[deleted]

Hahaha, maybe they will learn something.

But it is a funny title.

Edit: I have expressive aphasia, I have have a hard time programming. ChatGPT has changed my life, and now I can program. Since I know how to program previously I now know that chatGPT writes outdated code or code that doesn't exactly work, I have to update or add a lot to it so that it could run or be compiled.

I figured it’d be a matter of time. You still definitely need some awareness of programming to successfully implement some of ChatGPT’s code, but wow is it remarkable.

If you couldn't get this from the headline already: it is just a clickbait article of course.

ChatGPT is really good at creating small code snippets for easy tasks such as iterating files in python or simple encryption/decryption. But so is google, stackoverflow, github, ...

If you really want to do something more complex or have very specific requirements (such as abusing a specific security hole) it is useless.

The focus on malware here is misleading. Any resources that helps with basic programming could easily be used for simple malware shown in the article.

Might as well make an article how ChatGPT can help with robbing items from a car because it can recommend to use a brick to break a window. This is just the programming version of it.

This is dumb. If you still have to debug it and customize it (you do) it’s not really any different from grabbing code from any number of existing sources.

If it fails to compile, can we get ChatGPT to write a good question directly on StackOverflow for us? 🤔

I’m not a programmer but found it could create audio plugin code when I specified what I wanted (in C++) I wanted to try it out - is it likely to not work and for me to have no idea why?

My windows 2003 server is shaking in it's boots.

This is a self limiting system.

I gave it "Please write a parodic stack overflow post by a script kiddie trying to get their malware script to work and failing":

>Hey guys,

>I'm trying to run this malware script I found online and it's not working. I keep getting this error: "failed to execute script." Can anyone help me out? I'm really new to this whole hacking thing, so any help would be appreciated.

>I'm running the script on a Windows 10 machine, if that matters. I tried running it as administrator, but that didn't work either. I also tried using a different antivirus program, just in case that was the problem, but no luck.

>Any ideas? Am I missing something obvious here?

>Thanks in advance for any help!

(Then it told me that both my question and its answer might violate its content policies.)

I like that it slipped in there the implication that the root problem is that the script kiddie is trying to run antivirus on their own script.

>If you really want to do something more complex or have very specific requirements (such as abusing a specific security hole) it is useless.

I wonder though if you fed in certain CVEs if it could figure it out. I don't want to try it because that sounds like a good way to get kicked off.

[deleted]

[deleted]

For older OS's maybe, but for modern stuff they shouldn't even bother. It's not going to get through.

That’s not how any of this works.

Agree this is not a big deal. I can't say how far out AI malware is, but it's not here yet

It is just done via existing function in python / pip modules not manually

You will get stuck very quickly trying this. Good learning experience though

Anything complex and ChatGPT starts outputting broken (in both obvious and non obvious ways). It’s good, however, for writing small parts of relatively simple boilerplate code.

Just another example of how tools can be used for good or evil.

This is a good time to remind everyone, homebrew encryption is not the answer. If you don't know what you're doing, you are already making it unsafe. Use current known encryption for anything production worthy until you gain the skill to code for it.

Anything can be turned into a criminal case even a spoon. So should this be a sample to create a barrier to use such KI or spoons ? People should have more fear what this Technologie will change because it has the same Technologie impact as the ARPA net or iPhone had. Any company which host such a powerful KI would rule anything. I hope there will be also a free and open data one and not only those from MS or Alphabet with Censored content.

>abusing a specific hole

There’s a subreddit for that

err... script kiddies could just google it anyway

This is just offloading about 30 seconds of work

while(1 == 1){ //access some url }

Yeah it seemed to be doing seemingly very complex things in far too few lines of code, I was suspicious if it was really doing what I asked

Must be smarter than I, I still can’t figure out where the fuck you go to even use this thing.

[deleted]

[deleted]

Found the guy who clicked on a script kiddies malware

It’s honestly nuts to me that it gets this nuance so right. Like I’d expect it to write some generic thing but it looks like a legit post.

Oh no new technology!

[removed]

ChatGPT has been banned on SO for a few weeks now. It was causing a large amount of incorrect answers. It's tough to police though, because mods are basically having to decide if it sounds like ChatGPT or not since there isn't a way to tell 100% yet.

No, it is not.

I’m not sure if this is the start of some super-intelligent AI or the beginning of the loss of all modern human knowledge. Like what’s the speed of entropy in a system without error correction

[removed]

From what I have seen, chatgpt is like using google but 100x more efficient. With that said, is this just a fear baiting article? if a script kiddie/novice programmer wanted to create malware they already had the resources to do it. ChatGPT just makes it more efficient.

Admin scared of becoming obsolete

I totally agree, it's seriously impressive how well it gets all the little details in language. If you haven't checked it out yet, I'd definitely recommend giving it a try. I think you'll be blown away by how well it can understand and respond to pretty much anything you throw at it.

[deleted]

ChatGPT ask a question on StackOverflow.
ChatGPT answer a question on StackOverflow.
ChatGPT admin StackOverflow.

This seems to be the entrance to Matrix.

Cars are allowing terrorists to kill multiple people easily without guns. We need to go back to horses

So far though I feel like most of the answers it gives me are fairly generic.

Ask more specific questions, or twist the task a bit

of course

the chatbot won't write a story about a robot accidentally squishing somebody to pieces because it won't promote violence but it will allow for malicious code to be written.

That answer above was generated by it, so at least it passes for normal conversation.

This one is not, just to be clear!

You can iterate in a response. I got to the one above by making a prompt and then asking it to "rewrite the last response in a less formal style."

You do have to acknowledge its limitations. It doesn't understand current events, politics etc. It won't understand niche pop culture references.

You can ask it to e.g. write a response in the style of a Reddit comment reply or in the style of a ten-sentence children's book or in the style of a newspaper article, which helps enormously with getting tone or cadence correct. It's often better to ask it a normal question and then ask it to rewrite the response in a particular style.

IMO it excels at summarising information. "Write a 600 word essay on the causes of the Hundred Years' War" or something. Or simply paste a lot of information and ask it to condense it into 200 words.

Worth pointing out it's only the free version, and paid tools which are far more capable also exist.

You’re correct about the loss of knowledge. IA is just about what came often. Experts’ answers are not what it trains on, so only average answers and knowledge will be on that kind of ai.

Wait until it can train itself to be expert on a field. Then humans will really be obsolete.

Im trying to get him to write me a script in shell language to remove this nasty malware i have. I factory reset my laptop but it keeps coming back. I just dont know anything about programing

From what I read this no one doing this is a script kitty.

It’s very specific and they have to have deep technical knowledge already.

Looks like it more likely saved time.

This is actually very useful for generating believable email spam I think.

This tool might be more useful for spam generation than professional work.

It’s got just the right level of believability.

Yeah replying to people in spam emails too I imagine

This is why I always hated the Turing test. Data from Star Trek would fail that test.

The Turing test tests how well an AI can fake being a human. And as such it was passed by having a good understanding of human non answers pretending to be foreigners and children.

Basically any current event, political or human experience question is unfair because a bot won’t have human and real world experiences even if it was sentient. It could be trained on those answers though and provide a believable answer for something like “what’s your favorite sports team”, but it would never have watched sports.

An AI can be sentient without acting like a humans with fake human experiences.

Not saying chat bot is sentient but I think the line of what is sentient is going to get blurrier and blurrier in our lifetimes. While that google guy who claimed their bot was sentient was definitely wrong he may go down in history anyway as one of the defining moments of when what defines sentient and where we need to start redefining our definitions.

Well, as much as I believe this can be a dangerous tool (because people don't understand that it's a chat bot and experimental academic effort and not a proper knowledge source) I don't think we will get to a point were we have to worry, as this is likely a short public experiment by OpenAI team to get data on their model deficiencies and will like be either:
A) Replaced by a better, more accurate version.
or
B) Limited/removed from the public beta were anyone can use it.

That's the case for ChatGPT at least, AI as a whole is another can of worms that we are just opening.