From the Reddit comments:
"As a pediatric neuropsychologist, I can promise you that if this were anywhere near true, it would indeed be earth-shattering. It would entirely revolutionize our understanding of (a) autism, (b) mental ******ation, (c) how language develops, and (d) how children learn to write and spell.
Children who are profoundly autistic like this girl are can sometimes be taught, eventually, to use a communication system (usually pictures) to get their wants met, but that is not what they show this youngster doing. For example, you'll notice that she doesn't communicate her wants that way at all -- not once did we see her type out "Just give me the damn food and stop making me type" or something similar. Unfortunately, children with profound autism cannot be taught to communicate a rich inner monologue full of observations about things they've never experienced (e.g., what it would be like to go to school with normal kids who aren't mildly alarmed by her behavior), metaphors ("ants crawling" on her arms), and figures of speech (e.g., "it's not like there's an off switch").
The reason that children with profound autism can't do these things is because of what autism does to the brain. Autism appears to be associated with too much gray matter, which makes every new piece of information appear discrete, like it is unrelated to every other piece of information. For that reason, individuals with autism have trouble discovering general principles that tie discrete bits of information together, which is a necessary skill for learning language (e.g., you have to learn the general "rules" of syllabification, sequencing, and grammar through exposure).
They also have trouble understanding where one bit of information stops, and another starts, which causes they to 'chunk' together things that don't go together, or have trouble breaking a large chunk down into the parts that make it up. Here, for example, is a very recent study on computer-analysis of infants with autism showing that they can be reliably differentiated from infants without autism and infants with language delays based on this difficulty effectively understanding where to break language into its component syllables.
As a result of these difficulties, children who are profoundly autistic may never learn to speak at all, while children who are less autistic still demonstrate significant problems correctly 'chunking' language bits together (for instance, they often display echolalia, meaning that they repeat whole sentences or scenes from movies verbatim, because they learned it all as one single unit of information). They also have trouble understanding language rules such as how words change depending on point-of-view of the speaker (e.g., they refer to themselves as he/she or by their first name, or use 'you' instead of 'I'). By definition, autism also includes trouble understanding language pragmatics (e.g., metaphors and figures of speech), because pragmatic language requires seeing the connections between different types of information, such as the literal meaning and abstract/contextual meaning or a sentence.
Autistic children also have difficulty learning to write, for all the reasons listed above, but also due to the phonetics of spelling -- learning how to spell also requires seeing the general 'rules' that govern phonics, as well as breaking words down into their component sounds. Even when children with autism do learn speak and write, one of the cardinal features of autism is trouble understanding Theory of Mind, or being able to make connections and use context to guess what others are thinking. For example, saying something like "Don't judge me until you know me" would be extremely difficult for a person with profound autism, since it requires theory of mind to know that other people are evaluating you differently than you evaluate yourself.
I have absolutely no problem with the idea that a person with mild to moderate autism can learn to type, after she learns to speak and then spell (see Temple Grandin, as others have noted, along with thousands of other autistic individuals), which is the same order in which neurotypical children learn to do these things.
But the idea that a person with profound autism could overcome how her brain works -- but just while typing, not while speaking -- all at once? That she could, without training in any of these areas (remember, she just ran to the computer and typed "HURT" and "HELP") master:
Syllabification,
Word order and sequencing
Rules of grammar
Phonetic rules
Correspondance between sounds and letters
Correspondance between the capital letters on a QWERTY keyboard and the lowercase letters that type on screen, and understanding how a generative typing software program works (one that fills in the rest of the word after you type the first few letters)
Expressive vocabulary
Correct pronoun usage
Point-of-view
Non-echolalic speech
Figures of speech and metaphors
Theory of mind, etc
(Not to mention her mental ******ation)?
And only do so to communicate things like "what [an adult who is not autistic might think] it is like to have autism" and not really be able to do so when the camera is present?
Unfortunately, the likelihood of this is so terribly, vanishingly small that I am fully comfortable saying that it is impossible.
We may all wish it were true, but it isn't. And putting stories like this out there only clouds our understanding of what autism is and how to treat it (which hopefully will eventually lead us to how to prevent or cure it)."
http://www.reddit.com/r/science/comments/ctm81/autistic_girl_expresses_profound_intelligence/
I'm also very skeptical :/