Tag: machine learning

If Your AI Is Only For Chatting, You’re Doing It Wrong

August 22, 2017 / Joe Winn / 1 Comment

Updates:

June 2023: There are many issues with these systems and the speed at which they are being adopted, without care for the negative consequences. However, there’s definite value for your credit union. So long as you approach it with reason.

February 2023: With the release of OpenAI’s ChatGPT-3.5, this has all been both turned on its head, and also exactly the same. ChatGPT is good. Coherent and, frankly, perfect interview good. But only for appropriate uses. It does have a habit of “hallucinating”.

I’ve spent many hours working with it as well, and the results are stunning. It has helped me plan out a year’s worth of blog articles, then draft each of them (still up to me to format and post!). I tossed around some marketing ideas and got its hot takes, which were all valuable.

Some of the more fun and wow-worthy interactions will be seen over time. Hey, can’t spill all the beans at once!

Suffice it to say, this is the chatbot that will make them relevant, meaningful, and insanely useful. I’m not smart enough to explain how big a deal this is, so I asked ChatGPT to help me out. Here’s what it had to say:

ChatGPT is the future of AI customer service for credit unions. With its advanced language model, it can hold human-like conversations on any topic. Think of it as a smarter, more efficient version of Jarvis from Iron Man. Empower your credit union with the power of AI and give your members an unparalleled banking experience.
ChatGPT-3

Let this definitely-not-AI-but-people-will-call-it-that-anyway help you internally. Use a restricted form to manage questions. Then give members a chance to connect to a person, through whatever medium they prefer.

All that said, it doesn’t mean you should dedicate all communications, article writing, email drafting, and website updates to ChatGPT (or the bazillion related systems emerging).

Focus on the human touch. Embrace your mission. And remember: People. Helping. People. (And “AI” can help!)

Original article:

“Ask Our Friendly AI!” Your credit union’s website is excitedly promoting their new chatbot, there to answer questions 24/7. “Cool, so how can it help me save money or time?”

Whether they admit it or not, that’s what your members will be thinking.

In some cases, such tech is fielding member requests without burdening traditional staff time. And their resolution rates can be similar to human representatives.

What are you waiting for? Get Siri, Alexa, Cortana, and friends to every CU! (HAL is not welcome, sorry)

Interacting in a Financial World

If only it were that simple. “AI” support agents are uniquely programmed to understand financial world terminology. Plus, computers don’t excel at interacting like a person, since we learn and process the world in a different way.

As of the end of 2022, it looks like that day is finally here. With ChatGPT, you can have full conversations on topics far and wide. And it can even “hallucinate” other environments, opening the door to incredible education opportunities.

What may be even further off is having all systems talk to each other in the background. (ChatGPT is deliberately not connected to the internet, but the possibility is certainly there for closed-network interactions.)

Here are some ideas I had in 2017, when the “state-of-the-art” for AI was far less impressive. Amazingly, my insights on an increasing rate of change…of change are coming to fruition. So, back to 2017….

Imagine being able to ask Siri (remember that post?) to transfer money from one account to another and explain the tax implications of your specific IRA contributions.

Then find out the score for your favorite team’s last game.

But we’re not at that point…yet. And look who spoke too soon…we’re actually getting awfully close.

AI for Serving Your Members

Readers know my passion (that’s 3 links!) for the “AI Revolution”. With its arrival, a lot of ideas are being thrown around on best use. Right now, the most common answer is: Everywhere!!!

Patience, my young Padawan. A fancy chatbot might seem like the natural first step, but let’s look at it from a member benefit perspective. If they have a question, they don’t care who/what responds.

Your members want a quick and accurate answer. If your team is currently able to keep members served quickly and effectively (through any medium they contact), then this may not be a fit for you at this time.

Unless you have unlimited resources, in which case, yes, do all of this at once. Just make sure you have top-notch project management to ensure the focus is always on the unified credit union goals.

Big Data

For the rest of us, the AI which makes the most sense, if less “sexy”, is the Big Data side of AI: Machine learning. Here, you have solutions that can do things quickly which would take enormous staff effort.

Machine Learning Innovations

Fractal — *This is your data. Your AI can find the patterns.*

For example, you can analyze a member’s credit (beyond the report) and offer a rapid loan decision with a high rate of accuracy.

You can implement systems to monitor patterns in spending to identify fraud the moment it occurs, saving the institution money and the member frustration.

Machine learning is also enabling security of the body, biometrics. That’s the fingerprint sensor on your phone. Or FaceID on your newer iPhone. It’s also the Hello feature on Windows 10 computers.

Retina scanners combined with other biometrics, a security card, and “something you know” composes the “top level” of security at large financial institutions (and government entities).

Speed. Savings. Security.

Three great reasons to implement aspects of AI in your credit union. A recent post about this topic ended with a wonderful quote:

“When a bank…effectively uses AI, they run more efficiently and are able to connect more effectively with a segment of the population that will never be replaced by machines: their customers.” – Mohit Joshi, Innovations in FinTech

Ok, ok. I’ve given you way too much to consider. AI, Big Data, machine learning, algorithmic analysis…yeah, I get it. Overwhelming when you just want to know, “can this stuff help my credit union?”

So, I had a realization right after writing this post. Remember that series I did about tech in the financial industry? As part of it, I mentioned that financial institutions are at risk of becoming “dumb banks” in the same way that ISPs are “dumb pipes”.

What I mean is that you are at risk for simply becoming the corridor for other companies’ information. You hold the money, but your members use other services to move, spend, invest, even check on their funds.

The same is the case here with AI.

There will always be a place for information as you manage it now: Raw account balances aren’t going anywhere. But that’s “dumb data”. The future is in “smart data”.

Where your credit union and members can find patterns in spending, opportunities in lending, and personalized recommendations for minimizing debt (or maximizing wealth).

How will you become the “smart data” of the future?

Image credits: Mohamed_hassan and Seanbatty from Pixabay.

Deeper Connections – Part 3

November 29, 2016 / Joe Winn / 1 Comment

Originally published on CUInsight.com

It’s now been two posts on the “Two Peoples” topic. What have we learned?

There is a technology divide between those who are immersed and those who live more traditionally
The way people interact with the world and each other is evolving
Tech itself is changing beneath our feet (and around us in “the cloud”)

For nearly all of us, the idea of technology in our lives revolves around things mentioned in the first post. What phone you use, the devices you connect to it, even your “smart home” accessories. It’s primarily the hardware, and, as we learned in the second post, the services you use on them. So, that’s the future: Ever-improving devices with more interesting apps.

Not quite. There’s an area of growth which seems so far-fetched that we discount it as “distant future”. But it’s here today.

Artificial intelligence.

We aren’t talking the adorable bots from *batteries not included, nor are we concerned with T-1000 units “terminating” their target. AI (or more accurately, machine learning) today is in some ways like a traffic light. It does one thing. However, unlike a traffic light, it’s always improving how it does that one thing. And you use these 1st generation AI systems everyday.

Your Facebook feed is a form of machine learning. It tailors posts shown based on what it learns you enjoy. The more you use it, the better it gets. Your iPhone keyboard is the same way. It actually adjusts the size and location of each key by tiny amounts as it learns where your fingers press most often. It even figures out how you talk to better predict the next word you’re going to write (and it knows whether you’re typing in a social or professional manner).

Search Google for the image of a cat. You just asked their machine learning system. Their latest endeavor is a platform called Deep Dream (caution: highly geek). Besides trippy imagery, it shows how a computer actually learns. Fascinating, as Data would say.

Interesting, but, once again…why? The first two parts related to what technology you use knowingly. Those spawn the interest in visible tech. Modern app platforms. Game-like member engagement. All great, and important. But it’s the machine learning which will offer the “just what I wanted” capability of future financial services.

Computers are smartest with tons of data. Big Data, you could say. With it, a learning system can figure out when a member is at risk of overdrawing their account or might be in the market for a car. How thrilled would they be if you could suggest adding overdraft protection an hour before they make a costly transaction? Or notify them of a great auto rate and car research system the day after their vehicle has engine troubles?

Unfortunately, I’m not smart enough to even offer the breadth of examples this future will offer. But I’ve read a lot from those who are. A recent CU Broadcast interview dove headfirst into the data side (without mentioning the AI part). Coastal CU does data analysis for member habits. Affinity CU just expressed interest in the concept after a short chat. And that’s just in the course of a few days. Much like winter, change is coming.

So how can you stay ahead of your competition while providing historically-great member service? Focus on what you do best, and find partners which excel in their complementary areas. By working together, perhaps we stand a chance against our computer overlords.

I mean, serve your members in new and exciting ways.

For far too much detail on Artificial Intelligence and just how close we are to unbelievable changes, read this amazing post by Wait But Why’s Tim Urban.

Joe Winn Kicking in Rocky Mountain National Park

A Geek’s Thoughts on the Election

November 9, 2016 / Joe Winn / 0 Comments

Today, this country made a decision, but it isn’t a decision in which anyone should be proud. Whether you voted for your candidate or against the other one, or a mixture of both (or went third-party), the conclusion is a troubling result. We’ve said that experience is irrelevant. We’ve said that decorum is unnecessary. We’ve rewarded hateful speech and actions, probably because we share the fears from which they derive. We’ve legitimized a lot more, but it’s not even worth diving into it here.

In my research of machine learning (artificial intelligence), a common theme arose around the idea of logarithmic change. This means that as change (in this case, computer performance and “smarts”) occurs, it occurs at a faster rate than prior. Not only does a system get smarter as it “grows”, but it gets smarter more quickly. Think of it like a car which goes 0-60 in 4 seconds, 60-120 in 3, 120-180 in 1, 180-240 in 0.1, and finally 240-300 in 0.0001 seconds. Once it’s going 500, can you even process that type of acceleration? More importantly, how would you describe the velocity increase at 1,000? If you’re struggling to wrap your brain around it, that’s ok. You’re not alone. We perceive the world linearly, and this is at the core of many challenges.

Our world has been in the midst of this increasing rate of change for all of its history. However, only within the past decade or so has it become so impactful on the average person’s life. Minorities are rapidly becoming the majority, social norms are shifting at an accelerated rate, and the divide between what our knowledge contains and what the average person knows (or even *could* know) is growing exponentially. You could probably describe the basic idea of how your VCR worked. How about your iPhone?

This is why the challenges of today (and tomorrow) are so difficult to reconcile. We think in a linear fashion: Last year was that, this year is such and such, so next year will be a derivative of those. Except this no longer applies. Change accelerated and next year will be something we can hardly imagine.

And neither candidate appeared to grasp this fundamental concept.

This election was an expression of deep-seated fear of the unknown (be it gay marriage, traditional gender roles breaking down, ethnic diversification on a majority scale, expanding capabilities of a surveillance state, and any number of other topics). What many always knew to be true simply isn’t anymore. Like being in an earthquake, people’s “bedrock” is cracking. Anxiety over what an ever-increasingly changing future will bring led Americans to make rash decisions all the way through the election process.

I don’t have any answers. I’m pretty sure our President-elect doesn’t, either. So we’re going to have to work together and figure out how we will move forward while navigating this wildly-accelerating car.

Image credit: Me, seeking inner balance and focus.

Tag: machine learning

If Your AI Is Only For Chatting, You’re Doing It Wrong

Original article:

Interacting in a Financial World

AI for Serving Your Members

Big Data

Machine Learning Innovations

Speed. Savings. Security.

Deeper Connections – Part 3

A Geek’s Thoughts on the Election

Geekify Your Inbox

Most Recent

Affiliations

Brief Bio

Follow Me

Powered by