Blog #162: Machine Learning Review December 2023

Photo by Kelly Sikkema on Unsplash

[ED: There’s a bit of a mix of content here. On balance, it’s 3/5 propeller hats.]

Here’s my review of all of the interesting things that happened in machine intelligence in December 2023.

microagents This GitHub repo explores the concept of self-evolving, modular agents capable of self-editing their prompts and executing Python code. These agents are designed to automatically generate and improve responses to queries, with a focus on efficiently evolving and refining their approaches to various tasks, including executing Python code for specific functions. 1. #AIInnovation #SelfEvolvingAgents #PythonProgramming #MachineLearning #TechDevelopment

Microsoft Promptbase This is a comprehensive resource on prompt engineering (in the form of a GitHub repo), providing scripts, best practices, and techniques to optimise the performance of foundation models like GPT-4, with a focus on applying these strategies in various domains including medicine. #PromptEngineering #GPT4Optimisation #AIProgramming #MachineLearningResources #MicrosoftAI

What is Retrieval Augmented Generation? Retrieval Augmented Generation (RAG) is an advanced natural language processing technique that combines retrieval-based and generative AI models to create accurate, context-aware responses, enhancing applications like customer support, content generation, and market research. #RetrievalAugmentedGeneration #NLPInnovation #AIFusion #ContextAwareAI #RAGTechnology

The Wizard of AI: This is a 20-minute AI-generated video essay by Alan Warburton, commissioned by the ODI’s Data as Culture programme, explores the cultural, legal, aesthetic, and ethical impacts of generative AI on visual cultures and creative economies, highlighting its transformative effect on artistic and design practices. #AIInArt #GenerativeAI #CreativeTechnology #DigitalEthics #TheWizardOfAI

Is AI doomerism the new millennialism? The Singularity is Nigh! The article discusses the deep ideological divide in the AI community, contrasting the views of those who see AI as a path to a utopian future with those who fear its existential risks, reflecting a schism akin to religious fervour over the direction and impact of AI development. #AIFuture #TechnologyDebate #SingularityDiscussion #AIExistentialRisk #TechUtopiaVsDystopia

It turns out that Google’s GeminiAI demo was not quite as impressive as it looked: Google admits that a Gemini AI demo video was staged: The Engadget article reveals that Google admitted to staging parts of a demo video for its AI project Gemini, where the video editing sped up outputs, and the purported voice interaction with the AI was actually non-existent, raising questions about Gemini’s readiness and the authenticity of AI demonstrations. #GoogleGemini #AIDemoControversy #TechEthics #ArtificialIntelligence #TechTransparency

DeepMind’s Gemini is quite something. DeepMind’s new model represents a significant advancement in AI, capable of multimodal reasoning across various formats like text, images, video, and code, outperforming previous state-of-the-art models in a range of tasks, and comes in three sizes—Ultra, Pro, and Nano—to cater to different complexity levels and applications. #DeepMindGemini #MultimodalAI #AIInnovation #TechnologyAdvancement #AIforAll

Here are some of the more interesting Gemini developments:

Video and sight capabilities

Understanding the direction of light and plants

Multimodal understanding of audio

Helping a user with complex requests and showing some of the ‘thinking’ it is doing about what context it does/doesn’t have

Assessing the relevance of scientific papers and then extracting data from the papers

Marker converts PDF, EPUB, and MOBI to markdown.: This GitHub project by @VikParuchuri is a tool designed to convert PDF, EPUB, and MOBI files into markdown format efficiently and with high accuracy, utilising a pipeline of deep learning models for tasks like text extraction, OCR, and page layout detection. #PDFtoMarkdown #AIConversion #DeepLearningTools #TechInnovation #OpenSourceSoftware

Building AI Apps with Elixir: Charlie Holtz covers his experience at ElixirConf 2023, focusing on building AI applications using Elixir. He discusses three patterns for AI app development - the Magic AI Box, the Gen Server Generator, and Agents - and illustrates these concepts with practical examples, emphasising the synergy between Elixir’s features and AI functionalities. #ElixirConf2023 #AIDevelopment #ElixirProgramming #TechConference #InnovativeCoding

Originally published by M@ on Medium.