According to the Wikipedia page of the Socratic paradox 'I know that I know nothing', Latin version of the same is — 'Scio me nescire' (https://en.wikipedia.org/wiki/I_know_that_I_know_nothing).

However, Google Translate says that English translation of 'Scio me nescire' is 'I know I know'.

Why is their a difference between the info of the 2 sources and which one is right?

  • 2
    Computer translation programs are well known for being wrong. It's not that they just make mistakes (like humans are known for) -- it's that they calculate their translations (much like a math problem), but they lack the ability to suspect that their translations could be wrong. (At the very least, if they're programmed to have the ability to suspect wrong translations, they don't (yet) communicate it to the human user.) So always take computer translations with a grain of salt! (Meaning that you are allowed to be skeptical of them.)
    – J-L
    Commented May 28, 2019 at 17:10

1 Answer 1


Because Google Translate is wrong. It does not, (or not only) use the dictionary meaning of words, but learns phrases in context. In many cases this can help create a natural translation but (especially for short phrases out of context) it can lead to nonsense.

Nescire ("ne scire") means "to not know".

Scio me nescire is literally "I know myself to not know" - this is the normal way in Latin of expressing "I know that ...".

So a faithful translation into normal English is "I know that I do not know".

  • 1
    What is more, Google Translate doesn't use any model of grammar at all, or at least not until recently; I read they were surprised at how quickly and easily a small company created a superior translation website (DeepL), which does use a model or grammar, and so Google finally tried to integrate some grammatical model as well, although I don't know how well and to what extent they have done or are doing that. P.S. DeepL doesn't have Latin, but it is mostly superior to Google in the languages it does have.
    – Cerberus
    Commented May 28, 2019 at 15:04
  • 1
    I feel like this almost isn't complete without mentioning the connection between scio and nescire; that scio comes from scire ("to know") and nescire ("to not know") is literally just ne scire . Commented May 28, 2019 at 15:34
  • 1
    Added, @RaphaelSchmitz
    – Colin Fine
    Commented May 28, 2019 at 16:13
  • @Cerberus It's a big controversy in machine translation, whether to use rule based (i.e. human extracted syntax rules) or data-based (statistically learned patterns). Rather, currently it is not controversial at all because recent statistical based methods (DL, LSTM, BERT) are so much more successful. There's the classic quip by Jelinek, a speech recognition researcher: "Every time I fire a linguist, the performance of the speech recognizer goes up"
    – Mitch
    Commented May 28, 2019 at 17:37
  • @Mitch: Noöne suggests only one method should be used, but, from what I read, Google's disregard of grammatical models has proved inferior to that of DeepL, which no doubt combines models with statistical methods.
    – Cerberus
    Commented May 29, 2019 at 0:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.