DevThink.AI

News to assist you in navigating the AI tools and trends which benefit you as a software developer

⭐️ Top stories

Introducing 💫 StarCoder: The Revolutionary 15.5B Parameters Language Model Trained on Source Code and Natural Language

StarCoder is a language model trained on both source code and natural language text. It has a 8,192 tokens context window and has been trained on more than 80+ programming languages.

In my simple testing, I put StarCoder up against GPT-4. It did quite well with strait forward asks, Here the asks was:
“given me an example of joining 2 strings in python”

I appreciated that it showed me there are three ways to perform this in python.

In another example though I asked a bit more extrinsic task such as: “give me an example of AWS CDK for creating an s3 bucket”

Here, GTP-4 gave a very complete, yet succinct answer. StarCoder though went a bit off the rails, spouting a diatribe of irrelevant CDK code; crafting VPC’s and EKS stacks in addition to the S3 bucket 😕

It is still early for this model and my testing was very unscientific. I’m a huge fan of open-source/open-access so I have no doubt the armies of contributors for this model will soon astound us with the value it can bring.

ChatGPT Python Applications on GitHub - A Comprehensive Collection of Python Apps for Natural Language Processing

Explore the Power of ChatGPT in Various Applications with 'ChatGPT Python Applications' Repository on GitHub! This comprehensive collection of Python applications showcases the versatility of the ChatGPT language model in natural language processing, including chatbots and speech-to-text conversion. All applications are well-documented and open-source, allowing developers to integrate ChatGPT into their projects with ease. Subscribe to @qxresearch on YouTube for updates on new projects and join a community of Python enthusiasts.

Slack Unveils Vision for Generative AI in Platform: Slack GPT

Slack has unveiled its vision for generative AI in its platform. Slack GPT brings trusted generative AI to where teams already work, with features including AI-powered conversation summaries and writing assistance. The AI-ready platform lets customers integrate and automate with their language model of choice, using partner-built apps or building their own custom integrations. Slack GPT is being built with customers at the center of the experience, ensuring transparency and control over how AI is used. Availability includes the Claude app, the ChatGPT app in beta, and Slack GPT native AI capabilities and the Einstein GPT app are in development.

No matter the collaboration tool your team uses, expect these AI integrations to be already available or on the way. Link me to any you find interesting.

Open-Source: Embracing Collaboration to Stay Relevant in the Generative AI Landscape

Generative AI development is undergoing a revolution as open-source models rapidly advance, challenging the positions of companies like Google and OpenAI. Foundation models are becoming increasingly accessible, with innovations like LLMs on phones and scalable personal AI. Open-source models offer greater customizability and privacy at a lower cost. The key to remaining relevant is embracing collaboration, learning from the open-source community, and prioritizing third-party integrations. Focusing on smaller, faster models could result in quicker iteration and better long-term capabilities.

What is a vector database

Pinecone dives deep into some of the tech behind vector databases such as their own. There is of course a bit of a sales pitch from Pinecone in this summary but so be it, they’ve done a great job of explaining the technology so they are certainly entitled to recommend their product 🙂

🧰 Tools

Databerry: Datasource integration with LLM’s, as simple as it gets

Unlock the potential of your data with the Databerry platform, a versatile solution designed for users of ChatGPT, developers, low-coders, and no-coders alike. Seamlessly integrate a variety of data sources, and connect them with powerful language models like ChatGPT to streamline your workflow and enhance your projects.

I gave this one a try on the free plan, and I quickly found a great use for it; Loading Terms of service and asking questions of it. It took only 5 minutes to have the GitHub terms of service loaded and the interrogation started.

🏫 Education

When you have an extra 30 minutes here and there during the week, consider exploring these resources to expand your understanding of the technologies behind generative AI tools that can be utilized by you as a software developer.

ChatGPT Prompt Engineering for Developers

In the ChatGPT Prompt Engineering for Developers course, taught by Isa Fulford from Open AI and the AI OG 😎; Andrew Ng. Developers will learn to use large language models (LLMs) for application development through the OpenAI API. The course covers best practices for prompt engineering, LLM APIs, and hands-on practice with various tasks like summarizing, inferring, transforming text, and expanding. Participants will also learn to build custom chatbots. This beginner-friendly course requires basic Python knowledge and is suitable for advanced engineers too.

In total there is ~90 minutes of content broken up into roughly 12 minute’ish segments.

 Thanks for reading and we will see you next time

Follow me on twitter, DM me links you would like included in a future newsletters.