Skip to main content
Back

OpenAI Launches Coding Assistant for All ChatGPT Users

WATCH NOW

December 11, 2024 4 min read

Hey,

Welcome to AlphaSignal – the most read newsletter by AI developers. 


We bring you the top 1% of news, papers, models, and repos, all summarized to keep you updated on the latest in AI.

IN TODAY'S SIGNAL

Read time: 4 min 05 sec

🎖️ Top News

  • OpenAI releases Canvas, a real-time editing tool for text and code, supporting direct Python execution and debugging.


📌 Encord

⚡️ Trending Signals

📌 Assembly AI

💻 Top Models

  • HunyuanVideo generates high-quality videos from text and images, outperforming similar models.

  • TRELLIS creates versatile 3D assets from text prompts, excelling in local editing.

  • InternVL 2.5 is an open-source multimodal model series, supporting multi-image and video data.

🧠 Pytorch Tip

  • Use PyTorch’s pin_memory for faster data loading and improved GPU performance during training.

If you're enjoying AlphaSignal please forward this email to a colleague. 

It helps us keep this content free.

TOP NEWS

Coding Assistant

OpenAI releases Canvas, a real-time editing interface for text and code, to all ChatGPT users

⇧ 12,582 Likes

What's New

You can now access Canvas, OpenAI’s tool for editing text and code in real time, on the web and Windows app. Previously limited to paid plans, this expanded availability was announced during OpenAI’s “12 Days of OpenAI” series.

Key Highlights

  • Expanded Access: Available to all users on web and Windows; previously limited to premium plans.

  • Integrated with GPT-4o: Canvas automatically opens when you work on writing or coding tasks.

  • Data Visualization: Create, preview, and refine graphics from code without leaving Canvas.

  • Split-screen Workspace: You can edit text or code in a split-screen view without switching between tools.

  • Editing Shortcuts: Adjust text length, change reading levels, and add comments with one-click actions.

  • Version Control: Revert to previous versions of your work at any time.

Direct Python Execution

You can now run and debug code directly within Canvas. Test outputs, fix errors with ChatGPT’s suggestions, and preview graphics generated from code—all without leaving the interface. These features simplify iterative tasks like refining data visualizations and troubleshooting scripts.

Customize Your GPTs

New custom GPTs include Canvas by default, but you can enable it manually for existing setups in the settings. This gives you full control over how Canvas operates within your workflows.

Watch now

Automate Data Preparation with the only Agentic Data System

Gartner predicts that by 2026, over 60% of generative AI will be multimodal - up from less than 1% in 2023. But building multimodal AI requires large-scale, high-quality datasets—a complex and time-consuming challenge.


Encord Data Agents simplify this process. Use them as your ML copilot to automate data preparation with foundational models like PaliGemma 2, GPT-4o, OpenAI Whisper, Claude 3.5 Sonnet, and more.

Join Encord's ML Research team on December 16th to explore:

  • Automating multimodal data curation and annotation with Data Agents

  • Fine-tuning PaliGemma 2, GPT-4o, and OLMo 2 on your data

  • Best practices for building scalable multimodal pipelines

Spots are limited to 250 attendees. Reserve your spot now!

Register now

partner with us

TRENDING SIGNALS

Coding Assistant

Cognition launches Devin, an AI developer assistant for engineering teams, which automates frontend bug fixes and code refactoring

⇧ 3,139 Likes

LLM

DeepSeek updates DeepSeek-V2.5 with real-time internet search and 8% boost in math, coding, and writing performance

⇧ 1,934 Likes

Human-AI Interaction

Nous Research unveils simulators for studying human behaviors and AI interactions in digital environments

⇧ 1,529 Likes

Multimodal Model

Google rolls out Gemini Flash, its fastest and lightweight multimodal model through API with lower rate limits

⇧ 1,389 Likes

AI Performance Optimizatoin

Hugging Face presents TGI 3.0, an inference engine with 13x faster long prompt handling and 3x higher token capacity

⇧ 1,042 Likes

Transform Speech-to-Text workflows with a single API

Already using speech-to-text? Take it further with AssemblyAI. Tap into industry-leading accuracy for actionable insights, faster workflows, and easy integration.


Build conversation intelligence tools for sales, meetings, and more—optimized for scale.

Try the API ↗️

TOP MODELS

Text-to-Video

HunyuanVideo

⇧2,939 Downloads

HunyuanVideo helps you generate high-quality videos from text or images. It integrates a unified architecture for image and video generation using a 3D VAE, MLLM text encoder, and Transformer. Trained with 13B parameters, it outperforms models like Runway Gen-3 in motion quality (66.5%) and text alignment (61.8%).

3D Generation

TRELLIS

⇧ 102,105 Downloads

TRELLIS generates high-quality 3D assets from text or image prompts. It uses a unified Structured LATent (SLAT) representation and Rectified Flow Transformers for decoding into various formats like Radiance Fields, 3D Gaussians, and meshes. Trained on 500K objects with up to 2B parameters, it excels in versatility and local 3D editing.

MLLM

InternVL 2.5

⇧ 1,054 Downloads

InternVL 2.5 is a multimodal large language model series, building on InternVL 2.0 with improvements in training, testing, and data quality. It integrates the InternViT vision model with language models like InternLM 2.5 and Qwen 2.5, supporting multi-image and video data, and offering models from 1B to 78B parameters.

PYTORCH TIP

Streamline Data Loading with PyTorch's pin_memory for Faster Training

You can speed up data loading by setting 'pin_memory=True' in your DataLoader. This small tweak can reduce the time it takes to get data onto the GPU, especially with large datasets.

How it Works

pin_memory=True locks the data in page-locked memory, allowing the GPU to access it faster. This reduces the overhead of transferring data from CPU to GPU during training, improving overall performance.

When to Use

Use it when you train on large datasets or across multiple GPUs. This setting reduces data transfer bottlenecks and makes your training loop more efficient.



train_loader = torch.utils.data.DataLoader(
    dataset, batch_size=64, shuffle=True, pin_memory=True
)

At Alpha Signal, our mission is to build a sharp, engaged community focused on AI, machine learning, and cutting-edge language models, helping over 200,000 developers stay informed and ahead. We’re passionate about curating the best in AI, from top research and trending technical blogs to expert insights and tailored job opportunities. We keep you connected to the breakthroughs and discussions that matter, so you can stay in the loop without endless searching. We also work closely with partners who value the future of AI, including employers and advertisers who want to reach an audience as passionate about AI as we are.


Our partnerships are based on shared values of ethics, responsibility, and a commitment to building a better world through technology.Privacy is a priority at Alpha Signal. Our Privacy Policy clearly explains how we collect, store, and use your personal and non-personal information. By using our website, you accept these terms, which you can review on our website. This policy applies across all Alpha Signal pages, outlining your rights and how to contact us if you want to adjust the use of your information. We’re based in the United States. By using our site, you agree to be governed by U.S. laws.

Looking to promote your company, product, service, or event to 200,000+ AI developers? Let's work together.

Work with us