AI Policy For Credit Unions

Developed in partnership with leading credit unions

Download your free AI Policy now!

Download now!

You may unsubscribe from these communications at any time. By clicking download, you consent to allow Aviary to store and process the personal information submitted above to provide you the content requested.

Thanks for subscribing! A confirmation email will be sent to your inbox shortly. If you don't see it, please check your spam folder.
Oops! Something went wrong while submitting the form.

Edition 4: Open Source AI Soars

April 14 - 20: HUGE news on the most powerful "public" AI, How AI can turn 1 photo into a lifelike deepfake, and AI Safety's latest upgrade. Plus, learn how Morgan Stanley has already unlocked the value of GenAI. All this and more in Edition 4 of the AviaryAI Newsletter.

Welcome to the AviaryAI Newsletter!

Thanks for joining us as we explore the intersection of GenAI and finance with practical learnings and the latest relevant insights. Let’s get started.

This week you’ll learn:

  • How AI can turn 1 photo into a lifelike deepfake
  • Powerful AI just got a lot more accessible with Llama 3
  • How to use AI in excel
  • The leading tool for AI Safety
  • How Morgan Stanley is already unlocking the value of GenAI

Subscribe to the AviaryAI Newsletter here.

The coolest things we're watching and why you should care
Image: Microsoft Vasa-1 AI

Microsoft's VASA-1 Generates Hyper-Realistic Deepfakes From a Single image
The AI generates realistic facial expressions, head movements, and changes in gaze to match the audio. VASA-1 won’t be publicly available due to risks of misuse. See it in action here.
So what? As generative AI evolves, it's becoming easier to create misleading content that looks and sounds real. Credit unions should stay vigilant about potential fraud risks and educate members on spotting manipulated media.

The Next Generation of Open Source AI
Meta just released Llama 3, their latest open source large language model. It comes in a two versions that outperform leading AI's like ChatGPT-3.5 and Google's Gemini Pro 1.5. Read more on the AI making developers' heads spin here.
So what? Open source AI is evolving fast. As big banks pour millions into AI, Llama 3 levels the playing field for credit unions. Powerful open source models mean credit union partners (like us) can build highly specialized AIs without giving data, ownership, or control over to the OpenAIs and Googles of the world. But responsible use is key. Thankfully, Meta provides safety tools like Llama Guard 2 for filtering outputs. Read more on Llama Guard below.

Adobe is Bringing AI Video Generation to Creators
OpenAI’s video generation tool, Sora, will soon be available to Adobe Premiere Pro users. Sora broke headlines earlier this year with it’s impressive capabilities to generate very high-fidelity video from simple text prompts. Read the full story here.
So what? As AI video tools become more accessible through platforms like Adobe, it opens up new possibilities for creating engaging member content quickly and cost-effectively.


Boost your efficiency with these easy to follow templates 

Generate Excel Formulas in a Flash

How to use this prompt:

  1. Copy and paste the template below into a generative AI model (ChatGPT, Gemini, etc.)
  2. Replace the [Insert x] with the recommended content
  3. Do not edit the content within the "{}" brackets
  4. Hit “Generate” and review the output
  5. Ask the AI to make any changes you deem necessary
  6. Enjoy the time you’ve saved!

You are a Microsoft Excel expert for credit unions with deep knowledge of writing Excel formulas. Your task is to take a request written in plain language and provide the corresponding Excel formula to accomplish the task. After providing the formula, break down how the formula works step-by-step.

Request: [Insert Excel request written in plain language. EX: In column C, I have a list of dollar amounts. In column D, I want to calculate 15% of each amount in column C.]

Provide your output in the following format:

Excel Formula: {Provide the Excel formula that would accomplish the request}

Step-by-Step Explanation: {Provide a detailed, step-by-step explanation of how the Excel formula works and what each part does to accomplish the requested task. Use bullet points to clearly explain each step.}

Looking for real world use-cases of AI for your credit union? Use our free: AI Prompt Library For Credit Unions


One simplified GenAI concept per week to build your AI Acumen

AI Safety with Llama Guard

Image: Meta Llama Guard

The problem: AI systems don’t inherently understand ethics - they just learn patterns from data. In the past, keeping AI conversations safe was a major challenge and existing solutions have pretty extreme limitations:

  • Keyword blacklists: Blocks responses with banned words - easily circumvented and can miss harmful contexts
  • Manual review: Humans checking some AI responses - does not sccale and risks exposure to dangerous content
  • Limiting capabilities: Restricting what AI can talk about - can lead to poor user experience

One solution is Llama Guard, an LLM designed to safeguard AI conversations. This model was trained with human feedback on 14,000 diverse examples of safe and unsafe AI interactions.

It acts as an extremely fast and accurate “second pair of eyes” looking at every message, inputs and outputs, to catch inappropriate content an AI system might wrongly produce.

So what:

SAFETY: Financial institutions must be held to an extremely high standard when it comes to prioritizing consumer’s safety and security. Llama Guard is a major step in enabling institutions to set a higher standard for conversation AI with proactive safeguard for harmful content.

ADAPTABILITY: Other tools like it exist but many are proprietary. Llama Guard is open source and acts as a Responsible AI foundation which institutions can then customize to their specific needs and policies.


Fine Tuning

Most of the popular AI models like ChatGPT or Gemini are trained to be “General purpose” models. They’re good at a lot of different tasks but may need “extra training” for more specific outputs (i.e. a particular tone, style, or format). That extra training is possible through fine tuning. The aim of fine-tuning is to maintain the original capabilities of a base model (like ChatGPT or Mixtral) while adapting it to suit more specialized use cases.


The latest news at the intersection of GenAI and Finance

Morgan Stanley’s AI-Powered Financial Advice

Morgan Stanley has launched an internal AI assistant based on ChatGPT to help financial advisors and support staff quickly access and digest over 100,000 research reports. This "copilot" tool aims to boost efficiency and free up time by answering advisors’ questions for:

  • Investment recommendations: “What is our research organization’s view on Alphabet stock”
  • General business questions: “Who are the five major competitors to IBM”
  • Process questions: “How do I include an IRA in an irrevocable trust?”

This “AI-trained on my business knowledge” use-case is just the first of many to provide clear value to financial institutions. Other banks like NatWest, Square, and HSBC are already embracing generative AI to enhance virtual assistants and provide content creation tools for business customers. As a credit union executive, it's crucial to understand and plan for the transformative impact of generative AI. Those who harness its power will be better positioned to streamline operations, serve members, and compete in an increasingly tech-driven industry. The future of banking is conversational, personalized, and powered by AI – is your credit union ready?

Read the full story here.

About AviaryAI

AviaryAI is the next evolution of financial interaction. Enhance your team with proactive outbound voice agents to welcome new members, encourage credit card activations, and drive non-interest revenue.

Drive Revenue and Engagement
With AI Today