert it to editable text.

⭐ Created using Claude Haiku and Vercel AI SDK, it is accurate and fast.

⭐ The Next.js framework provides efficient web rendering, allowing users to upload an image to obtain text results.

Experience Address: https://top.aibase.com/tool/2txt

Project Page: https://github.com/ai-ng/2txt

Ctrl-Adapter: Controlnet Designed for Video Generation, Providing Effective Fine Control

image.png

[AiBase Summary:]

🎥 **Rich Control Features**: Provides multiple control options for image and video models.

📹 **Sparse Frame Video Support**: Effectively handles low frame rate videos, providing effective control.

🖼️ **Compatible with Multiple Base Models**: Supports multi-condition control generation process, improving generation efficiency.

Project Address: https://top.aibase.com/tool/ctrl-adapter

GPT-4 Cleans Up Academia! Retraction Prediction Matches Human Accuracy by 95%

[AiBase Summary]

🔍 GPT-4 predicts academic paper retractions with 95% accuracy, similar to human reviewers.

🔍 Discussing tweet prediction of paper retractions, GPT-4 performs excellently.

🔍 Researchers find GPT-4 has high potential in predicting retractions, but need to be aware of "hallucination" issues.

Google to Invest Over $100 Billion in Developing AI Technology

Google plans to invest over $100 billion in developing AI technology, demonstrating its ambition in the field of artificial intelligence. Public acceptance of AI systems is gradually increasing, although the systems still have flaws and errors. Google leads its competitors in computing power.

image.png

Highlights Summary:

1、⭐ Google plans to invest over $100 billion in developing AI technology, showing its ambition in the AI field.

2、⭐ Public acceptance of AI systems is increasing, despite the systems still having flaws and errors.

3、⭐ Google leads its competitors in computing power.

New Version of Anti-AI Scraping Tool Glaze2 Released, Protecting Artists' Works from Unauthorized AI Scraping

The Glaze project team has released Glaze2, a tool designed to help artists protect their works from AI model scraping, providing a faster and more secure tool. The new version targets new AI models and will expand protection to short videos and animation fields.

image.png

Highlights Summary:

1、🎨 New version of Glaze2 tool released, helping artists protect their works from AI model scraping.

2、🔒 Provides faster and stronger protection, specifically targeting new AI models like Stable Diffusion XL.

3、📹 Plans to expand protection to short videos and animation fields.

Details Entry: 🔗 https://top.aibase.com/tool/glaze

Jina-ai/Reader: Convert Any URL into a Format Friendly for Large Models

The Jina AI team has developed the Reader tool, an innovative solution that can convert any webpage URL into a format friendly for large language models (LLM), greatly improving the efficiency of language models in processing and understanding web content. Suitable for developers and researchers to use, it enhances the efficiency and output quality of LLM-based applications.

image.png

Highlights Summary:

1、💡 Improves input quality, formats and cleans URL content, ensuring LLM receives standardized and easy-to-process inputs.

2、💡 Allows the system to process large or real-time data in a streaming manner, suitable for fast response application scenarios.

3、💡 Supports multiple modes, adapting to different technical needs and application scenarios.

Details Entry: 🔗 https://github.com/jina-ai/reader

EleutherAI Releases Latest Model Pile-T5 to Address Limitations in Code Processing Tasks

EleutherAI has recently released the new T5 model Pile-T5, aiming to address the limitations of the original T5 model in processing code-related tasks and the possibility of the tokenizer missing important code tokens, marking a significant advancement in the fields of NLP and code understanding.

image.png

Highlights Summary:

1、⭐ Pile-T5 is trained twice as much as the original T5, reaching 2 million steps, learning richer language patterns and code structures.

2、⭐ Pile-T5 adopts the new LLAMA tokenizer, handling code-related tasks more accurately.

3、⭐ Pile-T5 shows significant improvements in fine-tuning downstream tasks, especially in code tasks, outperforming T5-v1.1.

Details Entry: 🔗 https://top.aibase.com/tool/pile-t5

Adobe Launches AI Assistant for PDF Reading

Adobe has launched an AI-based PDF reading assistant, with a subscription price starting at $4.99/month. The AI assistant can parse PDF content and quickly extract information through conversation, improving document processing efficiency.

image.png

Highlights Summary:

1、⭐ The AI assistant can parse PDF content, allowing users to quickly obtain the information they need through conversation.

2、⭐ Subscription price starts at $4.99/month, providing a chatbot-like interface for users to ask questions and obtain information.

3、⭐ The AI assistant also supports formats such as Word, PowerPoint, etc., improving document processing productivity.

Details Entry: 🔗 https://top.aibase.com/tool/adobe-acrobat-ai-assistant