As you might have gathered my name is Emerys
shey/they, just a dumb silly demigirl
existing on the interweb.
Currently studying mechanical engineering and
messing about with stuff i had learnt there such as
basic electronics and coding, doing hardware
projects whilst inhaling soldier fumes.
This page has gone through alot; github pages,
self-hosted with nginx and anubis, and for now
settling on neocities.
please dont spam reload the page it messes up
the counter, if you know a solution contact
me
I play some games in my spare time
currently playing -
- Deadlock graves, rem, victor
- Minecraft
- Cyberpunk
AI is an intresting topic, while it can be an
amazing assistive tool it isn't without its
flaws.
insert dead internet theory
Alot of the content you will see online such as
blogs posts, news articles, and social media posts
are generated by AI.
Hell even this content your reading right now has an
ai agent assisting me with my code editor helping
with massively repetitive tasks, for as long as
thats going to be a free feature...
In a vacuum this may not be an issue, but when a
study starts using a souce that was hallucinated by
an ai, and future models are trained on the same
'fake' data.
You might think its not that big of a problem, but
lets look at whats happening in the medical field
where ai is being used to diagnose conditions, where
"no accute findings"
where found at 4x the rate compared to older
generations of the same model.
Now while that may not be relevant to you there are
other issues. To which a paper from
MIT
is important.
In essence, it comes down to to maybe no shock to
some, that using ai personally to complete your work
doesnt actually teach you anything.
The study asked 3 groups to write an essay on the
same topic. One group having access to an ai agent,
second access to a search engine, and the last group
only has access to thier brain. During the study,
the brain activities of the participants were
recorded.
I recommend reading the paper and drawing your own
conclusions, but in my crude summary it showed most
brain activity in the memory retention part of the
brain, in the no help group, less in the browser
group, and least in the ai group.
In my eyes this would speak towards the groups with
the higher brain activies forming more neural
connections and retaining more memory, and such a
capactity for reasoning on the topic.
This is in the same way a coder with just a text
editor will think and reason about code, knowing how
to do things,
another one who would look at places such as stack
overflow for help, would be able to adapt others
code but would have a harder time reasoning and
solving issues, but still learning along the way
and finally a vibe coder who would be hopelessly
reliant on the ai not learning, and unable to solve
issues on their own.
OS: NixOS
WM: Niri (Wayland) w/ custom quickshell
Keyboard: Hand soldiered Nullbitsco Nibble
Rig: Samsung Galaxy Book 4 Pro + RTX 3080 Ti
eGPU
