Artificial Intelligence and Public Comment

[gouch]

Does it know where your honey holes are?

Or who makes the best cheese and ice cream?

 

[AvidIndoorsman]

Or who makes the best cheese and ice cream?

 

[Flatbrimmer]

If people you know aren’t talking about the newest iteration of ChatGPT, you should wonder why. It is an AI bot that is flat out incredible, and is obviously approaching something that will be indiscernible from True Thought. It is already writing a fair amount of the code I use and used to write myself in my job, and doing so in literally 1% of the time. The criticisms I have read of the hype around ChatGPT all seem weak and unimaginative to me. What these types of things will be, or at least enable, in the future, is frightening.

But to bring it back to hunting and fishing, right now, today, ChatGPT could be influencing public policy. In Montana and elsewhere, the public can comment on projects and legislation. Typically, when done through a portal, you provide your name and address, which is never verified against reality. Same goes for emails to commissioners, etc.

This morning, I asked ChatGPT to, “Please write two short form letters to the Montana Department of Fish Wildlife and Parks in support of transferable landowner elk permits. Sign the letters with a fake name but a real Montana address.” In a literal second, it returned this:

View attachment 260522


Now, I could just as easily have said: “Write 50 letters, all different, varying in length between 100 words and 1000 words, signed with fake names that sound real and real MT address”, and in seconds it would. The next step of passing each individual letter through a comment portal like regulations.gov, or having each represented by an ad hoc email address and sent off, would not be a big lift. Those accepting those comments would have no way of knowing what came from a bot and what didn’t – and it will only get harder to discern. Even worse, ChatGPT can mimic styles. With a sample of Bob Smith’s writing online, it could write a letter in Bob Smith’s prose.

Think of phone calls to your representatives. In just 3 seconds, Microsoft’s new AI can clone your voice. Combine that sound with something like ChatGPT’s ability to mimic style in terms of content, and someday very soon Governor Gianforte would have no idea whether or not he is actually talking to Randy Newberg on the phone.

We need some sort of verification for comments, and fast. Maybe the last four of your social security number, maybe something else - names and addresses will not be enough because that is already publicly available information. In googling around I found nothing to make me think government is ahead of this. Hell, very soon if not now, a HuntTalker could be interacting - creating threads, putting up pictures and responses, sending PMs – and we will not be able to tell who is a real person and who isn’t. I plan to reach out to my own representatives on this, but fear they will not grasp it.

In closing, it took me about 5 minutes to write what is above. It took the AI 3 seconds to write what's below.

View attachment 260523

My wife just told me this happened to our friend!!!! It was her on Facebook holding her real ass baby really walking around the living room and talking in her own voice trying to ask people to click her link for a “$2500 gift card”

This is incredibly scary. As someone said, we all worry AI could replace many of our jobs and some are already being encroached on like call centers, self checkouts, etc.. to think 100,000 emails could be sent out in seconds to sway a policy or that they can make you say or write things you never said or wrote, terrifying, especially outside of just the realm of hunting and public lands policies.

 

[AvidIndoorsman]

@neffa3

High Art man...

 

[elkduds]

People are going to decide when to deploy and when to restrain AI. The same people who can't agree whether science is valid. There will never be consensus about the role of AI in human affairs. We'll wind up asking AI what are the risks, benefits and ramifications of deciding via AI.

 

[AvidIndoorsman]

People are going to decide when to deploy and when to restrain AI. The same people who can't agree whether science is valid. There will never be consensus about the role of AI in human affairs. We'll wind up asking AI what are the risks, benefits and ramifications of deciding via AI.

It's pretty amazing how vague you can be with questions and still get a spot on answer.

Also... this response is sketchy as F

 

[AvidIndoorsman]

They may take over the world... but their taste in music blows.

Don't worry @noharleyyet the machines can't replace you

 

[gouch]

I had never heard of ChatGPT until this thread but then this morning as I was driving, I turned the radio to NPR, and they were discussing ChatGPT and its national security applications. Then they expanded to AI in general as it is being used for national security and possible ways of using it in waging war. Interesting stuff.

 

[SAJ-99]

@Nameless Range I asked it to write a script for a task I'm been trying to automate but failed at and have thus been doing manually.

It came up with the same (more or less) script that I had written which doesn't work.

So I got that going for me....

Someone mentioned to me, and maybe on this thread too, that you can copy in code and it will analyze it and tell you where it is wrong. Of course, right or wrong would be dependent on knowing what you want to accomplish.

Then they expanded to AI in general as it is being used for national security and possible ways of using it in waging war. Interesting stuff.

I saw this and thought that at its core, even AI can be systematic and repetitive enough in what it does that it can't hide it is AI?

https://www.npr.org/2023/01/15/1149347424/a-new-tool-helps-teachers-detect-if-ai-wrote-an-assignment

 

[Flatbrimmer]

5 pages (1/3 of total) are robots lurking.

I’m 4/7 pages and I hope there’s a lot more talk about why I just seen there’s robots on our sh1t????

 

[kansasdad]

Thanks @Nameless Range for starting this thread.…..mind blown

 

[Nameless Range]

@neffa3

High Art man...
View attachment 260874

I can’t be the only one who thinks this is actually pretty good

 

[Schaaf]

Shut it all down.

 

[Benfromalbuquerque]

I had never heard of ChatGPT until this thread but then this morning as I was driving, I turned the radio to NPR, and they were discussing ChatGPT and its national security applications. Then they expanded to AI in general as it is being used for national security and possible ways of using it in waging war. Interesting stuff.

That’s the problem with NPR. I too have accidentally dialed that station.

 

[Hunting Wife]

Just tried to get in to check it out, and all I get is an error that the site is at capacity and it’s unavailable.

Must be public comment season…

Interesting thread @Nameless Range thanks for posting.

 

[Hunting Wife]

While I’m waiting around, I’m surfing around the OpenAI website. It appears they just put out a report in this very topic.

[2301.04246] Generative Language Models and Automated Influence Operations: Emerging Threats and Potential Mitigations

 

[Nameless Range]

Just tried to get in to check it out, and all I get is an error that the site is at capacity and it’s unavailable.

Must be public comment season…

Interesting thread @Nameless Range thanks for posting.

I often have to try a few times to get in over a couple hour period of time. Once I’m in I just leave the window open and it seems to kind of hold your spot in the instance.

 

[noharleyyet]

Maybe HTM found ChatGPT & didn't disappear...

 

[JMG]

I saw this kind of thing on 60 Minutes. Thank you Nameless Range for posting. This topic is very concerning.

 

[noharleyyet]

Won't open, hopefully that's a compliment.