The Orca Pulse - Quarter 3 - 2023 Newsletter

The Orca Pulse - Quarter 3 - 2023 Newsletter

The trees are changing, and so are we!

We are excited to share our 2023 Quarter 3 news. Swiftly's new features and learn more about Small Language Models (SLM).

Recent News

P.A.M.

Here again, sharing our quarter-three news.

Welcome our Plan, Analyze, and Manage (PAM) bot. PAM is not only the methodology used for Swiftly; now, she is a bot.?


Explore with PAM

We are testing out a new feature in Swiftly. If your Product Management staff needs assistance determining how to create Epics, Features, or User Stories for a project, PAM is here to assist. We will be rolling out this feature in January 2024.



Learn more about Small Language Models (SLM)

Did you know ChatGPT has over 200 billion parameters?

In this article from Quanta Magazine. Read it here.

?

Large language models such as GPT-3.5 are known for their resource-intensive nature and the complexity of their training processes. To address these challenges, Microsoft researchers have embarked on a novel exploration: training smaller, more efficient language models using carefully curated datasets of stories. This approach offers a potential starting point for building large language models. This concept is particularly relevant for smaller organizations that may lack the resources to develop large-scale language models.

?

The article underscores the significance of commencing the neural network journey with compact language models, which can subsequently serve as foundational building blocks for constructing more expansive language models.

?

The key findings are:

  1. The research showed that even relatively small models, thousands of times smaller than large models like GPT-3.5, could provide valuable results as LLM. This suggests that downsizing models might be a viable strategy.
  2. The study used a data set called TinyStories, containing around 2 million stories, to train models ranging from 1 million to 30 million parameters. Larger models performed better, with the 28-million-parameter model generating a coherent story.
  3. It has been known that OpenAI's GPT-2, a 1.5-billion-parameter model, performed poorly on the same task, indicating that training smaller models might be more effective than relying on very large models.

In summary, the research presents an innovative approach to training smaller language models using a niche data set like children's stories, showing promise in efficiency and effectiveness. It challenges the notion that only very large models can perform complex language tasks effectively.



Recent Events

Event Images



Orca Intelligence was delighted to attend several prestigious events this year. We were present at the "Party on The Pier 2023 Gala" hosted by the Potomac Riverkeeper Network , the Association of the United States Army - AUSA Conference held at the Washington DC convention center, the "Small Business Awards Gala" organized by the U.S. Chamber of Commerce , and the National Minority Supplier Development Council (NMSDC) National Conference. At the latter, we showcased our cutting-edge Virtual Reality Experience.


DC Metro Reducing Regional GHG Emissions

Every trip with Washington Metropolitan Area Transit Authority (WMATA) Metro and not a car reduces the region's carbon footprint. On average, traveling by Metro results in 46% less GHG emissions per mile than driving in a single-occupancy vehicle. Calculate your CO 2?savings with Metro's?Trip Planner.


Need a refresher on our service? Check out our promo video below.







Our services

We are an 8(a), Economically Disadvantaged Woman-Owned Small Business (EDWOSB) and Minority Business certified by the United States Women Chamber of Commerce and National Minority Supplier Development Council, respectively. Our services include:

Agile Project Management, Human Center Design and Engineering, Business Systems Analysis and Design, and Solutions Architecture through the use of Artificial Intelligence

Primary NAICS Code: 541512 COMPUTER SYSTEMS DESIGN SERVICES

DUNS NUMBER: 078786922

CAGE CODE: 73PW7

OTHER NAICS:

541512 ? ?Computer Systems Design Services 541511 ? ?Custom Computer Programming Services 541519 ? ?Other Computer-Related Services 541611 ? ?Administrative Management & General Management Consulting Services 541690 ? ?Other Scientific & Technical Consulting Services 541715 ? ?Computer and related Hardware research And development laboratories Or services 541713? ? Research and Development in Nanotechnology

Orca Intelligence, 8720 Georgia Ave, Suite 302, Silver Spring, Maryland 20910, United States, (833) 921-ORCA

要查看或添加评论,请登录

Orca Intelligence的更多文章

社区洞察

其他会员也浏览了