Burak Aksar, PhD

This is a good article on using LLMs in the real world. It covers many aspects that we learned at Nurish over the last few months and more. Using LLMs is an inexact science, but has its own needs in processes and tools. While most of the attention has been on the LLMS, their capabilities, and RAG, there are mundane issues that need to be taken care of, issues where the solutions for pre-LLM development are not that well-suited. These include version management, integration, testing, data management, quality assessment, and more. This article covers many of those practical aspects. Eugene Yan, Bryan Bischof, Charles Frye & more https://applied-llms.org/

9

"It is anticipated that there will be 75% fewer employee-facing application software companies and 50% fewer categories." from Brett Queener's newsletter. The proliferation of software solutions has led to an overwhelming number of tools for employees, reducing actual productivity and mastery of their craft. Adding on top terrible UI/UX in B2B SaaS leaving users to navigate interfaces that suck. AI can simplify user interactions, understand user needs, and perform tasks autonomously, shifting the cognitive load from the user to the software. The application software market is poised for a paradigm shift driven by AI, leading to fewer but more powerful software solutions that significantly enhance user productivity. Companies that embrace this transformation will set new benchmarks for usability and efficiency, ultimately redefining what it means to be an application software company.

2

We recently announced our newest product updates at LILT AI Day. The highlights: * AI Review -- automated quality scoring based on custom MQM templates. * V3.5 Models -- 100x performance / cost improvement relative to GPT 4. * No code workflow builder -- orchestrate models, data, and people without having to write code. Link in the comments below! #ai #GenerativeAI #EnterpriseAI #Localization #Globalization #l10n #i18n

54

1 Comment

Our paper, "QGen: On the Ability to Generalize in Quantization Aware Training," (https://lnkd.in/gXXvS_9Q) is now available on Arxiv. This research delved into the generalization capabilities of quantized models, a relatively unexplored domain, examining them through the lens of sharpness in the loss landscape during optimization. I found the experiments and mathematical proofs particularly fulfilling. We're hopeful for acceptance at the conference to which it's been submitted. 🤞

47

I had a great time at ICLR 2024 last week presenting https://lnkd.in/ebr2JJZR. The future of AI hinges on compound systems of models that plan, reason, reflect, and coordinate with each other. It was exciting to see early signs of this future from top folks in the field -- here are some of my favorite papers from the conference, check them out! - A Real-World WebAgent with Planning, Long Context Understanding, and Program Synthesis (https://lnkd.in/eTvm_Mir): An LLM agent for web browsing with a custom multi-LLM workflow for high level planning and program synthesis. - Hybrid LLM: Cost-Efficient and Quality-Aware Query Routing (https://lnkd.in/ehsdybEM): A probabilistic model routing framework for optimizing cost-performance tradeoffs. - MetaGPT: Meta Programming for a Multi-Agent Collaborative Framework (https://lnkd.in/eJ_D9_BW): A multi-agent framework for software engineering tasks. - RECOMP: Improving Retrieval-Augmented LMs with Compression and Selective Augmentation (https://lnkd.in/eR7CZ84Y): A new LLM-based mechanism for pre-compressing context sent to LLMs, akin to re-ranking. - RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval (https://lnkd.in/eCEV49_H): A hierarchical indexing strategy for data retrieval. I’m a huge fan of hierarchical representations of unstructured data, as we at Arcus write more about in our blog here: https://lnkd.in/e5USx5am.

115

1 Comment