GatewayRedbirds.com

I’m a little skeptical about that company database story. I’ll have to read it later. Pretty sure they must have said yes to something they definitely shouldn’t have.

ghostrunner wrote: ↑

April 29 26, 3:10 pm

I’m a little skeptical about that company database story. I’ll have to read it later. Pretty sure they must have said yes to something they definitely shouldn’t have.

I sent it to a few people on my team. Our Sr. DBA responded with:

AI pressed the button, but this foolish startup stood it up at the button and said ‘promise me you aren’t going to press it’ and just trusting it wouldn’t. I could delete an entire company’s data store in record time too if they gave me direct access to it. The real issue is that they stored their backups in the same volume they were backing up. If there’s a story here, it’s that this startup has the world’s worst DR scheme known to man. No one should trust a company that operates like this.

Also, are we supposed to feel sorry for the AI when it ‘apologizes’? If I write and run a script that does something it isn’t supposed to and I find a way to make it say sorry, that doesn’t relieve me of responsibility.

To be somewhat fair, the CEOs and people like Musk overpromising should take a cut of the blame. Rather than publicly cautioning against this kind of trust, he’s saying things like robots will do surgery in 5 years.

On the one hand, I agree that it's insane to give the AI access to the delete button with instructions not to press it, but if you want to get the most out of your AI tools, you need to give it access to do real stuff, and it's going to be tricky to define those boundaries purely technically. As humans, we have access to various catastrophic buttons that we could theoretically push (killing people, trashing property, etc) all the time, but it still overwhelmingly works to put people in situations where they have power to do those things and expect them not to. If you had to put human workers in a box that physically prevents them from doing bad stuff, they wouldn't be able to get much useful done either. With AI, it's very unclear that you can do this, but it seems we're going to take more and more risk trying anyway.

But there's degrees in how much you can or have to automate. You CAN tell it here's full access and go do what you need to do, which seems fine if you're telling it where to pull data from and what you're having it do/create isn't that important or a finished product. But if it's stuff like your company data, you could save a little less time and have it just walk you through doing it yourself. Even then be sure you understand what it's telling you to do, or have someone who does do that part. And you could tell it to ask you before making changes. Just seems very avoidable to me.

The problem I am seeing right now is that folks (organizations) are just jumping on anything AI ("yeah, we have to have this tool and do all of this stuff with it") without strategy or understanding or proper education/training throughout the org.

It reminds me of a 5 year old soccer team just swarming the soccer ball from place to place on the field without any strategy...

It is absolutely disorganized. Very reminiscent of the dawn of the internet era in the mid to late 90's. Everyone rushing to have an online presence, whether it made sense or not. I'm positive there is a forthcoming AI bubble/crash in the stock market as well. A lot of circular spending for build-out, but these companies are going to need to be making billions of dollars very soon.

The real twisted thing is that the AI crash induced recession will be the impetus for the heavy push to do cost savings by entirely replacing workers with AI instead of just using it as a productivity aid.

Arthur Dent wrote: ↑

April 30 26, 11:44 am

The real twisted thing is that the AI crash induced recession will be the impetus for the heavy push to do cost savings by entirely replacing workers with AI instead of just using it as a productivity aid.

This!!!

Fat_Bulldog wrote: ↑

April 30 26, 11:15 am

The problem I am seeing right now is that folks (organizations) are just jumping on anything AI ("yeah, we have to have this tool and do all of this stuff with it") without strategy or understanding or proper education/training throughout the org.

You absolutely nailed it Bulldog.