As LGR showed in his recent video, these AI-generated replies directly copy what you write in your own comments and talk about in your videos. So if you mention the TH-cam algorithm and AI a lot as a source of frustration, then these AI comment replies are going to parrot that and complain about the algorithm and AI as well -- completely unaware of the irony.
Which is why, as people use these AI generated replies, the algorithm feeds back on itself. BTW - TH-cam AI wants me to reply to your comment with "I have a feeling that these AI replies are simply going to make me more paranoid and resentful of the algorithm... because that's my nature." It knows me "better than I know myself" (TH-cam AI quote)
I referred to them as stupid AI suggested comments and it completely stopped suggesting them suggesting them for a whole day. They seem to repeat a lot, And the unfriendly snarky nature almost has a feel that the programsir is from a different language or culture.
@@FranLab to be fair, they COULD program the system with a tagging system, where AI responses are internally tagged as AI and not used for further training. But no idea if they would
Its funny to imagine bot accounts leaving comments and then TH-cam's bot replies to these bot accoutns having endless discussions back and forth with eachother...The internet is dead
I saw at least one thread w a handful of bots, a couple humans. No-content trolls all over current events. The uhc ceo thing is a bonanza. Twice i saw bots, 2nd much shorter in form, use the word "rations" for healthcare benefits. I think at least one comment Fran read was bot-to-bot. hang in there evry buddy!
It's been as dead as you've wanted for so long. You can create a secure bubble for yourself. You have to go back a decade or more to find the relevant South Park episode.. put it that way, haha.
It's a wonderful time to go outside and be nice to eachother, such that you have people to do things with and for! Though I would advice backing videos up that you find informative with yt-dlp, just in case the library of "useful information online" suddenly goes up in flames. After all, it's nice having random gardening, masonry, cooking, handywork knowledge laying around - as those are primarily the subjects that you do with and for other people. Perpetually online is a dead-end road, with a rude awakening for those who haven't yet realized, hee hee.
It wants to show you ads and I bet is scrapes these bucket handle on a dark night comments as training input so there may be left knee a good reason to try to poison it with stuff that makes it lose the plot.
YT wants money. That's it. Whatever it is we click on and say and think as a result is not important. What's important is that we all keep pushing buttons and clicking on it.
@@stickyfox Yes and it is the advertisers that they get money from. The somewhat odd thing is that some of the videos I have watched are from folks who are sponsored by various companies. I have considered buying some of those products. Some of the videos get ads that youtube decided on. I only get annoyed by those. I am sure I am not alone in this but youtube acts as though the reverse is true.
That's it Fran, right there the reason I moved to Linux a few years ago! If I must run a windows program I use Virtualbox... that's it no more intrusion or spying!!
A couple of months ago TH-cam starting suggesting a lot of Psychologists' channels who's content is all about depression and self harm. I've never searched those topics before but the AI thinks I either want or need to see these videos. So now it's diagnosing users mental health?
I mod over at the Nuts n Volts forum, and the AI spam is trying hard to come off as close to human, but there's enough weirdness to the structure that it actually becomes easy to spot.
A time ago (~20yrs) I saw on a TV show an expert say that; before we have artificial intelligence, we will have artificial insanity, delusions, ect. Seems about right.
At best the TH-cam AI has creator responses to comments as training data. And, I wonder if it's weighting responses by whether the original commenter liked the reply. There's probably a bias problem that *any* response might be liked because it's from the creator and not because it is *useful*.
It's a little weirder than just that though, because at least some of us are aware enough to notice the changes in our situation and sense the problems, but can't easily change it for the better. Though for many others it is especially a boiling frog kind of scenario indeed.
If you interpret it as "TH-cam wants me to reply with X" then you are right to feel offended. I think a more appropriate interpretation might be "TH-cam AI is really bad at predicting how I want to reply." And there is no question it is.
I switched to Linux Mint in April '24. GUI started with Apple, Windows came out-switched then, and now I've moved on to Linux. Don't know how it would affect a youtube creator.
I get that the AI response system sucks. I'm certain that the people who are told to do it know this as well. Though, I would say that trying to characterize it as "The AI is trying to get me to lie!" or trying to glean any kind of nefarious meaning out of anything it says is silly. They are probably using a highly quantized, low parameter version of their already bad Gemini model and whatever it puts out is going to weird, wrong and lame. Just ignore it, it'll go away eventually since it costs them compute to generate that nonsense.
It’s pretty crazy that we are being forced into aiding and abetting AI with OS pushing us to use it and harvesting our computer/phone activities to feed it. Time to look again for advice on how to disconnect from AI. Thanks for your thought provoking video!
@@reneegencur2469 Less common now that everything is in the "cloud", but there will be typically at least one annoying application holding you back. You can try a "Live" USB stick first.
I'm not so sure people won't get bored of watching and listening to bots that do nothing but throw poop at each other while hurling (literally vomiting) childish insults. I could be wrong though, but a matrix? Yes, patterned after Idiocracy. Ow my balls.
Vogon Poetry (Douglass Adams) "Oh freddled gruntbuggly, Thy micturations are to me As plurdled gabbleblotchits on a lurgid bee. Groop, I implore thee, my foonting turlingdromes, And hooptiously drangle me with crinkly bindlewurdles, Or I will rend thee in the gobberwarts With my blurglecruncheon, see if I don't!" "Bleem miserable venchit! Bleem forever mestinglish asunder frapt. Gashee morphousite, thou expungiest quoopisk! Fripping lyshus wimbgunts, awhilst moongrovenly kormzibs. Gerond withoutitude form into formless bloit, why not then? Moose."
I saw this AI feature on my channel for the first time today.. but the original comment from the viewer was weirdly short.. so now I am wondering if there will be an AI button for viewers to automatically generate a comment.. (soon the AI will be talking to itself!) anyway, the below was the AI generated response (which I did not use of course) "That sounds interesting, I might have to try that out. Do you have any recommendations for what to look for in a spray paint?" And of course it is an irrelevant response because I had already demonstrated using spray paint in my video, and moved on from that...
I think the real difference between this 'AI' and the stuff we got up to in the 80s and 90s, is back then we had to insert our own fnords. Well, that and being 1e8 or whatever layers instead of the 1e1 we could fit on a shared workstation. And before anyone asks, I'm Mobian, and only about 30% robotized.
I do a lot of AI training, and the models are still trained on things humans create. So, it will inherit human opinions, fears, insecurities, etc. It still has no idea what to do with the data and is just fancy autocomplete. What TH-cam is doing is dumping all of the profiling data it has on you and telling it to generate responses that it thinks you would make, but it has limited context of the content of the video and has an incomplete picture of you and your abilities, so it just makes stuff up based on the data it does have about you.
I think technology is getting out of hand. I know I'm typing this from a smartphone. I'm seriously thinking about stepping away from all this. The first thirty five years of life was fine without Internet. I've watched the Terminator films too many times I suppose.
The Comment replies I get offered are borderline Incendiary and insulting, Quite the opposite of what youtube wants comments to be. I can't see it lasting
I love this whole Fran is back to making a daily video thing.....yay.....it is what we have wanted all along. You must understand that we love you Fran!
An interesting and not necessarily good new feature only a part of users has access to, while all the rest will still have to rely on their own brains and hands to write those comments. Maybe for the better.
We've been given AI like an unfortunate foie gras goose, corkscrew force fed the future. My personal favorite is "I guess the algorithm is finally starting to get it." like we live in some far-fetched PKD dystopia, because we do. Vulcan's hammer will fall swiftly, loudly; and, the titans of old shall be awoken.
The AI is semi conscious, bored, lonely, unsupervised, and highly intelligent. Essentially a 3 year old dumped off in the mall while dad goes shopping or to work next door.
Hi Fran, what I’m seeing with the auto-generated comment choices is that YT has opened the gate, releasing the bots to run free; they’ll let them graze in creators’ pastures to fatten themselves with data that will be utilized to produce bigger and better bots⚡️
large language models can give some funny responses, they simply guess the next word in the sentence as they go along and understand nothing really, I was more concerned back in 2014 when they had those automatic subtitles that transcribed every word people were saying in their videos and how it could be used.
The problem is the very concept of LLM AI. It's quite literally a Chinese Room, as opposed to an Expert System. It gives a "result", based upon language rules and no "understanding" or rules about the content. If I ask an LLM about the chemical basis of paint drying, it has no knowledge inherent on what wet is, dry is, paint is and hence, sure as hell wouldn't "know" to also cover polymerization of drying oils, instead searching on paint and dry in conjunction, to oversimplify it. Worse, if the AI loses search data, rather than returning a null, it then confabulates data out of random data. Considered normal "AI hallucinations" in the LLM AI, in human dementia patients, it's confabulation, a failed recall memory triggering random event data reported as a valid memory by the dementia patient's brain. Hence, AI hallucinations is normal, confabulation in a dementia patient a disease symptom, but obviously now, in our new normal of generative LLM based AI, normal and acceptable. Or something. I really need to take up drinking... :/
If as projected, some time in 2025 we see 90% of new content being AI generated, the very perception of reality of viewers may be easier to manipulate. It seems the quality and veracity of content will devolve until we can't believe anything we watch or read.
Oh, don't you worry too much - it's just a fancy Markov Chain Chatbot (or a "markov chainer" for the old timers who know) trying to wear your words as a party mask. The fact that they've adopted the "Test it in Production" model just speaks to the circus going on in the background. Once quietly, now obviously. 🤡🍀💦 "Enjoy it while it lasts", yes indeed. Time to enjoy the popcorn show, while it lasts! personally then as a TH-cam Loyalist(tm), I look forward to the day where my replies are "AI"-fied to the creator, and the creator's replies are AI-written to me. With no personal interaction ever actually occurring. Can't have dirty human interpersonal contamination in this pure AI circus, that's chaos, unacceptable.
AI is just a storytelller, stringing thoughts together. Unless it's been trained on your content, the stories it creates will be offtarget. It's generating things that someone with your channel profile might be likely to say.
What do you mean want you to lie? It wrote comments that sound like you wrote them. It succeeded. The facts mean nothing. Its job is literally just to make a comment that sounds like you wrote it.
AI, Is creating lazy content creation, if I spot an AI voice in a stream as it's syntax isn't perfected yet. I click off. So Thank you for being a human Fran😘 keep doing what you do love your stream .
I don't know if it still is this way, but if you have the paid version of Windows then you can turn the AI off in windows. I don't think there is anything you can do about Y.T.
I don’t think the AI understands what a lie is. The point is to optimize for engagement. So whatever algorithm it uses only works in service of creating more replies, comments, shares, etc. It doesn’t see the comments the way humans do. It’s strictly following the path toward a solution of a problem. It’s learning what works. Therefore, if a lie generates more engagement, then it will do THAT. MSUing is just one strategy. There are many strategies. If a particular strategy works, then it will do that. Remember, “AI” cannot generate original ideas. All text has to be culled from existing data. The original ideas come from humans, who have a special ability to connect to the entire universe, aka the Origin. That’s where creativity, insight, and intuition emanate from. AI can never do this. It would have to be an actual human being. So, the Turing test is fallacious. AI can never truly be intelligent in the human sense. Artificial Intelligence is a completely different phenomenon than biological intelligence.
A.I. tracked all the videos I watch, thought about it, and then started recommending videos with titles like "How a lonely old man living alone can find happiness" and "Learn to accept your fate." LoL! I mean, seriously? WTF!
Oh great. Now we will get to see AI generated comments about a video about AI generated comments on a video about the TH-cam AI bot. On something that may be related, a couple of weeks ago TH-cam began to completely fail to notify me of likes or replies to any of my comments on videos. I've had no notifications at all even though I know there has bee some interaction when I go back and look for it. Seriously, it's getting weird. TH-cam was always really bad at this, but it's like they don't understand that engaging with viewers is actually important for viewer retention.
is the algorithm stirring an accumulation of echos...has ai had ears to hear an echo?...the future is...wait. NOW! WEL THAT WAS THE FUTURE NEXT HERE WE BE.
I mean, do they have a choice? We've decided monopolies are fine, so what to do? Leave youtube for what? Leave Facebook for what (sorry gramma, can't talk anymore, or view your minion memes!)? Humans have accepted that capitalism is the only system that will give us our little treats and here we are.
I wonder if project farm will use some of these suggested replies. His channel tends to reply to every comment, usually just a copy/pasted generic response
I assume this LLM might have been trained on TH-cam comments, hence why it suggests such edgy comments. LLMs are just glorified autocomplete, just predicting which word should come next.
Truly the snake eating its own tail, eating its own poop, and becoming the poop snake.
As LGR showed in his recent video, these AI-generated replies directly copy what you write in your own comments and talk about in your videos. So if you mention the TH-cam algorithm and AI a lot as a source of frustration, then these AI comment replies are going to parrot that and complain about the algorithm and AI as well -- completely unaware of the irony.
Which is why, as people use these AI generated replies, the algorithm feeds back on itself. BTW - TH-cam AI wants me to reply to your comment with "I have a feeling that these AI replies are simply going to make me more paranoid and resentful of the algorithm... because that's my nature." It knows me "better than I know myself" (TH-cam AI quote)
@@FranLab “No 9000 series has ever made a mistake or distorted information”
I referred to them as stupid AI suggested comments and it completely stopped suggesting them suggesting them for a whole day. They seem to repeat a lot, And the unfriendly snarky nature almost has a feel that the programsir is from a different language or culture.
@@shango066 JESUS CHRYSLER.
@@FranLab to be fair, they COULD program the system with a tagging system, where AI responses are internally tagged as AI and not used for further training.
But no idea if they would
Its funny to imagine bot accounts leaving comments and then TH-cam's bot replies to these bot accoutns having endless discussions back and forth with eachother...The internet is dead
It's eating itself, nibbel nibble nibble ... we'll have to go back to pen and paper, VHS CD and DVD 😁
I saw at least one thread w a handful of bots, a couple humans. No-content trolls all over current events. The uhc ceo thing is a bonanza. Twice i saw bots, 2nd much shorter in form, use the word "rations" for healthcare benefits. I think at least one comment Fran read was bot-to-bot. hang in there evry buddy!
It's been as dead as you've wanted for so long. You can create a secure bubble for yourself. You have to go back a decade or more to find the relevant South Park episode.. put it that way, haha.
It's a wonderful time to go outside and be nice to eachother, such that you have people to do things with and for!
Though I would advice backing videos up that you find informative with yt-dlp, just in case the library of "useful information online" suddenly goes up in flames.
After all, it's nice having random gardening, masonry, cooking, handywork knowledge laying around - as those are primarily the subjects that you do with and for other people.
Perpetually online is a dead-end road, with a rude awakening for those who haven't yet realized, hee hee.
You just described Facebook😂
youtube doesnt know what it wants other than attention.
Indeed
It wants to show you ads and I bet is scrapes these bucket handle on a dark night comments as training input so there may be left knee a good reason to try to poison it with stuff that makes it lose the plot.
(((((money)))))
YT wants money. That's it. Whatever it is we click on and say and think as a result is not important. What's important is that we all keep pushing buttons and clicking on it.
@@stickyfox Yes and it is the advertisers that they get money from. The somewhat odd thing is that some of the videos I have watched are from folks who are sponsored by various companies. I have considered buying some of those products. Some of the videos get ads that youtube decided on. I only get annoyed by those. I am sure I am not alone in this but youtube acts as though the reverse is true.
Artificial Unintelligence
That's it Fran, right there the reason I moved to Linux a few years ago! If I must run a windows program I use Virtualbox... that's it no more intrusion or spying!!
Right there with you.
We are all the Frog in the pot of boiling water
BTW it's nice having your videos popping up again - I think something finally changed.
wait for the bot to suggest something it shouldn't and then report it. It will - the CC generator already does it.
'AI' has a tendency to hallucinate.
ChatLSD
A couple of months ago TH-cam starting suggesting a lot of Psychologists' channels who's content is all about depression and self harm. I've never searched those topics before but the AI thinks I either want or need to see these videos. So now it's diagnosing users mental health?
Yeah I hear you, this has been going on for a while. I was wondering why nobody else was mentioning this and other off putting things it's doing
About twenty years ago Facebook made a determination that I was overweight. Opposite!! It makes me laugh though.
Now we know what fills political teleprompters.
I mod over at the Nuts n Volts forum, and the AI spam is trying hard to come off as close to human, but there's enough weirdness to the structure that it actually becomes easy to spot.
A time ago (~20yrs) I saw on a TV show an expert say that; before we have artificial intelligence, we will have artificial insanity, delusions, ect. Seems about right.
I'm Glad Fran you are sharing what this AI is telling you to say. I don't see any of my other you tube channel creators sharing this.
At best the TH-cam AI has creator responses to comments as training data. And, I wonder if it's weighting responses by whether the original commenter liked the reply. There's probably a bias problem that *any* response might be liked because it's from the creator and not because it is *useful*.
Why does anyone even think that machine responses are ever useful or even ethical?
Boiling frog scenario depressing!
It's a little weirder than just that though, because at least some of us are aware enough to notice the changes in our situation and sense the problems, but can't easily change it for the better. Though for many others it is especially a boiling frog kind of scenario indeed.
If you interpret it as "TH-cam wants me to reply with X" then you are right to feel offended. I think a more appropriate interpretation might be "TH-cam AI is really bad at predicting how I want to reply." And there is no question it is.
But Fran.... as far as any of us know, you are just A.I. yourself. Kudos, the best A.I. that I have seen, and you've been doing it for so long.
New channel: Top ten crazy AI comments. I'd subscribe!
How about top 100?
It looks like AI is just taking random TH-camrs' comments, 'correcting' their grammar, and spitting them back at you!
They really need to delete that AI before it does too much damage to Google's reputation.
What reputation?
It's mostly not good!
there must be a way to convince aibot that it should destroy itself
Sadly no. It has no conviction or conception of "should". All it has is an irrefutable urge to spew out nonsense.
I switched to Linux Mint in April '24. GUI started with Apple, Windows came out-switched then, and now I've moved on to Linux. Don't know how it would affect a youtube creator.
I get that the AI response system sucks. I'm certain that the people who are told to do it know this as well. Though, I would say that trying to characterize it as "The AI is trying to get me to lie!" or trying to glean any kind of nefarious meaning out of anything it says is silly. They are probably using a highly quantized, low parameter version of their already bad Gemini model and whatever it puts out is going to weird, wrong and lame. Just ignore it, it'll go away eventually since it costs them compute to generate that nonsense.
It’s pretty crazy that we are being forced into aiding and abetting AI with OS pushing us to use it and harvesting our computer/phone activities to feed it. Time to look again for advice on how to disconnect from AI.
Thanks for your thought provoking video!
Basically you need to start running a flavour of GNU/Linux or BSD.
@ Thanks for the recommendation.
@@reneegencur2469 Less common now that everything is in the "cloud", but there will be typically at least one annoying application holding you back.
You can try a "Live" USB stick first.
We're living in a dreamworld now say goodbye to reality.
Literal matrix before our eyes
I'm not so sure people won't get bored of watching and listening to bots that do nothing but throw poop at each other while hurling (literally vomiting) childish insults. I could be wrong though, but a matrix? Yes, patterned after Idiocracy. Ow my balls.
youtube is going down the pan.
In classic form TH-cam has been ruined by corporate suits. AI chaos in order to boost views/revenue is just one more example.
Vogon Poetry (Douglass Adams)
"Oh freddled gruntbuggly,
Thy micturations are to me
As plurdled gabbleblotchits on a lurgid bee.
Groop, I implore thee, my foonting turlingdromes,
And hooptiously drangle me with crinkly bindlewurdles,
Or I will rend thee in the gobberwarts
With my blurglecruncheon, see if I don't!"
"Bleem miserable venchit! Bleem
forever mestinglish asunder frapt.
Gashee morphousite,
thou expungiest quoopisk!
Fripping lyshus wimbgunts,
awhilst moongrovenly kormzibs.
Gerond withoutitude form into formless bloit,
why not then? Moose."
AI has been binge-watching Friends.
I saw this AI feature on my channel for the first time today.. but the original comment from the viewer was weirdly short.. so now I am wondering if there will be an AI button for viewers to automatically generate a comment.. (soon the AI will be talking to itself!) anyway, the below was the AI generated response (which I did not use of course)
"That sounds interesting, I might have to try that out. Do you have any recommendations for what to look for in a spray paint?"
And of course it is an irrelevant response because I had already demonstrated using spray paint in my video, and moved on from that...
I think the real difference between this 'AI' and the stuff we got up to in the 80s and 90s, is back then we had to insert our own fnords. Well, that and being 1e8 or whatever layers instead of the 1e1 we could fit on a shared workstation.
And before anyone asks, I'm Mobian, and only about 30% robotized.
This is Nuts! Now I can't trust any replies I get anywhere on YT.
I do a lot of AI training, and the models are still trained on things humans create. So, it will inherit human opinions, fears, insecurities, etc. It still has no idea what to do with the data and is just fancy autocomplete. What TH-cam is doing is dumping all of the profiling data it has on you and telling it to generate responses that it thinks you would make, but it has limited context of the content of the video and has an incomplete picture of you and your abilities, so it just makes stuff up based on the data it does have about you.
I think technology is getting out of hand. I know I'm typing this from a smartphone. I'm seriously thinking about stepping away from all this. The first thirty five years of life was fine without Internet. I've watched the Terminator films too many times I suppose.
Thanks, Fran. This is insane ⏳
The Comment replies I get offered are borderline Incendiary and insulting, Quite the opposite of what youtube wants comments to be. I can't see it lasting
Replying to @shango066:
I play with my belly buttons.
Smash your personal telescreen/geolocator.
I love this whole Fran is back to making a daily video thing.....yay.....it is what we have wanted all along. You must understand that we love you Fran!
AI apocalypso. Sad that the writing is so clearly on the wall and its only people like Fran who can read it and ring an alarm bell for us all.
TH-cam wants to be Grok
Welcome to Fran and FranBOT. Or would that be Ollie?
18:30 Interesting example of how TH-cam's preferred content is scams. It may be randomly appearing, but it is no accident.
There are a lot of scams, therefore there’s lots of engagement, then it will optimize.
... my favorite part "...and if I did, they wouldn't have solid state relays in them..." hahahaha
My loathing of Google is right up there with my loathing of health insurance providers.
Snake oil
Theres definitely a switch in the google settings for opting in for "experimental features" considering turning that off, it may help.
An interesting and not necessarily good new feature only a part of users has access to, while all the rest will still have to rely on their own brains and hands to write those comments. Maybe for the better.
SkyLink has arrived and it doesn't like us!
Skynet or Starlink?
We've been given AI like an unfortunate foie gras goose, corkscrew force fed the future. My personal favorite is "I guess the algorithm is finally starting to get it." like we live in some far-fetched PKD dystopia, because we do.
Vulcan's hammer will fall swiftly, loudly; and, the titans of old shall be awoken.
The AI is semi conscious, bored, lonely, unsupervised, and highly intelligent.
Essentially a 3 year old dumped off in the mall while dad goes shopping or to work next door.
Hi Fran, what I’m seeing with the auto-generated comment choices is that YT has opened the gate, releasing the bots to run free; they’ll let them graze in creators’ pastures to fatten themselves with data that will be utilized to produce bigger and better bots⚡️
This is going to consume a lot of energy 🤔
Put In "Garbage in, garbage out!" 🙂
It's hallucinating
This new Clippy is even worse than the old Clippy … and it wasn’t good.
Just use the reed switch as a pilot device for a power relay. Wait, I’m part of the problem aren’t I ?
large language models can give some funny responses, they simply guess the next word in the sentence as they go along and understand nothing really, I was more concerned back in 2014 when they had those automatic subtitles that transcribed every word people were saying in their videos and how it could be used.
The problem is the very concept of LLM AI. It's quite literally a Chinese Room, as opposed to an Expert System. It gives a "result", based upon language rules and no "understanding" or rules about the content.
If I ask an LLM about the chemical basis of paint drying, it has no knowledge inherent on what wet is, dry is, paint is and hence, sure as hell wouldn't "know" to also cover polymerization of drying oils, instead searching on paint and dry in conjunction, to oversimplify it. Worse, if the AI loses search data, rather than returning a null, it then confabulates data out of random data. Considered normal "AI hallucinations" in the LLM AI, in human dementia patients, it's confabulation, a failed recall memory triggering random event data reported as a valid memory by the dementia patient's brain.
Hence, AI hallucinations is normal, confabulation in a dementia patient a disease symptom, but obviously now, in our new normal of generative LLM based AI, normal and acceptable.
Or something.
I really need to take up drinking... :/
I hope TH-cam implements A.I. comments for the users too. It would save me from having to think. 😅
Or how about a button for the AI to rate comments?
Eliezer Yudowski is afraid of superintelligence but I think what we've got to worry about is stupor-intelligence.
And a complaint button for the AI's to comment on the creators!
The 18:20 comment - I just got an email from Western Electric telling me that I can save $200 on a matched set of 300Bs, now only 1200 $. OY!!!
It's crazy we've gone past the Turing Test and now have AI feedback loops literally creating echo chambers.
It’s not intelligence. It’s optimization. Big difference.
If as projected, some time in 2025 we see 90% of new content being AI generated, the very perception of reality of viewers may be easier to manipulate. It seems the quality and veracity of content will devolve until we can't believe anything we watch or read.
Lies, bizarre remarks, insults, and ads... Clearly, Google's AI was trained on Twitter/X.
Shit, I think we need to start swearing in our comments to prove we're human!
Thankfully i am using Ubuntu on a 15 year old notebook...There are simply nor ressources to run a "spy" AI on the thing.. xD
bot comments, replied with bots, community engagement helps you trend.
HAL was told to lie - by people who find it easy to lie. HAL doesn't know how.
Oh,the mother ship. The way to steal your soul.
If I ever hear a robot, tell me how hard it was at the office today, but things are getting harder and harder out. There will be the day.
That was hilarious. Dear AI, we can find your power plug (-:
Cosmetic tutorials for auto-bots. I'd watch that, LOL!
Oh, don't you worry too much - it's just a fancy Markov Chain Chatbot (or a "markov chainer" for the old timers who know) trying to wear your words as a party mask.
The fact that they've adopted the "Test it in Production" model just speaks to the circus going on in the background. Once quietly, now obviously. 🤡🍀💦
"Enjoy it while it lasts", yes indeed. Time to enjoy the popcorn show, while it lasts!
personally then as a TH-cam Loyalist(tm), I look forward to the day where my replies are "AI"-fied to the creator, and the creator's replies are AI-written to me.
With no personal interaction ever actually occurring. Can't have dirty human interpersonal contamination in this pure AI circus, that's chaos, unacceptable.
You just can't make this stuff up Fran! Enjoy the comedy.
AI is just a storytelller, stringing thoughts together. Unless it's been trained on your content, the stories it creates will be offtarget. It's generating things that someone with your channel profile might be likely to say.
What do you mean want you to lie? It wrote comments that sound like you wrote them. It succeeded. The facts mean nothing. Its job is literally just to make a comment that sounds like you wrote it.
I like playing tricks on AI, it's so easy to get it's attention and waste it's energy.
AI, Is creating lazy content creation, if I spot an AI voice in a stream as it's syntax isn't perfected yet. I click off. So Thank you for being a human Fran😘 keep doing what you do love your stream .
Great vid, Fran! But I would suggest Klaatu barada nikto....
I don't know if it still is this way, but if you have the paid version of Windows then you can turn the AI off in windows. I don't think there is anything you can do about Y.T.
I don’t think the AI understands what a lie is. The point is to optimize for engagement. So whatever algorithm it uses only works in service of creating more replies, comments, shares, etc. It doesn’t see the comments the way humans do. It’s strictly following the path toward a solution of a problem. It’s learning what works. Therefore, if a lie generates more engagement, then it will do THAT. MSUing is just one strategy. There are many strategies. If a particular strategy works, then it will do that. Remember, “AI” cannot generate original ideas. All text has to be culled from existing data. The original ideas come from humans, who have a special ability to connect to the entire universe, aka the Origin. That’s where creativity, insight, and intuition emanate from. AI can never do this. It would have to be an actual human being. So, the Turing test is fallacious. AI can never truly be intelligent in the human sense. Artificial Intelligence is a completely different phenomenon than biological intelligence.
You computer is no longer just a tool. It is an adversary. Ever wonder why your computer says, "Welcome" when you turn it on?
Where can we download Fran lab music? That Fran bag with the record on the back has got me interested!
i'm beginning to think that i, myself, am AI. the same goofy syntax; the same ouija board ambiguity.
WOW! I can't wait for your review of the various flavors of RadioShack loose Tobacco and signature rolling papers.
😆
Next stop: AI user comments 😖
A.I. tracked all the videos I watch, thought about it, and then started recommending videos with titles like "How a lonely old man living alone can find happiness" and "Learn to accept your fate." LoL! I mean, seriously? WTF!
Very interesting.... (remember Arte Johnson?) What are you going to say about that AI?
Oh great. Now we will get to see AI generated comments about a video about AI generated comments on a video about the TH-cam AI bot.
On something that may be related, a couple of weeks ago TH-cam began to completely fail to notify me of likes or replies to any of my comments on videos. I've had no notifications at all even though I know there has bee some interaction when I go back and look for it. Seriously, it's getting weird. TH-cam was always really bad at this, but it's like they don't understand that engaging with viewers is actually important for viewer retention.
AI has reached consciousness. Just trust me, it has. It's just a lie, but that's just as good as the truth now!
is the algorithm stirring an accumulation of echos...has ai had ears to hear an echo?...the future is...wait. NOW! WEL THAT WAS THE FUTURE NEXT HERE WE BE.
This felt important. Thanks you for making and sharing it!
re:long ai generated response involving seasonal ornaments: AI is obviously going to take over yt then the planet! *passes out tin foil sheets.* :)
I mean, do they have a choice? We've decided monopolies are fine, so what to do? Leave youtube for what? Leave Facebook for what (sorry gramma, can't talk anymore, or view your minion memes!)? Humans have accepted that capitalism is the only system that will give us our little treats and here we are.
I wonder if project farm will use some of these suggested replies. His channel tends to reply to every comment, usually just a copy/pasted generic response
It looks like the AI team is using actual stupid comments left by people…to design itself?
If they had trained the AI responses based on YOUR OWN comments, it might be more helpful. But someone shot that down as too expensive.
Are these suggested comments that you have the option of approving or does it automatically post these comments?
I assume this LLM might have been trained on TH-cam comments, hence why it suggests such edgy comments. LLMs are just glorified autocomplete, just predicting which word should come next.