107 lines
5.1 KiB
Org Mode
107 lines
5.1 KiB
Org Mode
:PROPERTIES:
|
|
:ID: 1f142832-05f9-4280-a8ca-aa6f35209f91
|
|
:END:
|
|
#+title: AI and Death of the web as we know it
|
|
#+Author: Yann Esposito
|
|
#+Date: [2024-05-16]
|
|
|
|
- tags :: [[id:a5be1daf-1010-428f-a30f-8faf95c1a42f][blog]]
|
|
- source ::
|
|
|
|
* Constat
|
|
|
|
First, since a few years now, we are experiencing a huge acceleration of the
|
|
"enshitification" of the web.
|
|
Most of us relied on services and soon, most of these servies will be worse or
|
|
completely will not provide the same benefits as before.
|
|
|
|
A typical example, as a software engineer we often searched technical knowledge
|
|
using google. Quite often we ended up on Stack Overflow, or reddit, or twitter, etc…
|
|
Now with the promise of infinite AI Generated +Spam+ SEO content. Most search will
|
|
point to a terrible website full of ads, with the additional cost that the
|
|
content could not be trusted as it was not just copied from a reliable source,
|
|
no, worse, it will be invented by the AI that has a tendency to hallucinate its
|
|
answer and pretty often return wrong and even potentially dangerous ones.
|
|
Typically, imagine generated recipes, people made some experimentation and the
|
|
generated recipe are good to put you to the Hospital if you follow these advises.
|
|
|
|
So now, that's it. We are loosing our ability to more or less, trust random
|
|
content from the web.
|
|
Is this the end?
|
|
|
|
Perhaps.
|
|
|
|
Is there something we could do about it?
|
|
|
|
I think so yes :)
|
|
And this problem already is mostly solved using the notion of "Web of Trust".
|
|
|
|
Web of Trust is a decentralized system that help you trust resources.
|
|
But you have you word to say. For example, if you trust someone for a while and
|
|
they change, they start to put horrible ads, AI generated content on their
|
|
content. You simply "downvote" or "block" them. All your direct connection of
|
|
the network of trust will be impacted by your decision, and if enough people
|
|
like you start to dislike the new content. The content of this user will
|
|
disappear forever.
|
|
|
|
This is a bit like Reddit karma, but instead of the mechanism being centralized
|
|
and controlled by a single source. This is distributed on the customers. Some
|
|
might enjoy a user, for them that user will have a big note, for other it will
|
|
be not enjoyable and his note will be very low. So low, you will almost never be
|
|
exposed to the content produced by this user.
|
|
|
|
That will probably solve a first issue. Remove from our collective sight all the
|
|
SEO spam website/content, etc…
|
|
|
|
Now, what about discoverability? Being able to search for content using this new knowledge?
|
|
Here we have different multiple solutions:
|
|
|
|
1. Still rely on classical search engines but use a browser plugin to filter the
|
|
results with only website with a trust value that is high enough
|
|
2. Use the "Web of Trust" to the rescue. We could have servers taking care of
|
|
downloading the website from the most trusted websites (starting from a few
|
|
trusted people) and open source the algorithm so people could spawn that
|
|
system on their local computer or host it and provide their server to their friends.
|
|
And we will have a very small web at first, but with a quality value that
|
|
should be very high as compared to the "Big Web".
|
|
|
|
* BONUS
|
|
|
|
I think one issue with the "Web of Trust" is the ability for attacker to "steal
|
|
an identity" of a trusted producer and produce in its name.
|
|
In particular, if the "Web of Trust" simply uses domain names, these are know to
|
|
rot easily, and could be taken.
|
|
|
|
For this, one simple but efficient mechanism will simply be to cryptographically
|
|
sign your content.
|
|
So instead of having a "web of trust" that is using domain name, we could
|
|
additionally add GPG signatures. This could be added in the header of the HTML
|
|
pages, this way a browser that will be "Web of Trust"-friendly could display a
|
|
green mark saying "Hey this content was really produced by this user with this
|
|
value of trust".
|
|
|
|
This would probably change how we use the web, because it will forces us to
|
|
"vote" time to time. Probably with more and more subtleties. For example with
|
|
different level of like/dislike in order to be able to completely block some
|
|
sources, and not just make them less prominent.
|
|
|
|
* Last but not least some wise words from Socrates
|
|
|
|
Don't forget what Socrates has to say about the invention of writing:
|
|
|
|
"For this invention will produce forgetfulness in the minds of those who learn
|
|
to use it, because they will not practice their memory.
|
|
Their trust in writing, produced by external characters which are no part of
|
|
themselves, will discourage the use of their own memory within them.
|
|
You have invented an elixir not of memory, but of reminding; and you offer your
|
|
pupils the appearance of wisdom, not true wisdom, for they will read many things
|
|
without instruction and will therefore seem [275b] to know many things, when
|
|
they are for the most part ignorant and hard to get along with, since they are
|
|
not wise, but only appear wise."
|
|
|
|
We put our confidence in a shared memory, it was great knowledge sharing.
|
|
And with the recent changes it appears we will need to regress and use our
|
|
memory, read books, read man pages, go to official documentation website at
|
|
best.
|
|
|
|
I feel the potentially single way to solve this issue is perhaps with a "Web of Trust" that will drastically reduce the size of our shared memory.
|