COT: Generative Methods for Goal-Directed Planning in Narrow Domains

Hyper-personalization for #wellness management requires multimodal support, including goal-directed planning for behavioral changes. The Uplift Array AI and product teams on a data model and AI architectures that support generative ML for planning. Generative methods for constructing action plans for various domains of wellness require an approach that differs from querying large language models for two significant reasons. First, the expertise and symbolic reasoning necessary to build personalized wellness programs must be prioritized. We have experts and local knowledge bases with specialized information. 

Second, frozen LLMs are currently not optimized for generating goal-directed plans, especially for narrow domains. This paper's chain-of-thought (#COT) prompt methods support mathematical and common sense reasoning. Adding a local knowledge base tokenized and vectorized produces finetuned prompts and promising results in narrow domains and is a viable option for generative pipelines.   

#LLM #largelanguagemodels #prompt #promptengineering #symbolicreasoning #chainofthought #AI #ML #generativeML #GPT #LaMDA #PaLM

https://arxiv.org/pdf/2201.11903.pdf  

AI tools: June 2023

 

Organizational AI Govenrance

 The paper "Putting AI Ethics into Practice: The Hourglass Model of Organizational AI Governance" from February 2023 presents a new framework for organizational AI governance that is designed to assist organizations in translating ethical AI principles into practice and aligning their AI systems and processes with the upcoming European AI Act.

https://arxiv.org/ftp/arxiv/papers/2206/2206.00335.pdf

Generative Landscape: June 2023

 

Query Augmentation for LLMs

 Generative Question Answering with #LLMs and (local) Knowledge Bases

Architecting an AI platform with LLMs is exciting for the research community and enterprise AI designers. If we can augment a general model with local knowledge for generative question-answering #GQA we can improve general LLMs. At Uplift Array, we infuse LLMs with custom knowledge from our wellness and behavioral science experts. 

Kader Sakkaria and the team are excited about this architecture's transformational impact. Retrieval augmentation with #LangChain libraries is a robust solution to many well known #LLM issues. LLMs have several output quality problems at this time and possibly in the foreseeable future. The issue of model focus is hard to handle. Predicting if the model "saw" relevant data sets is challenging. This issue can be significantly addressed by chaining local sources of knowledge, "stuffing" queries, and optimizing with different chain types. The vectorization and indexation of #OpenAI embeddings often need significant optimization as well. However, it enables natural language retrieval access, and keeps the vectors in the enterprise environment. 

There is a lot of potential for generative AI to help students, coaches, and counselors to increase academic retention. #UpliftArray #ai #machinelearning #deeplearning #deeplearningai #queryoptimization #mapreduce #gpt #chatgpt #huggingface

https://arxiv.org/pdf/2305.14283.pdf

GPT-4 for FREE

 1️⃣ Perplexity AI: Delve into the realm of GPT-4 by logging in with your Google account. On their app, head to the Search page, and watch the magic unfold as you ask any question. 🪄

2️⃣ Merlin: Empower your web browsing experience with this Chrome ChatGPT extension. From summarizing texts to crafting professional emails, Merlin has got your back on any website you visit 📚

3️⃣ WNR AI: Accessing GPT-4 has never been easier! Join the app and create a new template to unlock the full potential of this remarkable AI model. Prepare to be amazed! 🌟

4️⃣ Poe: Quora's AI app is your gateway to an array of AI models, including the incredible GPT-4. Sign up, navigate to the GPT-4 tab, and unleash the power of intelligent responses to your prompts. 📝

5️⃣ Bing AI: Ready to embark on an AI-powered search adventure? Bing's alternative to Google is fueled by GPT-4, offering you a unique and insightful experience. Explore their website and start typing to witness the AI magic unfold! ✨

Tiny ML

 On-device machine learning and "edge intelligence" or "tiny machine learning" (#TinyML) architectures are the key topics of this book. Focused and individualized modeling on limited, fixed hardware is often possible through the "lite" version of machine learning libraries like #TensorFlow or #PyTorch. 

The authors provide an excellent overview of application design constraints, software engineering methods, and areas of application. Machine learning modeling and training procedures are discussed in combination with the distributed nature of the applications, network synchronization, and multi-client efficiency considerations. The chapter on neural networks includes a valuable discussion on tiny transformers, #hyperparameter architecture, and #GPU platform implementation. 

#machinelearning #architecture #neuralnetworks #ai #edgecomputing #edgeai #tinyml #tinytransformer

Why The Metaverse?

The metaverse industry will reach ~$650 billion by 2030

 https://www.fastcompany.com/90901722/every-brand-needs-a-metaverse-strategy-heres-why

Data Storage Architectures

 

General Electric

 General Electric and Jack Welch succession was probably one of the most in-depth analyzed corporate America stories. Power Failure by William D. Cohan is one of the most in-depth studies of GE as a business, leadership transition and the aftermath. The "Neutron Jack's" stellar success in building GE takes about half of the book. We learn a lot of details about the reorganization of GE's business divisions, big mergers, and c-suite hi-stakes games. There is also an excellent chapter on the famous "envelope" containing the names of possible successors in case Jack Welch were to become incapacitated.  

Since my years in business school, I've read many books and journal pieces on GE's rise and decline. Power Failure is probably the most detailed resource.   

GPT: api access