Holiday Digest in AI 🎁

🎄 Merry Christmas! I hope this newsletter finds you in good spirits and ready to celebrate the festive season. Here's a quick rundown of the latest updates and news in the world of tech and AI.

We've just launched a survey to gather your valuable feedback. You can participate by visiting this link: 📌 Feedback Survey. As a token of our appreciation, participants who complete the survey and leave their registered email address will receive a free month of Prompter Plus!

🚀 What’s New in Prompter

🌟 This Week's Release

• Live chat is now available for real-time support. Click the button in the lower right corner to get started

• Several bugs have been squashed.

• We're currently developing some new features. Stay tuned for updates.

📢 New Posts

• 📌 AI Model Costs: Pricing on GPT, Claude, Gemini & More

  This blog post aims to be your one-stop resource, simplifying the pricing information of major LLMs like OpenAI's GPT series, Anthropic's Claude, and Google's Gemini, etc.

• 📌 Understanding Token Limits in OpenAI’s GPT Models

  In this post, we explore token generation in OpenAI's GPT models, focusing on context window sizes and their impact on token limits. Understanding these aspects is key to optimizing GPT wrappers and cost management.

📰 Top AI News for Devs

• Vecel v0 is now open for everyone! You can generate UI in seconds with text or images.

  

• OpenAI Chat Completion API now supports log probabilities. logprobs and top_logprobs parameters are currently available for gpt-4-vision-preview. Learn more at OpenAI's API Reference and OpenAI Devs Twitter.

  • If logprobs is set to true, a list of message content tokens with log probability information will be returned along with the chat completion message.

    {
        "token": "Hello",
        "logprob": -0.31725305,
        "bytes":
        [
            72,
            101,
            108,
            108,
            111
        ],
        "top_logprobs":
        [
            {
                "token": "Hello",
                "logprob": -0.31725305,
                "bytes":
                [
                    72,
                    101,
                    108,
                    108,
                    111
                ]
            },
            {
                "token": "Hi",
                "logprob": -1.3190403,
                "bytes":
                [
                    72,
                    105
                ]
            }
        ]
    }

  • Also, check out how Microsoft utilized log probabilities for high-scoring Medprompt here.

• Anthropic's Claude Messages API Update - The new Messages API will help you catch errors early in development, particularly with respect to prompt construction, so that you can get the best output from Anthropic's models. Visit Anthropic's Website for more info

  

• Chatbot Arena Rankings - The rankings on Chatbot Arena seem intuitive. GPT-4-turbo is still far ahead. And OpenAI offers better quality than cheaper models while better pricing than those that outperform it. The latest GPT-3.5-Turbo-1106 seems to have deteriorated in quality. Stay updated with the latest rankings here.

  

🌐 Prompt Sharing

• IELTS Speaking Simulator: Just found this GPT which offered an authentic practice experience, helpful for anyone preparing for IELTS. The well-structured prompt is available here.

  

📖 Worth Reading

• 📌 OpenAI Prompt Engineering Guide: A must-read guide from OpenAI for anyone interested in prompt engineering.

• 📌 The Year A.I. Ate the Internet: A look back at how AI transformed the internet in 2023 from The New Yorker.

• 📌 Unbundling AI: A fascinating read on the evolution of AI.

• 📌 Why Most Agent AIs Don't Work (Yet): An insightful article discussing the current challenges faced by AI agents.

• 📌 Outsider Thinking and the Age of AI: Understanding AI from a unique perspective.

Merry Christmas to you and your family! May your holidays be filled with joy and laughter 🎉