Edition 2: It's Time to Phase Out Voice Authentication
![Edition 2: It's Time to Phase Out Voice Authentication](https://cdn.prod.website-files.com/65b0f7d963cdc6f21824f1ef/6613f44067310dd7f134f91c_Voice%20Auth%20Phase%20Out.png)
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:
- Plans for your Iphone's AI
- Why credit unions should avoid voice authentication
- How to find the right AI for your institution
- What credit unions should know about TransUnion's new AI platform
Subscribe to the AviaryAI Newsletter here.
GENERATIVE AI THIS WEEK
The coolest things we're watching and why you should care
![](https://cdn.prod.website-files.com/65b0f7d963cdc6f21824f1ef/660ed023a2fd0e69094ea246_425a76f8-927b-dd5f-20ed-5be8a13df271.png)
Google AI in your iPhone??
Bloomberg recently reported Apple is considering adding Google’s GenAI model, Gemini, to iOS 18. Read the full story here:. - Read the full story here.
So what? If the deal goes through, one of the most powerful LLMs could be on 2+ billion iPhones around the world. Since Gemini has access to the web, investing in SEO will be more important than ever.
Institutions May Need to Phase Out Voice Authentication Soon
OpenAI recently unveiled a new AI model that can clone any human voice with just 15 seconds of sample data. This model, called Voice Engine, is not yet available to the public. Check out the demo here and you might see why.
So what? Any banks relying on voice authentication will want to seriously consider new authentication systems sooner rather than later. If OpenAI has built it, others could be right behind.
GenAI + Analytics = Generative BI
AI has changed how we interact with data. All you need to do is request the insights you want. The system performs the analytical steps in the background and even presents the data it in the best possible format - See how GenerativeBI works here.
So what? With GenerativeBI, you don’t have to spend time writing DAX formulas, configuring graphs, or debugging queries. The insights to drive your decisions are just a conversation away.
GENERATIVE AI WORD OF THE DAY
Context Window
Consider a comic strip. If you only see one panel at a time, you might miss the punchline or misunderstand the story. But if you see several panels together, you get the whole joke. That's how context length works in language models – the more panels (words) it sees, the better it understands the whole story (sentence/conversation).
THE AI EXECUTIVE'S HANDBOOK
One simplified GenAI concept per week to build your AI Acumen
Finding The Best AI For Your Credit Union- AI Benchmarks Simplified
With so many options from ChatGPT, Claude, Gemini, or one of the 300,000+ open source models, finding the right AI for your institution can be overwhelming.
LLM Benchmarks are one piece of the puzzle. But understanding what these benchmarks tell you can be disorienting. Here are some things to keep in mind:
What are LLM benchmarks? Think of them as performance tests for LLMs. These benchmarks assess how well a particular model will handle tasks relevant to your use case.
How do they work? Different benchmarks throw different challenges at these models, measuring their accuracy, speed, and even creativity in handling different tasks.
Here’s an example:
![](https://cdn.prod.website-files.com/65b0f7d963cdc6f21824f1ef/660ecf57dcdd5d2bdb589567_24829e6f-03a7-4285-7856-beec1535d8eb.png)
The table above outlines the capabilities for Google’s LLM, Gemini 1.0 Ultra.
On the left you have the capability that’s being tested (Ex. “reasoning”), the name of the benchmark that’s testing that capability, and the benchmark’s descriptions of the tasks the model is tested on.
A Major Consideration
These benchmarks can be misleading. Some models reach their scores only through specific prompting techniques which can wildly impact performance. It’s difficult to compare two models unless you use the exact same prompting technique.
Finding the best model for your business
At AviaryAI, we’ve trained hundreds of models. In our experience, selecting the right one comes down to understanding your problem and experimenting. Look for models with best performance in benchmarks related to your desired use case and start testing.
GENERATIVE AI IN FINANCE
The latest news at the intersection of GenAI and Finance
TransUnion Launches New AI Platform
![](https://cdn.prod.website-files.com/65b0f7d963cdc6f21824f1ef/660ecf76fef0637c5ca42400_e89d592b-665b-1349-f378-0a27d43cfdd4.png)
Source: https://www.transunion.com/about-us/onetru
TransUnion has just launched their new AI-powered platform called OneTru. The platform touts its use of artificial intelligence will help banks and credit unions make better decisions about lending, marketing, and preventing fraud.
With OneTru, banks and credit unions can tap into various data sources, such as credit reports, consumer identities, and marketing information. To make the data navigation process easier, TransUnion is using large language models to enable analysts with the necessary permissions to generate reports and interact with private and publicly available data.
However, it's important to keep in mind that the Consumer Financial Protection Bureau’s (CFPB) guidelines restricts the use of GenerativeAI in credit decisioning. While TransUnion's OneTru platform seems promising, it's still too early for us here at AviaryAI to gauge how much value it will drive to financial institutions.
Using the platform? Respond and let us know what you think so we can share it with our readers!
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.