Re: Machine-Readable Records

My point is that if you take the term "machine processable" literally 
(as applied to an information format) it becomes completely meaningless, 
because ultimately *any* information format can be processed.  But terms 
of art are not meant to be taken literally.  A term of art is always a 
short hand for a concept that otherwise would be considerably more 
cumbersome to reference.

"Machine processable" formats (in the term-of-art sense) were *designed* 
for easy precise machine processing.  Natural language was not.  This is 
a useful distinction independent of AI.  Like any term of art, it's 
meaning needs to be explained to those who are not familiar with it.

Perhaps a more evocative term of art for this concept will emerge at 
some point.  If so I would welcome it.  But thus far I have not found a 
better one.

Thanks,
David Booth

On 7/7/24 05:03, Dave Raggett wrote:
> Hi David,
> 
> Whilst I appreciate the distinction you’re making, the rapid widespread 
> adoption of AI will change how people think of machine processable 
> information, so I think a different term is needed to distinguish 
> informal from formal information. Consider the contrast between a 
> business contract expressed in legalese (a subset of natural language) 
> and the information used to state a particular customer order where 
> precision is essential.  Trying to hold back shifts in the meaning of 
> words is reminiscent of King Canute and the tide, see: 
> https://en.wikipedia.org/wiki/King_Canute_and_the_tide 
> <https://en.wikipedia.org/wiki/King_Canute_and_the_tide>
> 
> p.s. I am convinced that the energy consumption of AI will come down 
> very considerably as the technology switches to sparse spiking networks 
> on neuromorphic hardware. This may however take another decade yet.
> 
>> On 7 Jul 2024, at 01:31, David Booth <david@dbooth.org> wrote:
>>
>> On 7/5/24 03:17, Dave Raggett wrote:
>>> Just to note that machine readable formats now includes natural 
>>> language, images, video and sound. AI is good at handling imperfect 
>>> knowledge,
>>
>> I don't disagree with AI's ability to handle imperfect knowledge, but 
>> I do wish to quibble about the term "machine processable".  I have 
>> been using terms like "machine readable" or "machine processable" for 
>> decades as a way to distinguish between formats that are readily 
>> amenable to precise, deterministic interpretation versus formats that 
>> can only be interpreted heuristically and not necessarily correctly. 
>>    This is a key difference, for example, between a data formats like 
>> RDF and natural English.  I have found this distinction useful, 
>> because The former requires a relatively tiny amount of processing 
>> power; the latter requires enormous amounts to reach high accuracy of 
>> interpretation, and even at its best you can never be sure that the 
>> machine guessed it right.  Applying the term "machine processable" to 
>> natural language muddies the water, making it more difficult to 
>> identify and discuss this important qualitative difference.
>>
>> Thanks,
>> David Booth
>>
> 
> Dave Raggett <dsr@w3.org>
> 
> 
>

Received on Sunday, 7 July 2024 15:34:39 UTC