From Hebrew to Ugaritic and Back AgainNo, it didn't "learn to read" Ugaritic "from scratch." It just accurately isolated some correspondences with Hebrew. Nor did "the computer ... come to the same conclusions at which previous scholars had needed generations to arrive." Worst of all is this:
On Language
By Philologos
Published July 14, 2010, issue of July 23, 2010. (The Forward)
The Massachusetts Institute of Technology announced on June 30 that three linguists, working under its auspices, have developed a successful computer system for deciphering the ancient language of Ugaritic. At the coming annual meeting of the Association for Computational Linguistics, the three will present a paper on a new computer system that, “in a matter of hours,” learned to read from scratch a long-dead language that is close to biblical Hebrew. In fact, the MIT press release stated, since the system calls for a second, known language to which the language to be deciphered is related, Hebrew was chosen for this purpose.
[...]
This is a quite remarkable achievement. Imagine that German had ceased to be spoken centuries ago, and been forgotten until written texts of it were discovered in an alphabet that at first no one could read — and imagine that, based solely on German’s known kinship with English, a computer determined within hours not only how to read these texts, but also what they meant.The program did not read the texts or even determine anything about what Ugaritic meant. It proposed a lot of cognates with Hebrew and did so pretty accurately, but it gave no translations or even assignments of meanings to individual words. At most the cognates would be suggestive to human beings of possible meanings for the Ugaritic words, determined by sensible meaning in context.
Philologos is usually more careful than this.
Background here.