My AI skeptic friends are all nuts

The author, a seasoned developer, argues that tech executives are making a mistake by mandating the adoption of Large Language Models (LLMs) for programming, but also believes that LLMs have the potential to revolutionize software development by automating tedious tasks and freeing up developers to focus on more complex and creative work. The author contends that LLMs can write a large fraction of tedious code, reduce the need for Googling, and don't get tired, making them a valuable tool for developers, but also emphasizes the importance of reading and understanding the code generated by LLMs.

Read more

Show HN: I build one absurd web project every month

The Absurd Website features a collection of unusual and humorous projects, including a waving cat that allegedly increases sales on e-commerce sites, a language learning tool, and various games and experiments. The website showcases a range of bizarre and entertaining projects, from a guide for aliens living on Earth to a slow delivery service and a dating site for eyes.

Read more

Show HN: Penny-1.7B Irish Penny Journal style transfer

Penny-1.7B is a 1.7 billion-parameter language model fine-tuned to emulate the 19th-century prose of the Irish Penny Journal, using Group Relative Policy Optimization (GRPO) to retain authentic Victorian-era diction while preserving general reasoning ability. The model is intended for creative writing, educational content, or stylistic pastiche in Victorian-era Irish English, but is not recommended for contemporary fact-based Q&A due to potential biases and archaic language.

Read more

AI makes the humanities more important, but also weirder

The integration of artificial intelligence (AI) into the humanities is a transformative and unavoidable development, with AI language models already being used for tasks such as language translation, data mining, and translation of archaic languages. Humanistic skills, such as understanding the relationship between language and culture, are also becoming increasingly important in AI research itself, with engineers needing to consider these topics to prevent their systems from breaking down.

Read more

How can AI researchers save energy? By going backward

Reversible computing, which involves running programs backward as easily as forward without deleting data, may soon power AI and save energy, as researchers like Michael Frank and Christof Teuscher work to develop this decades-old concept into a practical reality. By avoiding the energy loss associated with traditional computing, reversible computing could potentially improve power efficiency by orders of magnitude, keeping computational progress going as traditional computing hits fundamental physical limitations.

Read more

AI Persona Groupthink Makes Group Talk More Realistic

Casevo is a multi-agent simulation framework that utilizes large language models to simulate complex social phenomena and decision-making processes, enabling dynamic social modeling for scenarios such as social network analysis and behavior prediction. The framework has been demonstrated to facilitate realistic agent interactions, improving the quality of social phenomena simulation, and is available as an open-source code repository.

ArXiv

WINA: Weight informed Neuron activation for accelerating LLM inference

The growing computational demands of large language models require efficient inference and activation strategies, with training-free sparse activation methods offering a promising solution. A novel framework called WINA (Weight Informed Neuron Activation) has been proposed, which outperforms state-of-the-art methods by up to 2.94% in average performance at the same sparsity levels, offering a new performance frontier for training-free sparse activation in LLM inference.

ArXiv

Feeding AI personas media diets improves prediction

Public opinion can be measured and predicted using language models adapted to various media content, which can emulate the opinions of subpopulations that have consumed specific media. This approach has been validated through studies on COVID-19 and consumer confidence, showing it to be predictive of human judgments, robust, and more accurate for those who closely follow media, with potential applications in supplementing polls and forecasting public opinion.

ArXiv

AI Persona Opinion Surveying in Healthcare

Researchers compared survey data on healthcare decision-making with simulated responses from generative agents, finding that some large language models (LLMs) failed to reflect realistic decision-making, while others, like Llama 3, captured certain variations more accurately but introduced biases. The study highlights the potential of generative agents for behavioural research, but also underscores the risks of bias from both LLMs and prompting strategies, emphasizing the need for careful consideration in their use.

ArXiv

TradeExpert, a trading framework that employs Mixture of Expert LLMs

TradeExpert is a novel framework that utilizes a mix of expert Large Language Models (LLMs) to analyze diverse financial data sources, synthesizing insights to make predictions or decisions for quantitative trading. The framework demonstrates superior performance across various trading scenarios, and its effectiveness is evaluated using a large-scale financial dataset, showcasing its potential for stock movement prediction and quantitative stock trading.

ArXiv

Show HN: I built an AI Agent that uses the iPhone

PhoneAgent is an iPhone app that uses OpenAI models to interact with the phone, allowing users to perform tasks across multiple apps using text or voice commands. The app can tap, swipe, scroll, type, and open apps, and even listen for prompts in the background, but it has some limitations, such as struggling with keyboard input and animations, and may not always complete long-running tasks.

Repo

OCode is a sophisticated terminal-native AI coding assistant

OCode is a terminal-native AI coding assistant that provides deep codebase intelligence and autonomous task execution, built to work seamlessly with local Ollama models. It offers a range of features, including code generation and modification, project understanding, development automation, data processing, and system operations, and can be installed using a one-line installation script or manual installation with custom setup.

Repo

Show HN: Page Magic: Use AI to customize any web page

Page Magic is a Chrome extension that uses AI to customize the appearance of any web page, utilizing the Anthropic API, for which users must provide their own API key. To use Page Magic, users must install the extension, configure their settings, and then can apply changes to web pages by telling the extension what they want to change and clicking "Apply".

Repo

Show HN: Open-Source Terminal/SSH Automation Framework for AI Agents and LLMs

Termitty is a Python framework for terminal and SSH automation, inspired by Selenium WebDriver, that provides an intuitive API for automating command-line interfaces and interacting with terminals programmatically. It offers features such as terminal emulation, session recording, smart waiting, and parallel execution, making it a powerful tool for DevOps automation, testing, monitoring, and other use cases.

Repo

Show HN: Clai – Unixlike vendor agnostic LLM context feeder

clai is a command-line artificial intelligence tool that integrates AI models from multiple vendors, allowing users to generate images, text, summarize content, and chat using native terminal functionality. It supports various vendors, including OpenAI, Anthropic, and Mistral, and allows for easy comparisons between models, as well as customizable profiles and conversation history.

Repo