Bots in review comments?

[DUCK]

Has anyone else seen more of bots in the comment section? both for reviews and more obscure areas with comments, be it guides or anything else.

[DUCK]

引用自 crunchyfrog：

We can make general assumptions like that but all they are are basically guesses. We CANNOT know what their systems are, how they work, what would effectively block at the expense of genuine users and so on.

That's the point.

I could have said that differently, but think you might misunderstand me? never said anything about blocking or affecting genuine users. but as pointed out, having "stable UI", an "stable platform", and with the rise of malware software/games to AI/manual spam that will breach through the systems in one way or another.

[crunchyfrog]

引用自 DUCK：

引用自 crunchyfrog：

We can make general assumptions like that but all they are are basically guesses. We CANNOT know what their systems are, how they work, what would effectively block at the expense of genuine users and so on.

That's the point.

I could have said that differently, but think you might misunderstand me? never said anything about blocking or affecting genuine users. but as pointed out, having "stable UI", an "stable platform", and with the rise of malware software/games to AI/manual spam that will breach through the systems in one way or another.

No I know you didnt - I DID.

I'm making the point that is a common problem when you try to restrict or blokc bad actors - you get collateral damage. I thought that was obvious.

The fact is you CANNOT know unless you know their systems. It's impossible.

[miakisfan]

引用自 DUCK：

Has anyone else seen more of bots in the comment section? both for reviews and more obscure areas with comments, be it guides or anything else.

You need to prove they are responses from bots or you are just looking to get rid of people you don't agree with.

It looks like someone else is already pointing that out. Just pointing out they aren't the only person who is noticing it.

[crunchyfrog]

引用自 miakisfan：

引用自 DUCK：

Has anyone else seen more of bots in the comment section? both for reviews and more obscure areas with comments, be it guides or anything else.

You need to prove they are responses from bots or you are just looking to get rid of people you don't agree with.

It looks like someone else is already pointing that out. Just pointing out they aren't the only person who is noticing it.

Precisely.

There's also a saying that I'm fond of in the medical profession that relates to Occam's Razor and applies here.

"If you hear hooves, think horses not zebras".

The implication meaning if you have a piece of evidence or a symptom, you should always defer to the most banal and obvious cause rather than the exotic.

[Doctor Zalgo]

引用自 crunchyfrog：

Ai has a lot of major problems, but the big general ones are dilution. At the point of creation or release, AI has the entire internet of humanity to train upon (within reason of course) but as time goes on, it's OWN creations get fed back into that pool too, diluting the results which is similar to creating a photocopy, then photocopying the photocopy and so on. Eventually the results i useless.

This isn't actually true. The people that produced the paper saying this took the outputs of an LLM and fed that into a new LLM and then took the outputs of that and fed it into another new LLM and eventually said 'hey, we trained an LLM purely on low quality synthetic garbage and got garbage'.

That's like putting two people in a locked room and coming back in 5 generations time and being shocked that you've reinvented the Habsburgs.

Companies are spending billions of dollars on generating synthetic training data, its not like they forgot to read the paper before doing so.

[crunchyfrog]

引用自 Doctor Zalgo：

引用自 crunchyfrog：

Ai has a lot of major problems, but the big general ones are dilution. At the point of creation or release, AI has the entire internet of humanity to train upon (within reason of course) but as time goes on, it's OWN creations get fed back into that pool too, diluting the results which is similar to creating a photocopy, then photocopying the photocopy and so on. Eventually the results i useless.

This isn't actually true. The people that produced the paper saying this took the outputs of an LLM and fed that into a new LLM and then took the outputs of that and fed it into another new LLM and eventually said 'hey, we trained an LLM purely on low quality synthetic garbage and got garbage'.

That's like putting two people in a locked room and coming back in 5 generations time and being shocked that you've reinvented the Habsburgs.

Companies are spending billions of dollars on generating synthetic training data, its not like they forgot to read the paper before doing so.

Wrong. That's not what I'm getting at at all.

You're talking about something entirely different.

What I'm referring to is how ALL AI works right now.

At the point of creation or release, Ai has the pool of the internet to train on, right? So at that point in time, 100% of it is obviously entirely created by humanity, yes? It must be as AI didn't create before this.

So now it gets to work. People ask it things, and it creates things. That is now on the internet. So now the NEXT AI request uses the internet INCLUDING any previous AI created requests.

That's what diluting is!

So as time goes on and more AI results are created, it gets ever worse. This is really, really simple and obvious.

[DUCK]

why mix LLM's and AI? not really the main topic, although every platforms will use some data collecting AI filter system, be it banks, youtube, social media, and so on, and more tracking ID's. all AI doesnt work like LLMs, many different "AI" solutions for different things. self fuelled simulation generation will happen for some data points with only knowing "good but too few results".

[Doctor Zalgo]

引用自 crunchyfrog：

引用自 Doctor Zalgo：

This isn't actually true. The people that produced the paper saying this took the outputs of an LLM and fed that into a new LLM and then took the outputs of that and fed it into another new LLM and eventually said 'hey, we trained an LLM purely on low quality synthetic garbage and got garbage'.

That's like putting two people in a locked room and coming back in 5 generations time and being shocked that you've reinvented the Habsburgs.

Companies are spending billions of dollars on generating synthetic training data, its not like they forgot to read the paper before doing so.

Wrong. That's not what I'm getting at at all.

You're talking about something entirely different.

What I'm referring to is how ALL AI works right now.

At the point of creation or release, Ai has the pool of the internet to train on, right? So at that point in time, 100% of it is obviously entirely created by humanity, yes? It must be as AI didn't create before this.

So now it gets to work. People ask it things, and it creates things. That is now on the internet. So now the NEXT AI request uses the internet INCLUDING any previous AI created requests.

That's what diluting is!

So as time goes on and more AI results are created, it gets ever worse. This is really, really simple and obvious.

LLM's aren't just giant vaccums that suck in random internet content without regard for its source. The information being used to train them is getting more curated over time, not less.

[Doctor Zalgo]

引用自 DUCK：

why mix LLM's and AI? not really the main topic, although every platforms will use some data collecting AI filter system, be it banks, youtube, social media, and so on, and more tracking ID's. all AI doesnt work like LLMs, many different "AI" solutions for different things. self fuelled simulation generation will happen for some data points with only knowing "good but too few results".

I don't think that's a battle we're going to win unfortunately, I just say DRL these days because its faster to explain what I mean by DRL than say 'AI.. no not that one."

[CANCELCULTURE]

Bots can be both a blessing and a curse. They have the power to flood comment sections and make it difficult to decipher genuine feedback from automated spam. It seems like everywhere you turn online, there's a chance of running into one. Have you come across any recently?

Oct 13, 2024 -
The Best Game

https://steamcommunity.com/app/570/reviews/?p=1&browsefilter=mostrecent#scrollTop=53672

https://steamcommunity.com/groups/UKCS/members?searchKey=%F0%9D%92%9C%F0%9D%93%81%F0%9D%92%BE%F0%9D%93%83%F0%9D%93%80%F0%9D%92%B6

Valve has no shortage of illegal fake reviews flooded by The Best Game - look at store page graph for the days of unusual reviews activity spam that count towards the over all review score.

[Nx Machina]

引用自 CANCELCULTURE：

Bots can be both a blessing and a curse. They have the power to flood comment sections and make it difficult to decipher genuine feedback from automated spam. It seems like everywhere you turn online, there's a chance of running into one. Have you come across any recently?

Oct 13, 2024 -
The Best Game

https://steamcommunity.com/app/570/reviews/?p=1&browsefilter=mostrecent#scrollTop=53672

https://steamcommunity.com/groups/UKCS/members?searchKey=%F0%9D%92%9C%F0%9D%93%81%F0%9D%92%BE%F0%9D%93%83%F0%9D%93%80%F0%9D%92%B6

Valve has no shortage of illegal fake reviews flooded by The Best Game - look at store page graph for the days of unusual reviews activity spam that count towards the over all review score.

Illegal fake reviews?

Can you cite the law being broken or is it merely your own defintion of illegal based on criteria you?

[DUCK]

yes you do see fake reviews or "paid" reviews in some games. some sex games having a lot or spam games (previous content achievement to steam market items hunters/farmers), but that is a different topic.

[CANCELCULTURE]

The Federal Trade Commission (FTC) recently announced a rule that prohibits fake and AI-generated reviews and testimonials, and other unfair or deceptive practices involving reviews. The rule will take effect on October 21, 2024.

Oh now what is Valve to do? I'ts not like they are removing the 1000's of blatant ones weekly now on their own accord.

Tell them FEDs Valve - The Best Game - just 1000's of deceptive review bot AI scripted spam on record counting towards overall review score of your product, weekly. Valve did no wrong.

[Nx Machina]

引用自 CANCELCULTURE：

The Federal Trade Commission (FTC) recently announced a rule that prohibits fake and AI-generated reviews and testimonials, and other unfair or deceptive practices involving reviews. The rule will take effect on October 21, 2024.

Oh now what is Valve to do? I'ts not like they are removing the 1000's of blatant ones weekly now on their own accord.

Tell them FEDs Valve - The Best Game - just 1000's of deceptive review bot AI scripted spam on record counting towards overall review score of your product, weekly. Valve did no wrong.

Would those reviews YOU deem illegal fall under that? Doubtful.

[DUCK]

引用自 CANCELCULTURE：

The Federal Trade Commission (FTC) recently announced a rule that prohibits fake and AI-generated reviews and testimonials, and other unfair or deceptive practices involving reviews.

Oh now what is Valve to do? I'ts not like they are removing the 1000's of blatant ones weekly now on their own accord.

think some are already laws, also this doesn't help anyone unless you want to use your ID. most platforms have this issue, online shops too many to count, same for digital software platforms like steam, apple, google/android.