Not Latin.
You may ask yourself: How many people are likely to want something translated from Latin, and how does it compare with the number of people who are likely to want something translated from Upper Sorbian?
In the fifteen seconds he was willing to devote to the problem, Dr. Boli was not able to think of a good answer to how many people need to have things translated from Latin. But we can at least guess how many people are likely to be producing content on the Internet in some of those lesser-known languages. There are roughly 13,000 speakers of Upper Sorbian in the world. There are (by the last census) 1,316 speakers of Inuinnaqtun in the world. Let us assume that every one of them is a content producer. If they work eight hours a day producing content, it will take them decades to create as much content as is already available on the Internet in Latin.
What, then, does it mean that Microsoft Edge will not attempt to translate Latin? When we see the omission of Latin—which is actually quite typical for language tools (remember how long it took to get usable polytonic Greek in the computer world?)—we immediately begin to think that there is a giant cultural assumption at work here. The assumption is not just that the past is worthless: it is that it would be wrong not to repudiate the past. You would feel dirty if you programmed your translator for Latin. The eighteen people who communicate on the Internet in Inuinnaqtun should be encouraged; the millions of scholars of worthless dead things who need to read Latin should be told what we think of them.
Under the circumstances, we should give Google some credit for attempting to translate Latin, even if the results are usually comical.