Wie gefährlich ist ChatGPT?
Ich habe heute mal gesehen, wie man damit Schwachstellen-Analyse für Webseiten machen kann und würde fast sagen, dass man damit das halbe Internet sehr intelligent angreifen kann (zumindest alles, was auch einer OpenSource-Technologie basiert) und das ist vielleicht auch nur ein Beispiel, das ich kenne
I think improved search for vulnerabilities is the least problem. The problem is from my point of view deep fakes and lyrics that look good and are still wrong. Furthermore, there is a risk that certain jobs will no longer be asked, because ChatGPT writes beautifully. And in the end, quality and diversity suffers.
ChatGPT is nothing more than an improved search engine.
And if it were as dangerous as you would like to show that suddenly no website is safe, then we would see, read and hear clearly more about it. And it would also lead the operators to eventually invest more in their security.
But yes, just as you can abuse a hammer or a car as a murder weapon, you can also abuse ChatGPT for things that may not be so great.
Is that why we should ban hammer, car, etc? Because they can do evil in wrong hands? I don’t think, especially since we should ban the Internet, do we?
I would also be against a ban. There are too many interesting and good applications.
The security risk remains. I’m sure we’ll have a lot of comic experience.
I am also sure that it is much more than an improved search engine.
And with what do you establish this opinion? Do you have any expertise in this field?
False
ChatGPT is not a search engine
It only creates texts from an outdated own database and not from the internet.
A search engine searches the internet every day around the clock and collects the data to find and offer the appropriate website on a search request.
You are comparing apples with cherry kernels
This description comes from me. Experience shows that lay people can understand better than what they should look at ChatGPT.
🤣😂😅
And what is the test to prove?
This does not have to do with a search engine and not even with a tool that abuses the internet
Google’s LaMDA-KI also passed it, which subsequently proved its invalidity. The test is deemed to have been passed if a person cannot significantly assign the answers of another person and the AI to one of the two.
But what does that prove? That ChatGPT is able to formulate answers that act for an outsider as if they could have come from a person. No more and no less!
This does not make ChatGPT intelligent or gives it a consciousness or something. It only shows that the training data and parameters were sufficiently good to work intelligently.
ChatGPT didn’t solve anything. Other programmers have solved them and the AI has found sufficient examples of your problem in their training data to present functioning snippets.
Under warranty! ChatGPT cannot think creatively. It is 100% excludedthat it can solve a problem that has not yet been solved by a human being!
I work with ChatGPT myself and often it presents solutions that do not work. Especially in complex algorithms, it often fails mercilessly!
I can justify that ChatGPT passed the Turing test.
There are various very special programming problems that ChatGPT has solved excellently for me. That was nothing you can google, not on stackoverflow and co.
A chat is a conversation and has nothing to do with vulnerabilities analysis.
There are other AIs that are much better suited for this.
ChatGPT can only talk