Prompt Engineering

Author: Fraser

  • AI as a Partner in Reshaping Teamwork and Innovation

    —

    by

    Fraser
    in

    New working paper: The Cybernetic Teammate: A Field Experiment on Generative AI Reshaping Teamwork and Expertise AI as a Collaborative Partner AI functions more like a teammate than a tool, replicating benefits of teamwork such as improved performance, expertise sharing, and positive emotional experiences. This perspective requires organisations to rethink team structures, training programs, and boundaries…

  • TinyTroupe

    TinyTroupe

    —

    by

    Fraser
    in , ,

    TinyTroupe by Microsoft is an LLM-powered persona simulation. An opensource Python library to simulate people with personalities, interests and goals. They can interact with other personas in their TinyWorld simulation. It is designed to use GPT-4 to help understand human behaviour with a focus on productivity and business scenarios for success with projects and products.…

  • Improve Your Claude Prompts

    —

    by

    Fraser
    in , ,

    A new tool from Anthropic to improve your prompts by following best practices for more reliable output. Take your existing prompt and automatically refine it. It will optimise the prompt for use with Claude. The prompt improver uses methods such as Chain-of-thought reasoning, will improve structure, format examples correctly in XML format that Claude prefers,…

  • Stripe Payments with AI Agents

    —

    by

    Fraser
    in

    Stripe has introduced an agent toolkit that integrates with popular LLM frameworks (Vercel’s AI SDK, LangChain, and CrewAI) to enable AI agents to handle financial transactions. The key innovation is the ability to combine LLM capabilities with Stripe’s API, allowing agents to perform tasks like creating invoices, generating payment links, and managing virtual cards for…

  • Nomic Data Mapping

    Nomic Data Mapping

    —

    by

    Fraser
    in

    Nomic Atlas lets you take unstructured data and create an interactive map. Great for visualising clusters of information. And it is not only for text. It is multi-modal and can handle images, audio and video. The also have an opensource JavaScript library for zoomable, animated scatterplots in the browser that scales over a billion points.…

  • The state of Generative AI 2024

    —

    by

    Fraser
    in

    AI at Wharton has published a report on Gen AI adoption, business applications and future prospects. In 2024 the focus has moved from the initial hype to proving ROI. Marketing, Operations and HR have seen an increase in AI usage and the overall sentiment is more optimistic and excited. There is more focus on enhancing…

  • Exa Search

    Exa Search

    —

    by

    Fraser
    in ,

    Exa is a new type of search engine, powered by AI technologies. I found it especially good when you want to find websites similar to each other. They also recently added “semantic search over LinkedIn” which seems interesting although apparently limited to profiles in the US, and people reporting mixed results.

  • A Survey of Prompting Techniques

    —

    by

    Fraser
    in ,

    The Prompt Report: A Systematic Survey of Prompting Techniques is a comprehensive resource that offers a structured understanding of prompts, a taxonomy of prompting techniques, and a meta-analysis of the entire literature on natural language prefix-prompting. It provides a robust vocabulary of 33 terms, a taxonomy of 58 text-only prompting techniques, and 40 techniques for other modalities.…

  • Top 100 Gen AI Consumer Apps

    Top 100 Gen AI Consumer Apps

    —

    by

    Fraser
    in , , , ,

    The third installment of the Top 100 Gen AI Consumer Apps reveals interesting trends in the rapidly evolving AI landscape. Nearly 30% of the companies featured are new compared to the previous report, with creative tools dominating 52% of the web list. ChatGPT maintains its top position on both web and mobile, but faces increasing…

  • Flux: A New Frontier in Text-to-Image Generation

    Flux: A New Frontier in Text-to-Image Generation

    —

    by

    Fraser
    in

    Black Forest Labs has unveiled Flux, touted as the most advanced text-to-image model to date. The company aims to establish itself as the industry benchmark for generative media, leveraging the expertise of key developers behind Stable Diffusion technology. Model Variants Flux is available in three distinct versions: Key Features Technical Specifications Performance Claims Black Forest…

  • Midjourney, “sref” and IP-Adapters

    —

    by

    Fraser
    in

    Midjourney’s style reference, has sent the AI community into a frenzy with its cryptic “sref” codes promising innovative styles. Yet, a closer inspection reveals the secret sauce isn’t so secret. The kicker is that you don’t even need Midjourney to use these styles – and that changes everything. Enter the IP-Adapter. This powerful algorithm takes…

  • ComfyUI in the Cloud with Fal

    —

    by

    Fraser
    in

    ComfyUI is a powerful tool for generating AI images, and with Fal.ai, we can use it in the browser. We only pay when we run the workflow, and not while we are preparing the workflow. When you create a new workflow in Fal.ai, you’ll be given a basic template with nodes to create an image…

  • Llama 3 Groq Tool Use Models

    —

    by

    Fraser
    in

    Groq has introduced an open source finetune of Llama 3 for tool use. It performs highly on the Berkeley Function Calling leaderboard, making it a top contender for tool use LLMs. It outperforms proprietary models like Claude Sonnet 3.5, GPT-4, and Gemini 1.5 Pro. Notably, it was trained solely on synthetic data. The model processes…

  • Lummi: Stunning Royalty-Free AI Stock Images

    Lummi: Stunning Royalty-Free AI Stock Images

    —

    by

    Fraser
    in

    Lummi is a free AI stock images website. They are royalty free images and they are stunning with vivid colours, high quality and interesting subjects. A wide selection of photos sorted by category. There are a range of search options such as by colour and orientation of image but also number of people in photo,…

  • AuraFlow – open source text-to-image

    AuraFlow – open source text-to-image

    —

    by

    Fraser
    in

    AuraFlow is a new open-source text-to-image model that has the potential to become the new backbone of image creation workflows such as with ComfyUI. Backed by Fal, it has released a base version 0.1 with the promise of more to come. It’s a 6B parameter model, which is quite large, and some are complaining that…

  • Project IDX in beta

    —

    by

    Fraser
    in

    Google has announced several major updates to Project IDX, its integrated AI-powered development environment: The announcements highlight Google’s focus on streamlining AI-assisted development workflows within Project IDX. https://idx.google.com

  • Real-World Examples of LLMs in Retail: From Personalised Recommendations to Employee Assistance

    —

    by

    Fraser
    in

    LLMs are an amazing new technology, but we are still in the early days of figuring out the best ways to put them to use. This article covers some real-world examples of how businesses are already leveraging Cohere’s LLM technology to improve their operations and customer experience, as described here. First, LLMs are being used…

  • 01.AI and their Yi models

    —

    by

    Fraser
    in

    Chinese startup, 01.AI, has launched their first proprietary model, Yi-Large this week, and is comparable to the performance of GPT-4. The AlpacaEval Leaderboard currently has the model ranked third. https://tatsu-lab.github.io/alpaca_eval The model is available with sign up through their API and offer Yi-Large and Yi-Large-RAG. Yi-1.5 is an upgraded version with stronger performance in coding,…

  • OpenVoice V2

    —

    by

    Fraser
    in

    OpenVoice V2 is a text-to-speech model capable of cloning any voice and speaking in multiple languages. Developed by MIT’s Computer Science & Artificial Intelligence Laboratory (CSAIL) and MyShell. The technology has the potential to be used to create customised digital voice interfaces, multilingual virtual assistants, and automatic dubbing in any language. It is fully open-source…

  • Langfuse 2.0

    —

    by

    Fraser
    in

    Langfuse is an open-source LLM engineering platform to help teams build production-grade LLM applications. Its core offering is traces for debugging, letting users explore complex logs in a visual user interface. Especially useful for RAG or agent systems. Quality scores can be attached to each trace through model-based evaluations, user feedback or manual labelling. This…

  • AI Leads a Service-as-Software Paradigm Shift

    —

    by

    Fraser
    in ,

    The article discusses the shift from Software-as-a-Service (SaaS) to Service-as-Software, driven by advancements in AI. The authors explain how AI, particularly agentic AI, can read and interpret content, generate output, set priorities, and direct tasks, much like a skilled human performing a service. This new paradigm has the potential to automate work previously done by…

  • Generative Agents: Simulating Human Behaviour

    —

    by

    Fraser
    in

    This research paper explores the concept of generative agents, which are AI-powered entities that mimic human behaviour in interactive environments. The authors investigate how LLMs can be used to design and control these agents. Generative Agents: Interactive Simulacra of Human Behavior The paper highlights four key techniques used in prompt engineering: The paper also discusses…

  • Agentic Design Patterns

    —

    by

    Fraser
    in

    This series of articles explores AI agent workflows, a new approach to leveraging large language models through iterative processes.  It covers, Reflection, Tool Use, Planning, Multi-Agent Collaboration DeepLearning.AI: Agentic Design Patterns

  • Speech to Text: Leaderboard & Comparison

    —

    by

    Fraser
    in

    Compare speech-to-text transcription models and API providers based on word error rate, speed, and price. View the analysis: https://artificialanalysis.ai/speech-to-text

  • Could AI Become the Next Big Thing in Market Research?

    —

    by

    Fraser
    in

    Research from April 2023 suggests that artificial intelligence, specifically large language models like GPT-3, could revolutionise the way companies understand consumer preferences. Traditionally, market research relies on expensive and time-consuming surveys and focus groups. However, this research explores using AI to simulate consumer choices and predict their willingness to pay for products and features. Research:…

  • Gorilla LLM

    —

    by

    Fraser
    in

    Gorilla LLM is an open-source project that connects LLMs to APIs, accurately returning API calls for a specific task provided by the user in natural language. It has been developed by researchers from UC Berkeley and Microsoft Research. Its primary objective is to improve the ability of LLMs to effectively utilise tools through API calls,…

  • Mustafa Suleyman | The Coming Wave of Artificial Intelligence

    —

    by

    Fraser
    in

    The Jordan Harbinger Show with Microsoft AI CEO and Inflection co-founder, Mustafa Suleyman, discussing the current state and future of AI. Suleyman, co-author of The Coming Wave: Technology, Power, and the 21st Century’s Greatest Dilemma, provides insights into how AI learns and predicts, its potential impact on various industries, and the ethical considerations of allowing…

  • Many-shot jailbreaking

    —

    by

    Fraser
    in

    Many-shot jailbreaking is a new technique that can cause large language models to override their safety constraints and provide harmful responses. It does this by including a very long sequence of faux dialogues in the prompt where an AI assistant answers dangerous requests. After enough of these examples, the model becomes more likely to also…

  • Superhuman’s AI-Powered Email

    —

    by

    Fraser
    in

    Superhuman has used OpenAI’s API to build a suite of email products aimed at reducing the time users spend on their inboxes. It includes features such as: The products have helped customers process their inboxes twice as fast as before, while also doubling the speed at which they can write emails. There AI strategy has…

  • Chatbot Leaderboard

    —

    by

    Fraser
    in

    LMSYS Chatbot Arena is an open platform that allows anyone to evaluate and compare the performance of different large language models (LLMs) through interactive conversations. The evaluations are crowdsourced, meaning they are conducted by a community of users who engage with the LLMs and provide feedback on their responses. This approach enables a collaborative and…