10 Feb
2025
10 Feb
'25
11:09 p.m.
On 2025-02-10 16:35:26 deloptes via tde-users wrote:
Marvin Jones via tde-users wrote:
Quite right. Train medical AIs on research papers and medical texts; train
legal AIs on lawbooks and court records; etc. Training AIs on the noise from
social media is at best a waste of time, and dangerous in some situations
(e.g. law enforcement).
Leslie
--
Platform: Linux
Distribution: openSUSE Leap 15.6 - x86_64
Desktop Environment: Trinity
Qt: 3.5.0
TDE: R14.1.3
tde-config: 1.0
So, today I thought to try a Google AI search for
the issue.
I had low expectations, because Trinity is (sadly) not so
well known or employed in the wild.
There is no real AI. It is a commercial trick for statistics on steroids.
It also lies a lot (makes up things).
The great benefit is in saving time to generate new stuff (but in terms of
programming it mostly does not get it right yet). There are few application
areas (for example CAD or image and text generation) especially when
trained in closed domain, where it is really useful. For the rest it may be
right but it may be also wrong. A sober approach is more beneficial.