« first day (4381 days earlier)      last day (33 days later) » 

 
2 hours later…
10:01
@Gao I recently watched an "Ethical AI" course. Among other things, the speaker made an interesting point.
AI answer bots have the potential to be abused into giving answers that promote certain products.
A bad actor could give their bot finetunig or even just some grounding to promote a product for a company that paid them to do so.
Suppose for example you have a bot on Steam that "suggests games based on your play list". Can you be sure the bot would work in your interest and give actual valid suggestions or would it try to get you to play whatever game they were "promoting" right now?
Noticing a site trying to show ads relevant to your profile is still easy. A bot has the potential of doing similar things in a far subtler way.
@Memor-X just wondering....
I assume I am the only one getting the reference in this image?
@Memor-X Funny thing, some even call this the SECOND death.
There was already a big drop in effort when the "web population" went from a few very invested nerds to thousands zero effort passerby.
you can partially see this in effect on sites like GameFAQS. When you compare older content to newer one, often the older content is by magnitudes far more in depth than recent content.
Gao
Gao
11:18
@SPArcheon-onstrike i don't see how that changes the status quo, unless it's masquerading as human users, but then again companies already buy fake reviews.
@Gao because the AI assistants are generally more "trusted" by your Average Joe user than an ads on a site.
Gao
Gao
i'm more worried about mass rewrite of history online, followed by information flooding (well, more like information control)
by now, even the Joe knows that "ads are targeted at you and evil" so they are wary about them. But "ChatGPT is your best friend" so everything ChatGPT says is good.
There is no "maybe ChatGPT answers are targeted to" in the picture. Average Joe won't question the accuracy of their best AI chatbot friend
Gao
Gao
not sure how successful the mass media has been in making people, especially the younger generations, believe that
@Gao considering younger generations asking ChatGPT to do their homework... quite good.
Gao
Gao
11:23
@SPArcheon-onstrike nah, i question even that
it's all a good magic show by the west
From "let's ask the bot to write my homework" to "ehy, why don't I ask the bot to suggest me a game similar to GTA" ... the jump isn't that far.
Gao
Gao
@SPArcheon-onstrike what chatgpt suggests i think are the same as most review sites you come across, it's just parroting them anyway
the help me find this manga questions haven't stopped coming to this site, so ChatGPT hasn't been able to answer that either at least
true, but
1) Average Joe thinks ChatGPT actual understands what you say and what it outputs. Joe doesn't know it is just a probabilistic parrot, it actually thinks there is an "Intelligence" that generates new, original content based on actual reasoning.
2) you assume that the LLM is trained ethically, so that whatever biasis is will develop wasn't planned beforehand. What I said instead is that an ethical point I didn't consider is that a bad actor could plant those biasis by choice to promote its own agenda.
Gao
Gao
it didn't for me, giving me an outright fake answer, so i fed it fake data to regurgitate to the next victim
For example you could "poison" a bot to promote some political view, misrepresent your opposition actions, promote certain product brands etc.
you mentioned...
11 mins ago, by Gao
i'm more worried about mass rewrite of history online, followed by information flooding (well, more like information control)
allow me an hyperbolic example: a company could train a model to say that America was discovered in 1567.
Now, this would be easy to catch but... just imagine something a little more subtle.
We already get people claiming that the holocaust was never a thing... Maybe we will soon get a bot trained to say the same.
Gao
Gao
11:34
@SPArcheon-onstrike easy to catch now, but harder and harder for younger generations who will undoubtly learn a different history
@SPArcheon-onstrike that's still not i'm worried about because it's already true before ai was a thing
true as in there are people claiming that
but changing the very text of sources used in citations
it could even be realtime change on display so you don't detect a checksum change in the files
maybe it'll become feasible in the not so far future
12:27
0
Q: an elementalist manga where female lead goes to past

hollymollyI don't remember the name of the manga, but the female lead travels back in time. In her past life, she was sacrificed for her brother because she didn't manifest water powers. After traveling back in time, she develops incredibly strong water magic and forms a pact with the Spirit King. She also...

 
3 hours later…
15:01
@SPArcheon-onstrike Doctor Mario?
 
3 hours later…
18:42
@Peilonrayz @DaemonsMercy this is cursed on many levels
Luffy (G5) and who?
0
Q: Why doesnt Majin Buu have pointy ears if he's a demon?

PabloIn Dragon Ball Daima they mentioned all inhabitants from the Demon Realm have pointy ears. But Majin Buu is a demon from there and he doesnt have pointy ears, but "Freezer-like" ears. Why doesnt Majin Buu have pointy ears if he's a demon?

19:45
@SPArcheon-onstrike Yep

« first day (4381 days earlier)      last day (33 days later) »