The speed of development, networkization and globalization that is currently in progress these days has affected the field of translation, too. Therefore, the process of translation was never before as simplified as it is today, and this is due to the invention of machine translation. The great schools of translation as School of Jundishapur, Toledo School of Translators, Chang‟an School of Translation, etc., also played a huge role in the emancipation of translators – as carriers of knowledge and translation – as activity, but it was machine translation that fundamentally changed any fantasy or approach on translation. Eventhough the history of machine translation starts in the 1950s, it managed to become translator‟s best friend in terms of accuracy, time, cost, etc.  However, nothing is perfect in this world. Although Google Translate is considered to be one of the most popular and sophisticated machine translation it continues to be imperfect. Let us consider the following examples: Lady Gaga was translated from Malay into French as Britney Spears; Android 2.3.4 (German) was rendered as iphone 2.3.4 (English); Ce trist este cînd o mamă îşi uită copiii de dragul unui bărbat... (Romanian) – How sad it is when a mother for the sake of her children forget a man... (English) which is a total loss of meaning; or James Bond (English) rendered into Chinese as 007; Samsung (English) – SonyEricsson (Romanian); simpaticone (Italian) – rendered in English as Barack Obama; Я конечно могу писать по английски (Russian) – I certainly can not write in English (English) – the same loss of meaning; a night to remember (English) – гибель Титаника (Russian), etc. Well, the translations are pretty funny, it is true, but the main principle in translation is accuracy, therefore there can not be any space for such mistakes. Google Translate (GT) can be compared with a superhero that comes to the rescue of the people who encounter difficulties when learning a foreign language. However, even professional translators use GT to look for eloquent translation and to save time. The problem is that this superhero is not always at its best, and sometimes can make huge mistakes, thus people start to lose faith in it. Nevertheless, it still remains the most accessed database in the realm of translation. With regard to Romanian – English translations performed by GT, there are many issues which are probably common to many languages. Fortunately, GT does not consider the Romanian diacritics, therefore, when you enter a word without diacritics in its database, it gives you a correct translation with a number of appropriate examples. For example, for simtire (sim ț ire) GT has the following translations: (feeling, sentiment, consciousness, soul, pathos); for sarut (sărut): kiss, osculation; politete (politeţe): courtesy, civility, form, secency, etc. Another problem is that of Romanian abbreviations: Serviciul Hidrometeorologic de Stat was rendered by GT as SHS; CNA şi ÎS CRIS Registru – HE CRIS CNA Registry (eng.); ÎS (which is Întreprindere de Stat in Romanian) rendered as SOEs (we suppose this abbreviation provided by GT can be deciphered as State owned enterprises, however it is inaccurate); AIRM (ro) – AIRM (eng.). The first issue is related to proper names. Proper names are a bigger pain in the neck for GT. Everybody knows that names of authors are never translated, but we have entered a few names of Romanian personalities, in order to see how Google Translate would deal with the challenge. For example, Mihail Sadoveanu was translated just as Sadoveanu; Mircea Cărtărescu rendered as Mircea Mapping; Nicolae Văcărescu became Nicolae Cowboy; Zilot Românul was translated as Zealot Românul, the word zealot meaning a fanatically committed person; Nicolae Dabija was translated as Nicholas Dabija, at least in this case Google Translate has offered us an equivalent into English for Nicolae though it is inappropriate; for Iurie Leanca, GT provided the following versions: Iurie Leanca, Minister Iurie Leanca, Prime Minister Iurie Leanca, Mr. Iurie Leanca, which allow to the user to choose the version he needs. Other funny examples which should be mentioned are: the translation of Florin Piersic as Barry Peach and Boris Găină as Boris Hen or Boris (the) Chicken. Another issue is synonymy. Dealing with synonyms is a real challenge not only for GT, but for common users as well. For example, the word persoană fizică is translated by GT as follows: individual, natural person, physical person which may become confusing for a user. The same with companie rendered by GT as: company, society, companionship, group, crowd, partnership – thus a user should be very vigilant in selecting the best translation. For example, for Dumnezeu GT provides three translations: God, lord, master, all three words can be used in the same context, when we refer to the Almighty, but the third word has different connotations which would need a special context; the same with the translations for the Romanian rând: row and line, which can be both used exclusively according to the context, that is why they must be chosen carefully. Set expressions, proverbs and phraseological units also may create semantic gaps when rendered by GT. Let us consider the following examples: a spăla putina – to skip, to absquatulate and to pack off. It seems that the most appropriate equivalent for the Romanian expression is to absquatulate which is a deeply silly word that means to make off with something or someone, in other words, it is an archaism and it is accepted even if it is used only ocasionally. A călca pe bec – was rendered as to snitch; la paştele cailor – on the Greek calends; a fi prins cu mâţa-n sac – caught red-handed which in fact are very good equivalents. However, there are inappropriate translations, too: a face cu ou şi cu oţet – to make egg and vinegar; vrei, nu vrei, bea Grigore aghiasmă – Would not want, drink holy water Gregory (which is grammatically and semnatically incorrect); a-i face cuiva capul calendar – a I do someone's head off; a avea ac de cojocul cuiva – having one's coat pin; a bate apa-n piuă translated by GT as a knock-in water mill or a face din cal măgar şi din ţânţar armăsar rendered as a donkey horse makes a big deal which is pretty funny in English. Grammatical mistakes are also very frequent. Sometimes, GT forgets to translate prepositions and „swallows the articles”. For instance, pentru Elena was translated just Elena; la mijloc was translated as middle instead of in the middle; din 2000 was translated as 2000; în spatele magazinului was translated without article the, thus we obtained a meaningless translation such as behind store (an article goes a long way!!!); în seara de Crăciun was translated as Christmas Eve, when it should have been translated as on Christmas Eve; etc. The problem of tense is also present, for example: să-i fi dat lui banii rendered as to have given him the money. Let us take the following groups: eu scriam (imperfect) – eu am scris (perfect compus) – eu scrisesem (mai mult ca perfect) – eu scrisei (perfect simplu) – eu voi scrie (viitor simplu) – eu voi fi scris (viitor anterior) – eu oi scrie (viitor popular) rendered as I wrote – I wrote – I had written – I wrote before – I will write – I will be writing – I write sheep, the last example being the funniest and the most inaccurate. And finally, we should mention the gaps of translation made by GT in literary texts. As a rule, GT is efficient when it comes to scientific texts, terminology, different field-based constructions and units, but not so efficient in the translation of literary texts. Let us consider the following example: “Vai, şi va veni o vreme/ Când adormi-vom amândoi,/ Şi înstrăinaţi, prin cimitire,/ Va plânge toamna peste noi./” translated by GT as “Oh, and there will come a time/ When you fall asleep, we both,/ And alienated by cemeteries/ We cry Fall upon us.” It is semantically and grammatically inaccurate. As a result of our analysis based on a bulk of examples and sentences, we can state that GT is still an unreliable source of translation and has to improve in respect of grammar (tenses, syntax, cohesion and coherence), vocabulary (continuous-based updating) and text/sentence level approach. With regard to literary texts, it stays pending the question whether GT or any other machine translation will ever be able to make an accurate translation.