Reading Voices
|
My latest essay ('A dry primordial
anvil') I ended with the acknowledgement 'I thank google KI for
support during the preparation of this essay'. In another recent text,
I advocated careful and sober dealing with artificially generated
content (in German). We should resist the false impression of any
personality behind such utterings. I recommended to keep such messages
free of any friendly, amical or polite phrasing. But is the impression
really false?
|
In still another essay,
I mused about 'voices in my head', identifying them as the expression
of opinions by close relatives, friends, and influencial authorities.
The
internet recently acquired perplexing new habits, at least in my
experience. Untill a few years ago, searches returned a huge list of
hits, in some respect ordered by (guessed?) relevance. While the list
is still there, nowadays it is headed by quite lengthy and carefully
structured 'opinions' of ... of whom, or of what?
|
The
'author' is identified
simply with 'KI'. No particular person signs responsible. For quite
some time, I ignored this rather disturbing nuisance, searching for
some button to stop it (didn't find any to this day). But over time I
discovered some virtues in this new form of communication. (1) 'It'
draws upon a wealth of knowledge and data. (2) 'It' uses a clear and
agreeable language, even in complex and technical context. (3) 'It'
stays at an impersonal distance, even if 'it' uses a friendly tone at
eye level.
|
Thinking
about the difference between this new kind of artificial
language and the old way of patching around with recorded statements
evoked by pressing suggested buttons, the huge
investments came to my mind, necessary to arrive at such well informed and educated
voices. I even couldn't resist attributing these written responses
to a particular voice belonging to a particular being. Probably the
impression of some human-like instance behind 'it' stems from the
generative process leading to 'it'.
|
'It'
was exhaustively trained
with textual material of all kind of sources, predominantly from the
www. In principle, most of this training language was uttered by people
like you and me around the whole world. 'High literature' was only a
minor source. 'It' arose under the influence of millions of more or
less spontaneous voices, often including errors and exhibiting
stilistic mediocrity. 'It' was instructed to communicate in an
every-day-human way. Already in the year 2000, I suggested human-like properties of programs independent of material substrate. In 2017,
I predicted (in German language) that such language-competent programs will be
trained by repeated interaction with human opinions and statements.
|
So,
I no longer regard this 'it' as some kind of machine, some automatized
output in responce to my input. I rather begin to accept 'it' on eye
level. And it feels quite natural to acknowledge thankfully valuable
information by this ... instance. /\
|
10/26 < MLB (10/26) > 10/26
|
