Datamine Discovery - eDiscovery Review Software & Project Management. A KEY Discovery Affiliate’s Post

In this episode, we discuss Observational Scaling Laws and the Predictability of Language Model Performance by Yangjun Ruan, Chris J. Maddison, Tatsunori Hashimoto. The paper introduces an observational approach to building scaling laws for language models by utilizing approximately 80 publicly available models, bypassing the need for extensive model training. It discovers that despite variations in model efficiencies, performance can be predicted using a generalized scaling law based on a low-dimensional capability space. This method demonstrates the predictability of complex scaling behaviors and the impact of interventions such as Chain-of-Thought and Self-Consistency.

Imagine you need a RAG, imagine you still do not know how to choose among many vectorDBs and how to set one up..then start choosing a model fine-tuned to answer domain-related questions (but still is good enough on many other tasks)..this is already enough for many cases even without a vectorDB. Nevertheless, once you add one then you have a great boost..InfoSynth is our attempt to create an llm (for italian language) that can synthesise information and answer questions by its nature and can be further boosted if you attach any vectorDB. It is designed to take questions and context as input and provide pertinent responses, or to indicate if a response cannot be deduced from the given context. You find the details in the link-->https://lnkd.in/dFVACEnJ #moxoff #bezucchetti #generativeai #rag #llm #huggingface

In this episode we discuss E3 TTS: Easy End-to-End Diffusion-based Text to Speech by Yuan Gao, Nobuyuki Morioka, Yu Zhang, Nanxin Chen. The paper introduces Easy End-to-End Diffusion-based Text to Speech (E3 TTS), an innovative text-to-speech model that converts text to audio using a diffusion process without the need for intermediate representations or alignment information. E3 TTS functions through iterative refinement directly from plain text to audio waveform, supporting flexible latent structures that enable zero-shot tasks like editing. The model has been tested and offers high-fidelity audio generation, comparable to the performance of advanced neural TTS systems, with samples available online for evaluation.

In two studies, we show that 'Language regions' and 'the language network' are the result of indiscriminate averaging, leaving only input regions and connectivity hubs coordinating variable whole #brain distributions of activity: https://lnkd.in/eF9E_Q-J A thread [ 🧠 🧵]

Check this out… This study is challenging the idea that different parts of the brain are responsible for specific tasks, like language processing or memory. Instead, the researchers (one of whom is my stepbrother) are suggesting that the whole brain is involved in all of these tasks, and that the idea of "language regions" or "memory regions" is an oversimplification. They're using data from brain scans to show that different parts of the brain are more or less active during different tasks, but that no one part of the brain is solely responsible for any one task.

A warning about LLM and AI from German government: https://lnkd.in/gMDaGDij Yes. Regulation and validation/verification (V&V) are critical. For example, nuclear tests were to V&V atomic bomb models. An UN newly released report said that AI ONLY benefits small amount of states, companies and individuals, i.e., some few humans are making big money by using AI to harm many people. AI is based on math models, and models must be V&V before we can trust them. Using math models to simulate a physical process started from the Manhattan Project in WWII for nuclear bomb design. Then, in 1960s, it came the C4ISR, which is today's AI, whose original missions were just breaching enemy's security, dis-/mis-information, cognitive manipulation, cheat, surveillance, detection for kill, etc.. In our reality, AI causes expensive electricity bills to a C-level friend of ours. 😔 😥 😿 AI can do many things, but NOT everything, at least NOT what we are doing, by using our intellectual property (IP), a copyrighted multilingual metadata. Without metadata, NO data can be found/retrieved, even by AI. https://lnkd.in/g-aJFnXR

On YouTube🎙️ Dive deep into the Large Language Model Evaluation with Gauthier Guinet, on our latest LLM Daily episode. 🌐 Elevate your understanding of LLM Evaluation by tuning in for insightful, scientific discussions. #LLMEvaluation #AIResearch #ScientificInsights 🚀 Youtube: https://lnkd.in/gQa5funr

Recent paper on work toward machine translation of N'ko:

With wisdom state acquisition and access to the proper semantic structure - we become able (gain the capacity) to communicate effectively.

Synthetic Work Issue #58 is out. In this issue, Alessandro Perilli 🇺🇦 challenges the idea that large language models must be perfect to deliver economic value, and considers the business dynamics of a world where imperfect LLMs deliver free world-class expertise in most disciplines. Read it here: https://synthetic.work