A Look Inside OpenAI DevDay

Plus Pika Labs New AI Model...

Author

Jelani Abdus-Salaam
October 02, 2024

In partnership with

Welcome back, AI’ers!

Here’s What’s On The Menu For Today

  • 🤖 A Look Inside OpenAI DevDay

  • ✨ Pika Labs Drops 1.5

  • 📰 News Hot Off The Press

  • 💰️ Newly Added AI Tools

A MESSAGE FROM OUR PARTNER

The fastest way to build AI apps

  • Writer Framework: build Python apps with drag-and-drop UI

  • API and SDKs to integrate into your codebase

  • Intuitive no-code tools for business users

TODAY’S TOP STORIES

👾 OpenAI’s DevDay brings Realtime API and other treats for AI app developers (LINK)

OpenAI's 2024 DevDay unveils exciting new tools for AI app developers, including the Realtime API for low-latency, AI-generated voice responses and vision fine-tuning capabilities. Despite recent executive departures, the company remains focused on innovation and maintaining its competitive edge in the AI industry.

  • The Realtime API enables developers to create nearly real-time, speech-to-speech experiences with six provided voices.

  • OpenAI introduced vision fine-tuning in its API, allowing developers to use images and text to improve GPT-4o's visual understanding capabilities.

  • The company also launched features like prompt caching and model distillation to help developers reduce costs and enhance performance.

OpenAI also announced the following:

  • An increase of the API rate limits on the o1 model equal to that of GPT-4o (10,000 requests per minute)

  • A reduction of the price of GPT-4o API calls with automatic prompt caching, making repeated API calls 50 percent cheaper with no extra developer effort

  • A multi-modal fine-tuning API that allows developers to fine-tune GPT-4o with images in addition to text

  • Triple the number of active apps are on the OpenAI platform from last year to this year, and there are 3 million active developers.

🔴 New AI models are more likely to give a wrong answer than admit they don't know (LINK)

New AI models are becoming less transparent about their limitations, often providing incorrect answers instead of admitting uncertainty. A study from the Universitat Politècnica de València tested various large language models (LLMs) on math, science, and geography questions, revealing improved accuracy on complex problems but decreased reliability on simpler ones.

  • Newer LLMs like GPT-4 showed a significant drop in "avoidant" answers compared to previous versions1

  • While capable of solving challenging problems, these models still struggle with basic questions.

  • The study found no apparent improvement in solving simple problems across newer AI models.

🖥️ Udemy screws over teachers by stealing their content (LINK)

Udemy, a major e-learning platform, offered instructors a brief window to opt out of having their course content used for AI training, which has now closed. The company plans to use course materials to train AI models for enhancing the platform's features and potentially creating new AI-powered tools.

  • Instructors had until September 30, 2023, to opt out of having their content used for AI training.

  • Udemy's AI initiatives aim to improve search functionality, content recommendations, and course creation assistance.

  • • The company emphasizes that AI-generated content won't replace human instructors but will complement their efforts.

SPONSOR US

Get your product in front of over 20k+ tech & AI enthusiasts

Our newsletter is read by thousands of tech professionals, investors, engineers, managers, and business owners worldwide. Get in touch to learn more.

OTHER NEWS YOU MAY LIKE

Microsoft brings AI-powered overviews to Bing (link)

Meta won’t say whether it trains AI on smart glasses photos (link)

Chromebooks Are Getting a Dedicated AI Key, but You Won’t Use It for AI (link)

Y Combinator is being criticized after it backed an AI startup that admits it basically cloned another AI startup (link)

The Art of the OpenAI Deal (link)

Japan's Shimano gives cyclists an easier ride with AI gear shifter (link)

Man tricks OpenAI’s voice bot into duet of The Beatles’ “Eleanor Rigby” (link)

OpenAI Pitched US on Unprecedented Data Center Buildout (link)

California Enacts Additional Generative AI Bills Touching on Training Data and Healthcare Decisions (link)

FROM THE WEB

✨ Pika Labs Drops A New Video Model - 1.5

Pika Labs released an updated video model that is impressive.

This update brings significant improvements in realism and creative capabilities for video production.

Pika 1.5 introduces "Pikaffects," allowing for hyper-realistic human and creature movements with dynamic physics and sophisticated camera techniques.

The update includes new motion control features like Bullet Time and Vertigo shots, enhancing the cinematic quality of generated videos.

While offering more powerful capabilities, the new model requires more credits per video generation, reflecting increased resource demands.

AI QUICK TIP

Onboard AI as if it’s a new co-worker by giving it instructions and context.

Without providing the proper context or instructions your out put will be garbage.

Other things to remember:

  • Set specific targets

  • Specify structure and style

  • Stay concise and focused

  • Check for clarity and typos

🔮 MYSTERY LINK

This link might showcase Mark Zuckerberg's metaverse vision or how to develop immersive virtual reality experiences or anything in between.

PROMPT OF THE DAY

Copy and paste this prompt 👇

"Analyze [trend] and give me ideas on how to uniquely use [trend] to achieve [goal].

Trend = [Insert here]

Goal = [Insert here]"

Want more prompts?

We have a database of over 8,000 of the best prompts we’ve collected, which you can access by sharing Best of AI using the link below.

THAT’S IT FOR TODAY

Until next time, stay safe.

-Best of AI Team

Remember to click to unsubscribe if you don’t want us to land in your inbox anymore.

What did you think of this issue?

Login or Subscribe to participate in polls.

Please reply to this email if you have specific feedback or anything interesting to share.