Hi All, I step in here and reply to a couple posts at one time. Am 03.03.2010 um 20:42 schrieb Pascal J. Bourguignon:
On 2010-03-03, at 19:24, Jim Adcock wrote:
However, The DP/PG approach is extremely expensive compared to what Google is doing. Consider: Google Books == about 10 million books photo scanned. DP/PG == 30,000 books "fully restored." So Google's approach is about 300X faster than the DP/PG approach. My Conclusion: In the best of all world's there would be some measure of VALUE in choosing which books DP/PG chooses to put effort into fully restoring -- the idea that somehow DP/PG is going to be able to fully restore all the world's books is surely false.
Google produces scan sets. Sure they put some in amore pleasurable form, But they are not interested producing books or even conserving them. The quality of the work is proof in fact. My personal opinion is that Google is simply interested in producing revenue, by what even means! That does not mean that Google does not have any merit. DP want to produce pleasurable eBooks. Personally, DP/PG has more value.
I think that the bet made by Google, is that sooner or later, sufficiently smart AI and OCR technology will be developed to allow to process its scans and do the job of PG automatically. I doubt this very much. AI proper has been dead since the failing of the ELISA project. Yes, the term is still used today to refer to anything that a computer does that seems to be intelligent. But it is hardly AI.
In the 80s machine translation was all the fad. The japanese said they would have a MT system that would translate in real time your telephone conversations by the 90s. Well, here we are some 20 years later and we can have the most horrific translation made online. The standard is that of my introductory class I had in the 80s. Googles service does not even use the half of the developments made MT.
The only question is when it will happen, and some think that singularity will occur within 20 years.
BB if it was realistic I would take you up on your bet. In 50 years their will not be a system finished that will do job of creating proper output anything above 95% fully automatically That is without any human interaction whatsoever.. Already, in the 90s it was said that with faster computers and cheaper storage the problems of knowledge engineering. Again, here we are and all is vaporware. It has been proven already in the 80s that human language Type 0 is and it is know that Type 0 can not be processed completely automatically by a computer. So the emphasis has change to simulate as much as possible. Yet, this will be always far from prefect. Sorry, for being here more than OT. But it was needed to prove the fact that anything having to do with language can not be handle by a computer program by itself. regards Keith.