• Bits & Neurons
  • Posts
  • GPT-4 Turbo with 128K context, lower prices, Assistants API, DALL·E 3 API, and more

GPT-4 Turbo with 128K context, lower prices, Assistants API, DALL·E 3 API, and more

Author

Ts Yotov
November 07, 2023

🤖 Everything from OpenAI’s first DevDay 🤖

Key Takeaways:

  • GPT-4 Turbo Introduction: an enhanced version of GPT-4, with a larger context window of 128,000 tokens and a knowledge cut-off in April 2023. This development aims to provide more depth and updated information for users interacting with the model.

  • 2.75x Cheaper Pricing Structure: OpenAI has slashed GPT-4 Turbo pricing to $0.01 per 1,000 input tokens, roughly 2.75 times cheaper, making AI integration more accessible for developers. GPT-3.5 prices are slashed too.

  • Custom GPTs for Users: Users are empowered to create their own GPT models without coding knowledge. This capability could democratize AI usage, allowing non-technical creators to customize AI for specific tasks or entertainment.

  • OpenAI GPT Store: The upcoming GPT Store will allow users to publish and potentially monetize their custom GPT creations. This move could foster an ecosystem of niche AI models catering to diverse market needs.

  • Assistants API Launch: The new Assistants API allows developers to embed AI-driven assistants into their applications. These assistants can perform a variety of tasks, from coding support to vacation planning, potentially enhancing app functionality and user experience.

  • DALL-E 3 API Release: The DALL-E 3 text-to-image model is now accessible via an API, complete with moderation tools. This API is priced per generated image, making advanced AI image generation more available to developers and creatives.

  • New Text-to-Speech Models: A new text-to-speech API featuring six preset voices has been launched, widening the scope for developers to integrate human-like speech in applications at a low cost per input character.

  • Copyright Shield Program: With the Copyright Shield, OpenAI commits to legally defending businesses using its products from intellectual property lawsuits. This support could encourage more companies to integrate AI into their offerings without fear of legal repercussions.

  •  Whisper v3: the next version of OpenAI's open-sourced speech recognition model, with improved accessibility and accuracy of voice-driven applications.

The advancements from OpenAI's developer event herald a brighter future for AI application, with easier access, greater customization options, and robust legal protections fostering a nurturing environment for innovation and growth in the AI sector.

📰 Quick News 📰

🧑‍🏫 Join the First NVIDIA LLM Developer Day: a virtual event focused on teaching developers how to build LLM-based applications. It's hosted by the NVIDIA Deep Learning Institute on November 17 and offers practical guidance on how to use key technologies, pros and cons of different methods, and a chance to create and deploy applications using various APIs and LLMs.

⚙️ Following the launch of Grok, the xAI team has now introduced PromptIDE, an integrated development environment revolutionising prompt engineering. This tool equips users with an SDK to apply sophisticated prompting techniques and offers rich analytics, drastically improving the knowledge and efficiency of users in prompt engineering and interpretability research.

🎞️ YouTube is experimenting with a new AI chatbot that provides additional information about videos and recommends related content. Another AI tool aims to categorize video comments into topics, making discussions easy to follow.

💰 Binny Bansal is venturing into a new AI startup following his success with Flipkart. Centered around providing AI talent, products, and software to corporate clients, the startup is primed for the legal and e-commerce sectors.

🧑‍🤝‍🧑Microsoft announces collaboration with Inworld AI to create a high-tech AI toolkit for Xbox game developers. This innovative toolset will enable the incorporation of AI-powered plotlines and characters to create dynamic gaming narratives.

📚 Nerd section 📚

🖥️ Understanding the impact of OpenAI's GPT4 Turbo on "Wrappers". Many startups could suffer at any point due to 'native' new tools from OpenAI itself.

⚒️ Unlocking greater potential with ChatGPT’s Advanced Settings - if you consider various parameters of ChatGPT like Top P, frequency and presence penalties, stop sequences, and maximum length, you could significantly optimize its application.

🌐  OpenAPI introduction and 11 practical examples of using ChatGPT models - such as code explaining, grammatical correction, translating text to emojis and vice-versa, and text summarization of data in business settings.

🌐 Web section 🌐 

🕸️ What's this new section about?

On days when there's a lot of news, we aim to keep our newsletter short and sweet.

So here (but not in the email version) we're adding AI-related stories that may be slightly less attention-grabbing, but still could move the needle in specific sectors and fields.

🏭 Baidu, a top-ranking AI firm in China, opted for 1,600 Huawei's AI chips against Nvidia, effectively reshaping the landscape of China's tech industry by favoring an internal player instead of an American company.

👷🏽 The Google contractors who contributed to the Bard AI chatbot as part of Accenture voted to unionize, a key development following the lay-offs in August, amid complaints about treatment at work. 26 out of 36 eligible voters agreed to this move, which was primarily aimed at getting better control over work assignments, better pay, and improved benefits.

🚕 Cruise's robotaxis are stated to require remote human assistance every 4 to 5 miles, accounting for 2-4% of the total drive time, primarily for navigation support rather than remote control. Despite its claim of 'autonomous' driving, Cruise’s reliance on frequent human interventions questions the actual autonomy and effectiveness of such vehicles.

🏭 Intel has suspended an investment that would have nearly doubled its operations in Vietnam. This decision came as a surprise as it happened after President Joe Biden's supportive deals for Vietnam's chip industry. Power supply instability and heavy bureaucracy are suspected to be the factors behind Intel's decision.
+
💸 Reports suggest that Intel is potentially in line to receive billions in government funding for creating secure defense-chip facilities, a pivotal move to limit U.S. reliance on East Asian chip imports, particularly from Taiwan.

📃 The Artificial Intelligence (AI) tool, Family-Match, conceived by Thea Ramirez, based on a scoring algorithm is under scrutiny. Despite optimistic projections, in multiple states it failed to notably increase successful adoptions but has injected a technological angle to traditional methods.

Thank you for reading today’s edition!

We love to hear back from you!

Feel free to reply to our emails with questions, suggestions, or topics you'd like to see covered, or drop us a message on Twitter or Facebook.

Until tomorrow,
- Ts (Bits and Neurons)PLUS: