Hacker News Reader

Comment by enoughalready | Hacker News Reader

There’s a tedious, time consuming task, we all dread doing, but yet are compelled to do several times a month: interact with our bank’s UI.

Regardless of the convenience of mobile over web, it is still a multiple step process that requires several presses/clicks, screen loading icons, scrolling up and down the page, getting bombarded with calls to action, promotions, etc.

As we gaze into the future of digital banking, it’s clear that the next logical step is to incorporate functionalities such as trading stocks, dabbling in cryptocurrencies like Bitcoin, and even applying for loans, all within the same natural language interface.

View on HN

Please no. Prompt engineering is even more cumbersome and error-prone.

Some people will talk it into doing something it shouldn't. The bank will get in trouble because it seduced you into making an unwise investment.

Yeah, that was my first thought. Adding a layer of UI with documented unintended behavior is a liability landmine.

A chatbot for banking wouldn't necessarily be a bad idea but it's behavior would have to tightly controlled. On the other hand, that might not be hard to do because the domain that it needs to work over is very small, it's like the old Expert systems of the 1980s.

Indeed -- why would it even need to be a GPT system?

But I personally do not want a chat interface. I find communicating with a computer via natural language to be more cumbersome than using a decent traditional interface. Perhaps some banks have worse interfaces than others, but I'm happy with what I have now.

If others want a chat interface, then great! Hopefully it can be an addition rather than a replacement.

I think it'll likely end up being voice commands, and typing is just an interim solution. For instance, the ChatGPT iOS app which was just released has voice to text.

It would work no different from the current apps on the backend; (chat)gpt will fill a typed object, for instance a Transaction, software will validate this object as it does now (you can fill an object in an illegal way now and send it to the api; no difference), show it in the chat window as a strict UI to ask you if you are sure and then execute. The gpt part is only for you, not for the backend.

So people can try to make it do whatever; the banking backend logic won’t execute it, so it doesn’t matter. The user will just be wasting their own time.

Until of course neo-neo-web3-gpt-banks pop up that are ‘built up from the ground with ai’. But those will fail regulatory before they launch anyway.

While not perfect at the moment, huge strides are being made. There's a noticeable improvement in ChatGPT-4's ability to understand intent. Asking to "Pay 300 towards my credit card from my checking account" isn't a very complex ask, and is unlikely to be misunderstood. It's also much simpler than navigating through 6 traditional UI screens.

Confirmations will obviously be a requirement, just like they are in traditional UIs.

Reply on news.ycombinator.com