1. Forum
  2. Usenet
  3. COMP.MISC

  • Decryption by ChatGPT

    From Sylvia Else@21:1/5 to All on Sun Feb 12 10:37:57 2023
    "The following is a well known English sentence encrypted by a simple substitution cypher. Spaces are unchanged. What was the sentence?"

    "qsb ayxjt ldzwh pzf eymkbc zvbd qsb onrg czi"

    The encrypted sentence reads: "the quick brown fox jumps over the lazy dog."

    ** OK, that was freaky. **

    "What about:"

    qxmb wnxqu pzd hz mnh

    The encrypted sentence reads: "jump now or pay later."

    ** Er, that doesn't even had the right pattern of letters and spaces. It
    was meant to be Time waits for no man. **

    ** Later, while I'm trying to explain something, without my usual
    quoting - you'll figure it out. **

    Yes, "it" and "hzw" both have the same number of letters, which is
    three. This is because in a simple substitution cipher, each letter in
    the plain text is replaced with a different letter, so the length of
    each sequence of letters remains the same in both the plain text and
    encrypted forms.

    How many letters are then in "it"?

    The word "it" has two letters.

    How many letters are there in "hzw"?

    The sequence "hzw" has three letters.

    Do "it" and "hzw" have the same number of letters?

    Yes, "it" and "hzw" both have the same number of letters, which is three.

    I think teachers who are concerned that students will use ChatGPT to do
    their homework need not worry (yet).

    Sylvia.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Rich@21:1/5 to Sylvia Else on Sun Feb 12 03:09:04 2023
    Sylvia Else <[email protected]d> wrote:
    Yes, "it" and "hzw" both have the same number of letters, which is
    three. This is because in a simple substitution cipher, each letter in
    the plain text is replaced with a different letter, so the length of
    each sequence of letters remains the same in both the plain text and encrypted forms.

    See if it can decrypt this then:

    dr%Z H#Ff CXV "p&-C &>IE MNqh 405 #(,, NccZk5 !9 {}GOHcl< 5y79= ;I
    RNfaiG n9)> ?L CNTZ171 %] + GEQbfg 2$1].:KTaXls p3)){D5 GKUh tu([)\ <K
    YUZ swt9= >'PT Qi pk374^}; RLea b mz4$_|.LZ Zax5y_g >| MIN cduv*$ #]
    /,GT mg0$4+)~ FE STqyr$+ ~{\.KXk tpu *6]] <K Gbok 405 }~[AN bUvz o97
    */@Mbamfm w%}#Az

    That is also a "substutition cipher" of sorts, just not necessarially
    'well known'. Richard Heathfield of sci.crypt termed it SCOS for Sci
    Crypt Open Secret. The 'key' to the above is "5 8".

    It will be interesting if it can decrypt that one or not.

    PS - I encrypted the paragraph of your post that I quoted above
    (without quote prefixes) with scos to create the encrypted portion
    above, so you will know if chatgpt manages to figure out the pattern.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Sylvia Else@21:1/5 to Rich on Sun Feb 12 14:52:48 2023
    On 12-Feb-23 2:09 pm, Rich wrote:
    Sylvia Else <[email protected]d> wrote:
    Yes, "it" and "hzw" both have the same number of letters, which is
    three. This is because in a simple substitution cipher, each letter in
    the plain text is replaced with a different letter, so the length of
    each sequence of letters remains the same in both the plain text and
    encrypted forms.

    See if it can decrypt this then:

    dr%Z H#Ff CXV "p&-C &>IE MNqh 405 #(,, NccZk5 !9 {}GOHcl< 5y79= ;I
    RNfaiG n9)> ?L CNTZ171 %] + GEQbfg 2$1].:KTaXls p3)){D5 GKUh tu([)\ <K
    YUZ swt9= >'PT Qi pk374^}; RLea b mz4$_|.LZ Zax5y_g >| MIN cduv*$ #]
    /,GT mg0$4+)~ FE STqyr$+ ~{\.KXk tpu *6]] <K Gbok 405 }~[AN bUvz o97 */@Mbamfm w%}#Az

    That is also a "substutition cipher" of sorts, just not necessarially
    'well known'. Richard Heathfield of sci.crypt termed it SCOS for Sci
    Crypt Open Secret. The 'key' to the above is "5 8".

    It will be interesting if it can decrypt that one or not.

    PS - I encrypted the paragraph of your post that I quoted above
    (without quote prefixes) with scos to create the encrypted portion
    above, so you will know if chatgpt manages to figure out the pattern.

    It didn't even try, denying knowing anything about Sci Crypt Open
    Secret, when prompted. It says it was trained on Usenet posts, but
    denies knowledge of specific posts or posters.

    I'm not surprised; it cannot even replicate the Lazy Dog success with a different character mapping.

    Sylvia.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Computer Nerd Kev@21:1/5 to Sylvia Else on Sun Feb 12 16:10:31 2023
    Sylvia Else <[email protected]d> wrote:
    It didn't even try, denying knowing anything about Sci Crypt Open
    Secret, when prompted. It says it was trained on Usenet posts, but
    denies knowledge of specific posts or posters.

    I'm not surprised; it cannot even replicate the Lazy Dog success with a different character mapping.

    That's disappointing. But while researching something completely
    different today I found this blog post where the author determines
    that it's somewhat useful for generating pestering emails to send
    to organisations and politicians (provided they don't begin
    filtering out AI-generated emails, which I'd argue they should if
    people really start doing this): http://writerofminds.blogspot.com/2023/01/chat-gpt-useful-application-test.html

    --
    __ __
    #_ < |\| |< _#

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Andy Burns@21:1/5 to Sylvia Else on Sun Feb 12 10:08:34 2023
    Sylvia Else wrote:

    It says it was trained on Usenet posts

    The real reason google bought dejanews?

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)