Neural networks being used to create realistic phishing emails

[crazycells]

@phenomlab recently, I have observed something interesting…

When I asked chatgpt for a recently published work about two scientific fields (in this case, epigenetics and aging) , it gave me an answer and list some article names.

At first, I was shocked to miss these papers, and not seeing them before… but later I figured out these are not real papers. Although, I have to admit, it sounds and looks legit, it is definitely how those article titles are constructed, even some of the authors are real scientists who are working in these fields… But I could not find any of these publications, so they do not exist

Somehow, chatgpt understands and learns how to give these answers, but it could not make the connection that these references should be real and represent something that exists…

[phenomlab]

@crazycells very interesting indeed. Particularly to provide works that do not exist, and yet reference known professionals in those fields as the authors when they are not.

[crazycells]

@phenomlab said in Neural networks being used to create realistic phishing emails:

@crazycells very interesting indeed. Particularly to provide works that do not exist, and yet reference known professionals in those fields as the authors when they are not.

yes, actually I have to add something…

I only identified two people, the rest of the authors did not exist. But one of those two people is a very well-known person in the field that has a lot of review articles… so that might be the reason… chatgpt assumed his name should appear on the list

[crazycells]

@phenomlab you may remember that people were writing a lot of assays or articles using chatgpt in the first week and many people were discussing how a lot of students will use this in their coursework.

I heard that a teacher was able to identify the homework assays written by chatgpt by directly asking chatgpt if it wrote or not lol , I did not read this but heard from a friend. that is funny, so chatgpt is a snitch

And I have read a few days ago, chatgpt makes the assays by known/predictable structures (how many sentences in a paragraph, how many words in a sentence, what kind of words should be used etc.) I guess that is the average of the many assays it scanned… that is why the assays look legit…

[phenomlab]

@crazycells chatGPT is a snitch… Love it.

On a serious note, you can’t help but wonder if the same algorithm is being used by teachers who use AI to detect plagiarism in essays to eliminate cheating.

[crazycells]

@phenomlab yes, I believe chatgpt (and its competitors) will be the first step of action for a lot of people especially young people soon

I have just asked for a job application email sample and a CV sample separately… it gave me very good samples… Of course, it is nowhere near polished or finished , but as a “first step” , I think it is very acceptable. Many people can utilize it especially if you are not professional. I do not think it is as sophisticated yet for professionals, but will become there at some point I guess…

you can even make it write a code for you, lol… I have just asked this:

[phenomlab]

@crazycells I bet the CV from chatGPT is of a higher quality than some I’ve received from real people in the past. Seriously, they are so bad in terms of basic spelling and grammar - which is inexcusable given that you have a spell and grammar checker right in front of you - that they often don’t make it out of the gate, and to interview stage.

Obviously, I make allowances for English not being the primary language or “mother tongue” when accepting CV’s from potential candidates overseas (and I do the same here - I won’t correct anything anyone posts on sudonix provided it remains within the guidelines), but these are from UK residents which makes it so much worse. The point here is that you could use something like chatGPT to compose your CV, but you’d be very quickly exposed if your grammar and spelling weren’t up to scratch

[crazycells]

@phenomlab said in Neural networks being used to create realistic phishing emails:

@crazycells I bet the CV from chatGPT is of a higher quality than some I’ve received from real people in the past. Seriously, they are so bad in terms of basic spelling and grammar - which is inexcusable given that you have a spell and grammar checker right in front of you - that they often don’t make it out of the gate, and to interview stage.

Obviously, I make allowances for English not being the primary language or “mother tongue” when accepting CV’s from potential candidates overseas (and I do the same here - I won’t correct anything anyone posts on sudonix provided it remains within the guidelines), but these are from UK residents which makes it so much worse. The point here is that you could use something like chatGPT to compose your CV, but you’d be very quickly exposed if your grammar and spelling weren’t up to scratch

yes, CV sample is quite acceptable as a first draft… I agree it would eliminate all the small mistakes. I will try to use it more commonly and figure out more ways to get benefit from it

[phenomlab]

Just came across this which made me laugh. Very much in keeping with this topic

[crazycells]

@phenomlab lol, good… chatgpt is adapting into human culture…

[phenomlab]

@crazycells just came across this. Looks like Google has finally jumped into the bandwagon with it’s own offering called “Bard”

https://news.sky.com/story/google-launches-ai-chatbot-bard-to-rival-wildly-successful-chatgpt-12804958

[crazycells]

@phenomlab so here it begins… AI wars… we have to protect John Connor no matter what happens…

[phenomlab]

@crazycells wondered how long it would be before the Hollywood connotation got a mention . In all seriousness, it’ll be interesting to see how this inevitable battle of the giants will play out.

[phenomlab]

@crazycells Interesting

[crazycells]

@phenomlab said in Neural networks being used to create realistic phishing emails:

@crazycells wondered how long it would be before the Hollywood connotation got a mention

lol…

what if Bard starts chatting with ChatGPT and they realize that H**o Sapiens is inferior to them, so they join forces to form SkyNet to enslave us ?

[phenomlab]

@crazycells this perhaps?