G. Watson vs. Humans

On the television program Jeopardy, a computer named Watson played against two human contestants.  Did you see the program?  Watson did TOO good, taking an early lead, eventually winning, but I think it had an advantage.  I sent an observation to the Jeopardy website and never heard  back from them.

I wrote:  “When human contestants played Watson yesterday, you said the computer received the question as a text file.  I suspect you sent the entire question instantly to the computer as soon as Alex Trebak started reading the question to the contestants.  If this is the case, the computer would have an advantage of receiving the question faster than the humans, as the reading of the question takes a number of seconds to finish (as few as 5 and as many as 20?).  The humans had to wait for the finish of the reading of the question before they could answer but Watson would get the entire question right off the bat.  This seems to favor the computer compared to the human contestants.”

BTW, was Watson required to use voice recognition for Alex Trebek’s questions?  I never heard back from Jeopardy about this.  Doesn’t it seem like he should?  Isn’t feeding Watson the question as a text file PREFERENTIAL treatment?  Of course, we are dealing with the tyranny of the host, here.  He gets overpaid for enforcing the rules, not for giving participants the benefit of the doubt.

In any event, did you read about the computer?  It was something like 10 heavy duty servers in something like 28 racks (10 desktop-type computers per rack) each and it couldn’t even be in the same room as the contestants.
What would a similar comparison in medical transcription between computer VR (voice recognition) and a person doing traditional transcription be like?  Is it even comparable and what would the criteria be?
If the criteria was expense, speed and quality of printed report, I am sure the human would be better, because the VR system would have to factor in the expensive hardware, software, HOURS of training the computer AS WELL AS the doctors dictating AND the VR would require a human editor to review the computer’s output (not to mention the ADDITIONAL expense of a middleman/medical transcription service).
The only cost for my system of abbreviating EVERY repetitive word and phrase is a person, a PC and InstantText7 ($200).  The structure of the files, the formula and the categories are distributed for free (in hopes of advancing my humans are better than computer belief).
If the criteria was speed, the length of the report might be a consideration.  However, a good transcriptionist using an efficient, organized system of typing a report (pulling up a shell “boilerplate” document and using a system of unlimited abbreviations) can do a report once and submit it, whereas the VR would have to be administered by a human (who assigned the report to the computer, retrieved it when done and reviewed it/made edits for accuracy).  Since the turnaround time for getting dictation on the patient’s chart is a number of hours, if the VR was a few minutes faster there would be no significant savings by the computer.
If the criteria was economic benefit, doing transcription locally increases local jobs, benefits the local economy and allows the medical facility to have more ability to manage their transcription costs.
When VR salespeople hold out VR as better than traditional transcription, it is not true, but it definitely is more expensive, compared to a smart method of doing medical transcription.

Leave a Reply

You must be logged in to post a comment.