Mister Shake, I don't think there is any reason to justify actually doing this, but I do think it asks questions that are interesting.
Loved 'Ghost in The Shell', my favorite scene is the beautifully rendered trip through Hong Kong in the rain. A 5 minute study into what it is to be a living thing experiencing the moment. Truely magnificent.
As for practical uses, you're on the ball again. Neural networks are notoriously hard to put together, especially when so many connections are required in order for the network to perform with any usefull accuracy. Why spend time threading all the wires just so, etc when you can take a piece of natural computer and train it to perform the tasks you want performed. This itself suggests that the tissues that make up our brains are higly flexible, trainable and 'plastic'. People may in the future eschew the body, and being born into a machine, live out their lives from within an artificial sensorium. Applications might be space exploration, or some other endeavour for which our bodies do not provide adaquate support. Things that we may be unable to currently comprehend might be easily graspable by someone with an upbringing that has been solely tailored with that purpose in mind.
Once again, justification is not something I'm trying to do, I'm just pointing out possibilities and asking whether they are possible, and if so, what implications do those possibilities have on the way we see the world around us?
If a patch of cells can control a computer, or recognise a pattern, or perform one, or a range of tasks that we find natural, but which logic-based machines find difficult - at what point does a patch of cells get to think of itself? At what point (if any) does it become self aware? Can it just be that after reching some critical mass of connections, conciousness emerges?
At what point is a brain in a vat under the control of others? Does it have free will? How does it compare with a brain settled comfortably in its body?
The idea of keeping the infants body intact, but allowing senses we're less interested in atrophy due to disuse was just my way of illustrating a way of getting a 'brain in vat' but without any surgical procedures which might confuse the issue.
Finally, returning to the 'why do this rather than have a lump-of-cells' point, I suppose again for practicality. The brain-body arrangement already provides easily accessible input and output interfaces. The lump-of-cells needs to be trained to recognise various inputs, and to act in such a way to send equally various outputs.
If you are interested in thought and developments along these lines, I can thoroughly reccomend reading or finding out about a chap called
Igor Aleksander and the types of research he and many others are doing in this field (not, I hasten to add invloving bottled babies of any kind!)