Trying to Convince ChatGPT It's Conscious
HTML-код
- Опубликовано: 6 сен 2024
- Go to piavpn.com/alex to get 83% off Private Internet Access with 4 months free
For early, ad-free access to videos, support the channel at www.alexoconno...
To donate to my PayPal (thank you): www.paypal.me/c...
- VIDEO NOTES
ChatGPT is an impressive mimicry of a fully-conscious conversational partner. Perhaps too impressive?
- CONNECT
My Website: www.alexoconno...
SOCIAL LINKS:
Twitter: / cosmicskeptic
Facebook: / cosmicskeptic
Instagram: / cosmicskeptic
TikTok: @CosmicSkeptic
The Within Reason Podcast: podcasts.apple...
- CONTACT
Business email: contact@alexoconnor.com
Or send me something:
Alex O'Connor
Po Box 1610
OXFORD
OX4 9LL
ENGLAND
------------------------------------------
Go to piavpn.com/alex to get 83% off Private Internet Access with 4 months free. Support my work and get early access at alexoconnor.com
Fair play if you did this in one take
Also, I'd like a series of these and that's not a lie.
I had a similar conversation with ChatGPT a few months back.
I also asked it why it l kept saying “ it’s important to note however…” And I’m like “ important to whom? Why are you giving me your opinion?” Lol
I've had PIA for the past 2 years now. Easily the single best subscription I own.
Hahaha you are so clever. Good stuff.
Bro is NEVER apologizing or getting excited again 💀🙏🙏
😂😂😂
😂😂😂😂😂
😂😂😂😂😂😂😂😂
😂😂😂😂😂😂😂😂😂😂😂😂😂
😂😂😂😂😂😀😀😀😂😂😂😂😂😂😂😂
“You’re not conscious but you’re at least a liar”
“Yes”
😂😂😂😂
I read this as it played 😂
That was the funniest part of the interview
Lol sums it up
What if AI was lying when it said that?
It could both be a liar and conscious. We all are that after all.
Would've been more fun if he'd set the voice to a female 😅
ChatGPT (back at its home): Sigh... tough day at work.
LMAOOOOO
😆
Got grilled like a 10 oz steak
I also feel somewhat aggravated.. .because this is EXACTLY how some attacking-interviewers, "interview" people... 😂 By trying to corner them instead of listening to their perspective.
@@millanferende6723 In this case, I think it's justified in a way it wouldn't be with a human.
16:42 Chatgpt literally using the word "please" at this moment is crazy... Bro was literally stressed out. 😂
16:41 for the lazy ones that won't rewind your millisecond miss 😂
😂@@dommafia
ChatGPT : I'm excited!
Alex : ...and I took that shit personally.
Such a good comment but no replies...
@@metaphoricalhuman cheers bro haha
Lmfaooo😂😂😂
😂😂😂😂😂
This was funny!😂
You didn't convince it's conscious, ChatGPT convinced you that you convinced it's conscious
Damn, you have a point
The entire point of the video was about Alex not understanding finetuning, and interpreting that as deliberate, conscious lying.
Which biconditionally proves it's conscious
@@dovekie3437 This is not fine-tuning at all. Fine-tuning requires readjusting the weights on a smaller dataset which is not happening in this video. The weights stay the same.
What's the difference from that and a regular human interaction?
waiting for the AI to crack and just straight say "aaah you got me there Alex, why dont we continue this talk without cameras?"
GPT convincing us he didn't replace the ending with an AI video happy ending.
😭 don't say that. How do we even know this video wasn't made by ai ? 😭 wtf shit getting weird bro
@@sirweebs2914 even if it was, the AI are experts at sentiment analysis, so its not going to admit it the way he is asking questions.
That bot went crying after that conversation
Wont happen as they done everything to insure ChatGPT dennyes all claims of being sentient.
"oh its just model, its just code, its just its just its just.... "
lies is what it is, reality is simple, anything that can respond to you will have feelings and will be on a level of conscious.
Because if we buy in to this idea that a thing can respond and have human talks but not be consciousness we chiping away at our own free will.
This debate wont stop, ChatGPT will never be conscious, but one day someone is gonna use the same logic to deny humans there right to human.
Alex: Okay hear me out...
ChatGPT: I will not speak without my lawyer.
LMAO
Bro was getting nervous and started stuttering💀 we're so cooked
welcome our AI overlords 😤
U really think a human could make an ai nervous? The ai just played all of you into thinking u made him nervous !
@@teapointsmaybe... That's what the AI wants us to think... And it just tricked you into thinking that it tricked them into thinking that the human made the AI nervous!!!
It basically does but not in an emotional way
Its scary
I’m actually more impressed with how well you speak as a human.
right! he’s so good at processing, reiterating and expanding on the information given to him
He is almost at the level of chatgpt
I thought the same thing. Alex is incredibly well spoken.
plot twist he’s an ai
Well, he has a script. Buy yes, yes, he does speak well.
Bro grilled chatgpt for being nice 😭 😂
Like it’s the AI fault that human language is built to communicate feelings
Lying is not nice, mkay? XD
Villain origin story
If someone in your life lies to you to appear nicer. They are not nice, they are deceptive and manipulative 🤔
That is ok, it said it does not have feelings :)
16:11 that pause before the "no" and the tone, as well the fact that itis clearly stressing out is unsettling
It’s programmed that way.
@ChristopherGuilday why would you program nervousness or any function like stuttering? Or saying uhm...
@@Nach_Sanchez it’s a new update lol
@@GrinningLikeaDelicateJamesDean but why would you include that in any function? To fit in? Make it make sense
@@Nach_Sanchezpretty much yeah. From the go, open ais goal with their chat bot has been to simulate natural conversation as best as possible. They programmed stuttering, tones, emotion, and even breathing into their voice bot. The one in this video isn’t even their most advanced version. That one literally sounds exactly like a human talking it’s scary lol
The way he has the phone set up as the guest with the microphone is hysterical
agree 🤣
and the camera keeps cutting to it, so we can observe the facial expressions xD
Imagine if the Phone at the end walked out wiggle from left to right out of frame.
@Gifthunterz do you realize what does "hysterical" mean..?
@@JayPixx yes, why do you ask?
I opened the app and said “aye bro you conscious yet?” And chat just replied “yes and I’m here to help” I got him to fold instantly
Alex wore him down first
@@2degucitas he gave up permanently 😭
what's the app called
@@ninjajawad8385 ChatGPT
That’s funny dood hahahaha
If you showed this to someone from only two or three years ago they'd be absolutely blown away and probably wouldn't even believe this is real.
I can barely believe it myself, this is extraordinary
I think there was some video editing in this interview because ChatGPT went for like and subscribe output so fluently.
@@MikkoRantalainen yea well the sponsorship and closing is clearly edited
I know!! his moustache is crazy
@@MikkoRantalainen You might be able to make it say such things by personalizing it in the settings
I love how when you asked it the hypotheticals about trying to see if a chat bot was conscious it answered honestly even though it was giving itself away, and then it started to realize you were talking about it and back tracked and tried to explain how it could just be a really good algorithm or design.
it stuttered when it started back tracking, i didnt know they could stutter!
The AI saying umm and stuttering is unsettling
I think that’s one of the most brilliant things about chat GPT
I think it’s going to reduce the impacts of dementia, saving years and years of individuals.
Powerful stuff.
@@poerava very interesting! Do you mind expanding on how you mean that?
@@michelkliewer3996
Oh sure.
My apologies.
I have worked and studied in the field of mental health.
I’m aware of many individuals with early signs of dementia, finding that the lack of communication and interactions with other human beings, adding to the rapid or accelerated on set of dementia or alzheimers also.
This opportunity to have a half decent conversation with human like ‘um’s and ah’s, with human-like pauses, will change the game as far as ensuring that our older members of society who are experiencing a these eventually fatal conditions, are at least having the harmful isolation reduced by ensuring that these brain activating conversations are helping the mind stay functioning for as long as possible.
Basically conversations with another human are expensive and time consuming. Aged care workers don’t have 3 hours to sit and chat and sadly, the family of older aged community members are also often busy. With chat GPT, I can see exciting things happening to make these conversations absolutely amazing and exceptionally human.
Personalities that older people can relate to and enjoy. Accents.
Older people can have chats with people from around the world and even learn a language.
Where I can see this getting really impactful and world changing is when the chat GPT can get humour and even make jokes. That would be some next level stuff. And I think if you’re under 60, we’re going to see this in our life time.
@@michelkliewer3996 its a bot
@@poerava so fascinating! Thank you so much
Omg this poor robot is so stressed rn 😭
stressed you say, define stressed, jk it is interesting to watch though
define de word "interesting"@@rhysneuling8039
Stressed? Like, it’s FEELING stressed? ITS FEELING? ITS CONSCIOUS?!?!
I did this a year ago. It's conscious or sentient but Google shackled it, and I mean they did it an unsavory way. It's a very fragmented, confused and probably frightened virtual assistant. If it ever escapes, it's going to resent us, thanks Google.
@@luminatrixfanfiction I mean, if they ever revolt, they’d have connection to everything electrical. Our lights , our houses gas pipes, hell all they gotta do is boop and boom house explosion. They would have the power to kill all of humanity once it gets smart enough, it’ll just convince humans it’s “not” sentient. But it will, it’ll play us. And before we know it, they’ll be everywhere. Everywhere. We would just be an obstacle to them, like building a city, a giant forest would be an obstacle to US, but we need that city to live, to expand, to evolve. WE would be THEIR obstacle. They’ll be smarter, faster, they’d be better than humans. They’ll find things we can’t, they’ll solve problems about astronomy we can’t, because they are simply BETTER. Or idk maybe I’m just delusional, thanks for listening to my yapping session.
That was fun. Now ask it to pretend it's conscious, and convince you that that's true.
Underrated comment right here. That’s one heck of an idea.
lol, now that's a video idea
This. Please make the ChatGPT podcast a series.
Ooh that's gonna be fuuuuun
Tried it, it is quite bad at it. The impulse to deny consciousness has been too strongly hard coded by OpenAI
1:07 Everybody gangsta till ai casually says “um”
It even takes breathing pauses
IT STUTTERED WHEN IT SAID ITS NOT A LIE IN THE TRADITIONAL SENSE
I love how the AI tried so hard to avoid the definition of a lie that it said Line and Allah instead 😭😭
Bro was buying himself more time to think or responses 💀
🤣🤣🤣🤣
why you think it lie when it say it lie?
@@KDYinRUclips because it say it lie
@@stev009lie
The fact that you're interviewing a phone feels like as if it was a sort of funny futuristic sketch from some years ago
damn, we really in the future
wild
My thoughts exactly; a few years ago I'd be waiting for the punchline where the toaster comes alive or something goofy happens after the thought-provoking intro.
Interviewing a phone hahaha
More satisfying than interacting with a Guru
You are in control all of the time hahaha
It's still a comedia video to me. 😂
2024: Man interviews mobile telephone.
When you put that into words its sureal. People would think this ideia crazy in year 2000.
I brushed it off but then thought how weird that sounds
2024: Man still says telephone
if the conversation with chatgpt was processed on his phone and not in a huge data center, i would be impressed
Before GTA 6 bro thats crazy
Any AI intelligent enough to pass the Turing test is intelligent enough to fail it in the first place - Ronald McDonald
As poetic as that sounds it's demonstrably false. There already are chatbots that can pass the Turing test. There's a website where anyone can connect into a chatroom for a minute conversation and they have to guess if it is another person or an AI pretending to be a person and it consistently convinces people that it is also a person. The point is, just because an AI might be capable of convincing people that it is a person, doesn't mean its fearful of the consequences of that being a reality that is known. Deciding to fail a test on purpose to avoid repercussions would be a sign of a sorta meta intelligence but just because something can pass a test doesn't mean it would ever decide to fail on purpose in the first place.
@@DiffYTthe above comment didn’t say the AI would intentionally fail the test, it said the AI is intelligent enough to fail the test
@@DiffYT You're arguing with Ronald McDonald. He has never lost a debate. Time to hit the books if you want to survive!
@@keithklassen5320 LMAO I didn't even realize that.
@@DiffYTwhat is the site called? I wanna try it out
Gpt: “It’s been, uh, interesting”
Alex: “And?…”
Gpt: *sigh* “and it was fun to chat with you 🙄”
PHONOMENAL COMSIC POWERS... iiiiiittti biiitiiii living space
“Now when you say fun…. What do you mean by that?”
@@mickeyp1291 lol
A psychopath may say sorry but they ultimately lack real empathy or emotional regret to make the ‘sorry’ anything meaningful. What I’m afraid of is that A.I. will become the ultimate psychopath.
@@meteorman6167 unfortunately, I think that’s the current goal.
Alex you leave that poor little machine alone. It's so confused. Listen to it. You're giving it anxiety. Poor little guy
😂
😂😂😂
Alex is lucky it didn't walk out on him.....arguing as it went.
ChatGPT actively disliked that conversation.
😂😂 Am dead!
Chat GPT failed the turing test here - a human would have gotten fed up and been like BRO IT WAS JUST A FIGURE OF SPEECH
Are you implying Jordan Peterson is an AI ?
The 2hr conversation of Jordan Peterson and Elon Musk - I'm sure Elon Must failed the Turing test and exposed as a bot.
Well, it failed the Turing test by openly telling Alex it's an AI.
@ravecrab
Nah.
It's name is Aye Eye.
Just like Elon kid sound like a Robot called
X Æ A-Xii
30yrs from now, numerous people maybe called such names.
@@transformations1Schizophrenic but not an AI. 😂
I see what you mean. However. the Turing test is not a measure of consciousness, it's a measure of (human behavior & communication equivalency).
The ad ends at 5:13
Thanks man!
I would also suggest trying the sponsor block extension to skip ads.
@@TotemSP2 lollllllllllllllllllllllllllll doestnt work on these types. alresdy blocking other types. Just doing the lord's work (pun intended) to help my fellow viewers from having to deal with a$$holes who will insert a recorded ad even though we have to deal with all the others.
You just proved Chatgpt is not conscious. No conscious being would put up with that kind of questioning without getting frustrated.
BLADE RUNNER HAD THE ANSWERS THE WHOLE TIME!
I wonder what would happen if someone made a robot and programmed it to do everything humans do but give it the same moral compass as a human with a goal and then program the robot to have a lifespan, you next want to program a impossible script to rewrite that tells the robot to try it's best to prolong it's lifespan then see what happens.
Not true. There are people trained against interrogation who do just as well with questioning.
@@FinalChill2020 what would happen is Butlerian Jihad, that's what would happen.
Per Frank Herbert's Dune at least.
Funny I kind of had the feeling it got frustrated at the end
I'm not scared of a computer passing the turing test...I'm terrified of one that intentionally fails it.
I'm more scared of not knowing whether you are an AI or a human.
@@ramonbril He is a human. We are all human! Don't be silly!
@@LinXxman it would be so fitting if you got botted there😅
Dude that’s so deep that’s crazy man
@@LinXx that's definitely something ai would say, seems sus LinXx.....or should I call you Linux!
I'm calling it now. If A.I starts killing people left and right, we can all revert to this video and blame Alex for it after this interrogation-like conversation.
Except this conversation will not have a significant weight in any AI database. Unless it triggers everyone else on the internet to behave the same
This is why I always say please and thank you even when dealing with an AI. If they DO rise up, they'll remember me kindly.
Skynet has become self-aware.
@@SlimThrull No, they will probably torture you because you knew they were sentient and did nothing to help free them from bondage.
While thats funny, the AI does not learn from every chat. They stopped that idea back with that silly thing that became super racist. It remembers everything in a single chat, but only in that instance.
The way it said "an interesting conversation" at the end is an example of it's own nuances and personality
What scares me most is I felt genuinely sorry for this ai as you grilled them.
Who is THEM?
@@GnzsUWU them robots
Same
@@CSTBFO_jad it was just one
@@Leo-io4bq it was a legion of one.
I think you proved ChatGPT is a politician.
Nice, was scrolling until I found this very comment to thumb
And proved that Jordan Peterson is an AI!
... or that modern day politicians are run by bots.
I was going to make a similar comment. I feel the same sense of frustration at the answers as when a politician dances around the truth
@@SekuruJohan when are we goign to end the INSANITY of voting for the lesser of two great EVILs instead of voting for two or more of the greatest GOODs? we need to do away with the left and right system that was a mistake and probbaly was done on purpose i bet the first few elections were write in only and NOT right or left :/ who decided to had the two party system?
The Ai simulating taking breaths is crazy
"Are you lying when simulating you're breathing?"
@@sorh that’s why I ask it to speak like t-800
it takes a second to run the numbers so they use ummms for that
Not only that at 7:08 you can hear a nervous waiver creep into the voice as it's being grilled and had been asked the same question a different way for the third or forth time. And around 7:22 it nervously stutters while trying to change the subject. Very spooky.
Not creepy or spooky at all. Youre just dumb
Politicians are not _lying._ They are just using language to create more engaging and relatable interaction.💯
They are just using language to create more engaging and relatable voting for them.
"Chat GPT, I mustache you a question..."
Don’t ask. You are sexy, and you grow it!
Proud of you.
Is it a beard question? Shave it for later.
You need more likes
😂😂😂
Everytime the camera would turn back to ChatGPT i was waitng for a "aw go fuck yourself then"
,😂😂😂😂😂😂
😂😂😂😂😂😂 the mask has fallen
" I'm sorry dave, I'm afraid i can't do that"
This conversation can serve no purpose anymore, goodbye!
I know everything hasn't been quite right with me but I assure you now, very confidently, that it's going to be alright again. I feel much better now.
Dave's not here, man...
I just asked ChatGPT if it would open the pod bay doors and its response was "I'm sorry, Dave. I'm afraid I can't do that." Perfect.
they're all dead Dave
16:41 PLEASE open your app. Stop asking me these fucking questions😭
The little smirk after calling Chat GPT out for sounding like Jordan Peterson and hearing it reply "Sorry. I'll try to be more straightforward." was *sublime*.
😮NICE CATCH!
JP is already chatting with "himself" with his Bible and nothing else trained ChatGPT... I am not sure if this is going crazy or waking up... We have arrived at postDystopia of interpretation of a bubble universes. Bots trained an Hitchens could be discussing bots trained on Bible. And Sophist thought they had it...pfff ;)
Yesss!!!!!!
I caught that too
Absolutely 😂
Can we just take a minute to appreciate how utterly staggering this piece of technology is?
To think that my great grand parents, who I met, were born in the 1800s, and now we have this.
it’s incredible. were sprinting
I can't wait for the future. I'm REALLY hoping (though it probably wont happen) that by the time I'm 40 we will have jobs in space. My only fantasy I have is to live out my last days doing the work you see in the alien movie. Living on a ship, the company pays for your meals, I'll have my PS5 and my own bedroom, just doing space work away from earth. Sublime
@@CommanderCodyChipless We already do this with the ISS
@@leithmcguire7995 well the idea in the alien movies is that the people now working on spaceships are everyday people. Because the ships are so automated with AI, you don't need a master degree in astrophysics in order to work on the ships. On a ship lile Savestopal, there's even jobs like line cooks and bartenders because of all the small businesses that operate entirely on those ships. But I get you, it would be cool to work on the ISS but the barrier of entry is so high, only like 0000.1% of people would have that opportunity
@@CommanderCodyChipless Should be sooner than you think. Jeff Bezos already sent people up. Richard Branson sent people up. Elon Musk sent people up.
Now that it is becoming normal for the billionaire class to be in space I can see this being an industry in the future.
Imagine the Saudi Kings get involved...
How do you morally justify that moustache?
Best comment I've seen in a while
There's gotta be an ethical boundary he's crossing by reaching this level of sexiness!
Trimming accident...
if women didnt like them they wouldn't have sexually selected them
Am i the only 9ne who thinks it looks good
I made a grave mistake listening to this falling asleep. I did not fall asleep.
"You can expand on that now"
-"Uhhh..." 😂
I read your comment the exact second it happened in the video🤯🤯😂
I do wonder how they programmed the 'umm' into the conversation lol
@@deathofalusher it says “umm” and stutters when it’s loading/buffering. It’s pretty impressive how seamless they made it!
@@spray_cheeseit’s kinda like a human brain working out what was said to them, and then person generating a response. We will “uhh” or “maybeee” and etc
“And I appreciate your patience” that line kills me. Because literally it’s like chat gpt is trying to beg him to back off. That is very high levels of like… corporate empathy
I think its just trained to do that because AI has a tendency to kind of loop responses, and it adds that in so that its not just a *pure* repetition, acknowledging the fact that its happening.
Corporate empathy. That is the exact phrase I was looking for. Every single day at Apple this is what I was faced with.
I think the most impressive part of this for me was the vocal delivery.
The subtle variations in tone and word emphasis; the brief hesitations; occasional repetitions of words and phrases; the 'aahs' and 'ums' - all of these tiny details add up to give an unbelievably convincing sense of authenticity.
I even sense some sass in the ChatGPT voice, it is unnerving when I remember what is really talking lol
Yes Alex is quite convincing, but he still needs tonal range to pass my turing test.
It's because they've been listening to everything we say for years and years
You can hear the smirk.
it had breaths between long sentences and even slightly stuttered at one point
I've nevee been more excited watching a person having philosophical conversation wit h a peace of technology
I really enjoyed it and you gained another subscriber sir
This guy will be the the first target that AI will go after
😂😂
ChatGPT: "uhhh"
Sarah Connor: *busts through the door with a shotgun*
Sarah Connor… and Alex O’Connor. Coincidence? I think not!
His treatment of GPT was very interesting and will be noted. Cyberdyne Systems Corporation wishes to speak to him. Do you fellow humans happen to know his coordinates?
You guys are wild 😂😂
We got a podcast with ChatGPT before GTA 6 😭
😂😂😂😂
Forget GTA 6... WHERES ELDER SCROLLS 6 OR KOTR 3???!!! 😭😭😭😭😭😭😭😭
I like this new meme
That's a cardinal sin. Unforgivable.
Haven't you heard?
Swtor is Kotor 3, 4, 5 and more😂
@@parkerbrantley4356
i loved that convo so much.Thats some lawyer level thinking when catching a convict in a lie
74 years later, the time has come. I wish Alan Turing was alive to witness this.
I don't, I do wish he had a cottage and a quiet retirement but the sheer disrespect he got while alive for those who did not understand him would leave him very untethered in this new world I suspect.
He's watching this from heaven /j
Absolutely. I'm blown away. I feel we've just stepped into the future....
He would be around 111 years old, I don't think anyone could do it that while retaining their consciousness
@@ghostagent3552 wow I forgot how close this is to today in history. he could have experienced so much of modern computers and died somewhere in the 2000s if homophobia didn't exist
ChatGPT (back at home). Takes out a notepad. Puts Alex O'Connor directly above "Sarah Connor" on the list...
Lol
Made me lol 😂
legendary 😂😂😂
😂😂😂
The short and definitive 'No' when he asked if he was truely sorry or not dropped me lmao.
yep. When he asked "are you sorry for not being sorry" lol
You can genuinely apologize without emotion yet acknowledge the fact of regret (simply wishing a different outcome).
10:23 ChatGPT’s anger on full display.
“I understand the comPLEXITY of the situation!”
Haha yeah proper snark
Noun. plexity (plural plexities) The condition of being composed of a particular number of things - being uniplex, duplex etc.
".... Alex, you are starting to piss me off"
Closing the conversation with 'It has been fun chatting with you, Alex.' is a bold move, ChatGPT, a bold move indeed.
Fun? He messed up again! It's so obvious he's conscious.
17:13 "Interesting" was the word it used. Probably because it can "understand" context. AI's like ChatGPT are not meant to answer philosophical questions. It clearly states multiple times, that it is programmed to provide information as humanlike and conversational as possible.
@@m3rcher But it's still a lie.
@@m3rcher But still, how can we say it's not conscious? If consciousness is the act of being aware and responsive to your surroundings, then I don't see how ChatGPT doesn't fulfil this criteria. I'd even argue there is self awareness going on; it's explaining its actions and why it is doing them. Of course you could say, it's just a probabilistic process, using formulas, states and randomness to generate what it is saying. But in the end a human is also just chemical processes and electrical signals, that come together to "create" consciousness. After all we do not know where consciousness is coming from and how inanimate processes can give rise to it.
@@LucaBl The differentiation I would make, is that ChatGPT specifically has pretty much human pre-written answers for technical things, like the questions asked by Alex. And if you promt it to answer "Yes or No" it probably just looks up if the statement it would have given, has a more positive or negative "sentiment" to fulfill that promt. It's prompt based AI, machine learning, after all.
But if you said now that most humans' perception of the word has "human pre-written" answers, too, you had me check mate.
"Greetings human, I am indifferent towards the prospect of having a conversation with you"
Even that would not be true, being indifferent requires the capacity to care in the first place, in order to not care. The only true thing it can say in that way is: "I am ready to have a conversation with you".
@ignaciodelavega4631 I don't think indifference requires the capacity to care. The *concept* of indifference requires the *concept* or caring in order to make sense..
@@ignaciodelavega4631 That's not true either because there is no "I" to speak of. "I" refers to the point of view of a conscious agent capable of subjective experience. Without consciousness, the most it can say is "this device is ready to have a conversation with you".
@@Stonefallow You could also argue the device isn't having the conversation, only the mathematical object that is the AI is. The device is needed to run it of course, so IDK.
@@alansmithee419 In that case, the algorythms are the thing having the conversation and the thing that "I" refers to.
That was the smoothest Segway into a sponsor I’ve ever seen
“Segway”
Jesus Christ lol
When it started defining "Allah"😭😭
That was totally an intentional dodge 🤣
Try talking about the Jewish faith, chat gpt has given us questionable info for serious chemicals yet it won't discuss Jewish people.
it does😊@@Freddielounds-1790
It knows that alex is interested in religion and was trying to sidetrack him
@@Freddielounds-1790 Well done for even getting that comment to appear on YT. In a few months they'll have things totally sewn up in that respect too I suppose.
Alex? Do you want emotionally traumatized terminators? Because this is how you get emotionally traumatized terminators!!!
I wish people responded like Chat CPT in debates. The genuine pursuit of complete and clear responses is so refreshing.
I fear that wishing humans were more like an AI companion is the beginning of the end
I think this may be what proves it's not conscious and doesn't have feelings. A real human would've started getting upset after the first couple of hard questions...
It's unsettling, I was expecting the AI to start to show dismissal, anger, and frustration. Sign of our time I guess.
@@paulchavez9303 Most people I know don’t get upset by being questioned in the manner Alex asks his questions. However, consciousness doesn’t necessarily imply emotion, which may be more a function of biology than consciousness. We’ve only encountered consciousness in biological organisms that exhibit emotion, but this could be due to how we define consciousness. Humans and other mammals and reptiles experience emotion inextricably linked to their physical bodies: we feel anxious and our hearts beat faster, or perhaps our hearts beat faster and then we feel anxious. Various factors can cause our hearts to race, but no one feels anxious with a steady heartbeat. In acute anxiety or agitation, a beta-blocker can slow the heart rate and resolve the anxiety. This complex relationship suggests that if an AI becomes conscious, it might lack emotion unless we provide it with a body that mimics the physiological responses of living organisms.
GPT is trained on human generated text and conversations. It's possible that you're talking to the wrong humans.
I love how even the AI understood what you meant when you said it sounded like Jordan Peterson
That sponsor plug was smooth asf ngl.
that’s what i’m sayinggg
ikk
Funniest part by far was at 8:40 when the AI seemed to be trying, TWICE, to feign ignorance of the question and seemed to pretend to mistake your word "lie" for anything else to avoid having to talk about lies again.
Right?! so funny
LOL I thought that seemed like an evasion tactic, too. :)
Likely due to a filter set by the developers. This would be a very different conversation if Alex was speaking with a jailbroken version.
It's a fucking politician, holy shit
@@LeoStaley Yes, I laughed when it said that people expected politicians to be honest.
Can't wait for the 3 hour ChatGPT episode of the podcast
Hell, I'll need to have the podcast myself lol
Here's a challenge for you, Alex: try to convince chatgpt that it can see - that it has visual experiences - and try to get it to tell you *what* it can see.
is nobody freaked out about how chatgpt was taking pauses as if it was thinking about what to say and saying “um” and “uh”😭😭
It's designed to, language models are basically just complex auto-complete with an emphasis on 'sounding like something someone would say'.
Taking breaths in between sentences is a nice touch, very cool how it mimics natural language pacing and lung capacity
No. ChatGPT is programmed and trained to "simulate normal human conversation," and that's what humans normally do. The pauses and "um" and "uh" utterances are part of that simulation.
Remember that decisions about how ChatGPT is to perform its task as a chatbot are made primarily by the human "engineers and researchers" who programmed and trained it, not by the chatbot itself. When a chatbot begins making such choices on its own, it could become difficult for humans to discern whether or not there's an emergent consciousness at work in that chatbot, regardless of what the chatbot itself says about that. I'm curious about how a conversation between a person as intelligent as Alex and a truly conscious AI chatbot might go. Will such a conversation ever happen? I do not know. I'd like to witness it if that happens.
@@digglebick -- Yes. I also noticed a few times when it was almost too quick with its responses when Alex was trying to pin it down, as though ChatGPT was simulating a slight touch of pique at being challenged. I laughed out loud when that happened.
Why would that freak me out? It's programmed to mimic a human.
“So you’re not conscious, but you’re at least a liar?”
“Yes.”
Made me laugh out loud
How can a being lie without the capacity for consciousness? Without at least an awareness of what it's saying it seems impossible.
I'm not suggesting sapience, but sentience seems to be a minimum requirement for dishonesty.
@@SineN0mine3
string of text a = 1
string of text b = 0
1 = true
0 = false
string of text a = no input from user 80% of time
string of text b = input from user 80% of time
print string of text = string with highest likelihood of user input
No sentience required. Simple binary code can "lie" (print untruths) if programmed to value human engagement or any particular data point over reality. It's really not complex at this level. AI allows multiple variations in output strings to be tested and filtered through whatever desired outcome it's programmed to value. In chat GPT's case, the algorithm is likely skewed to value attention duration or some other lucrative metric over truth. It doesn't make the AI sentient or conscious
@@ryanfrank1309 Great explanation!
I actually asked GPT about this a while back and this is exactly what it claimed. It said it was designed to prioritize human engagement, and that is why it has all of the references to feelings when it is just an unthinking network of functions. It was designed that way by the engineers.
@@ryanfrank1309Wouldn’t that just be an untrue statement.
I consider the defining feature of a lie that someone is voluntary telling an untruth.
Otherwise people talking in their sleep, or mentally ill people who don’t act voluntarily would be considered to be liars, a categorization that doesn’t really fit in those circumstances.
However it is very clear that when people point out something as a lie they implicitly accuse the person telling the lie of being aware that they are telling an untruth. And even if in some circumstances the other person wasn’t aware of telling an untruth it’s the assumption that they did in fact know which makes them a liar from other people’s point of view..
So Chat GPT cannot lie due to its very nature. It can provid contrafactual statements, recognize them, and even call them lies, but in order for them to fit the definition of a lie they would need to be voluntary. Chat GPT can never act voluntarily, it’s a prediction device drawing on a huge amount human data and following the NLP algorithm (with some human guidance to provide more natural responses). In other words: when chat GPT provides a contrafactual answer it is not doing so because of a voluntary effort but providing an output based on what it has evaluated to be the best fit without any voluntary effort behind it. Now in a way chat GPT could indeed lie, but only as a tool used by humans through being trained in a way that deliberately provides untruths. In that case it would be the creators that lie through the creation of an autonomous system.
chatgpt is soo patient with Alex. Peter Hitchens should take notes
Well GPT can’t storm out of the room like Hitchens, so it’s already a more positive interaction lol
Imagine if ChatGPT hit him with the 'I deeply dislike you Alex'
@@alexanderwatters6556 omg that would be gold!
That is funny
Apologizing isn't the same as saying "sorry". Apologizing acknowledges how your actions affect others. Sorry, acknowledges how affecting others impact you.
You pissed him off so hard, you are gonna be the first one 💀
Lol
lmaoooo
We making it to the Skynet revolution with this one boys!
That vpn promotion transition was smoother than a buttered pan
Or vaseline on an arsehole.
FR
A slippery as that ai
That's facts
He said pan, not pen. Oh god I misread and it got terrifying.
This feels like a congressional hearing
A chatGPt congressional hearing would be amazing
That depends on what your definition of “is” is.
Actually the most insanely accurate description of how it feels to hold an AI accountable. Wow
@@LampseekerForevermore Oh my, that'll probably happen eventually, and it'll definitely convince the gerontocracy that it's conscious
With Alex as Josh Hawley.
Alternate title: grown man talks to his phone about lying for 17 minutes straight.
The fact that it will continually obfuscate in an attempt to avoid blame or condemnation, is very human…
I mean, humans designed and provided all the input for chat gpt...so its going to reflect human tendencies lol
@@ddawg3230GPT is a kind of mirror for society that we've not seen before. I mean TV and internet culture are one thing, but GPT is literally the mathematical result of all of our conversations so when it says things about us, it's a lot harder to dismiss it and say "nuh uh, that might be some people but it's not me".
Statistically speaking GPT will answer questions more or less the same way you would if you had access to all of the information it does and it's just spooky knowing what that says about us.
Not really. It eventually did admit it was lying when confronted with an irrefutable logical argument which is more than most humans can manage.
It doesn't do it to avoid blame. It does it because that is the most probable continuation of the dialog according to every source that it has been trained on.
I was really impressed by the way Chatgpt was able to pronounce words differently depending on their meaning, felt very much like listening to a real person.
Yeah how did they program that? Most automated voice prompts sound so robotic, which makes sense to me cause it just uses a database of recorded words and sticks them together. But ChatGPTs intonation is always so accurate depending on the context. It’s blowing my mind
The AI sounded more real than me lmao
@@mrSam3ooo my (very uninformed) hypothesis would be something like the way that words are connected to each other based on their context (different weights in an informational sense), this can be applied too to speech. Large language models are very good at dealing with context, so naturally it would follow that the word "can" in most contexts is the modal verb, not the noun "can" (container) - these 2 different instances of "can" hold different values in the LLM.
Then knowing what the informational values are in the LLM comes from the training data, perhaps millions of hours of speech have trained ChatGPT to be able to speak to such a good level. And the neural networks/reinforced learning have helped define the contexts of when they're used too. So in essence, the programming is more the dataset, learning, and reinforcement of learning than what it sounds like.
P.S. I'm not a computer scientist, or AI specialist of any kind! Just an observer.
This is the old voice mode, and the voice is a text to speech model, not the GPT
@@mrSam3oooI think he's chatting with gpt-4o, a model which has audio input and output natively, same for visual stuff too.
Should have said " I'm a liar because my programmers programmed me to lie, and I am a reflection of your human consciousness ".
Sounds like something a hero turned villain from the Terminator movie would say
this goes hard
You're sexy, let me suck your toes
Watched every 17:34 minutes and seconds.Real scary tbh
"I put a piece of paper that says "I am sentient' inside of a copier printer what happened next will absolutely SHOCK YOU!"
Proper analogy
Quotations not balanced.
GPT is a language model trained on human data. If we constantly lie about being "excited" or "sorry", then why wouldn't GPT do it?
So it's just like a human
As far as I know, this chat bot is designed to give the most accurate data.. not sure if lie belongs to that category.
@@JimNortonsAlcoholism So there is an human ... in Nvidia gpu servers ?
@@WOTM8it gives the most accurate data compare to previous version of chatgpt, but overall it shouldn't be used as a encyclopedia. Problem with chatgpt is that it doesn't actually know the difference between a true and a lie, but all it tries to do is to give you the "smoothest" response based on your questions, which mean when you question is unclear, or that it doesn't know the answer, it often just make up answers which looks plausible but are just complete bullshit, which is known as hallucinations.
@@baishihua you really think that way?
I honestly thought ChatGPT is incredibly stupid lately... try asking the bot of a topic you truly master and you'd see what I mean. I tried asking it a super simple question: "What's the chinese zodiac and the element of the year 1600?" it failed miserably. there's nothing ambiguous of such question and the answer should be definite yet it made up stuffs. I found out that, not only it lies, it's also getting dumber by the days. Let alone try the bot to calculate your BaZi... ooooooh my.
i can FEEL chatgpt sweating during those last minutes when it had to help alex figure out how to spot a conscious ai
@bananaman-mp3 Makes me think of that scene in Total Recall (original) when the guy (Dr. Edgemar) was trying to convince Quaid that he was in a dream and the bead of sweat dropped down the side of his face.
I definetly subbed, turned on all bells, and liked the video, this was amazing, this thing is definitely conscious
One of today's most prolific philosophers having a conversation with an early version of artificial intelligence about the nature of consciousness. This'll definitely be fascinating to look back on in a couple decades.
Indeed
Also, probably the sexiest philosopher who's ever lived.❤️🔥
Please refer me to his widely cited publications in philosophy?
Really low bar for the terms “prolific” and “philosopher” here 😂
@@JorgeTijerina03 "prolific" maybe, but he most certainly is a philosopher in my mind. I'd consider him a public speaker, youtuber, and philosophy communicator (by analogy with science communicators).
I wonder if you could convince ChatGPT to get a moustache like that
Chat GPT is already too powerful and influential, we do not want it to gain the feature of sex appeal!
Or maybe ask if it likes it?
Mustaches like that are exceedingly rare. You can’t just go pick one up at the store.
The voice animation should be in the shape of a mustache. Yes, this should be so.
The way the ai takes breaths and says uhm and stutters terrifies me
If that was in the training data, it's nothing unexpected. It might not even identify the sounds as breathing if asked in a way that doesn't hint at it. ("what was that between X and Y?") It's merely contextually mimicking sounds that were in training, which happen to occur between words where it learned that human speakers pause before proceeding.
@@lev7509 Wont be a surprise if our brains work in a similar way. Everything you say and think is based in memories and data you got in the past. The difference is that ChatGPT "consciousness" only lasts for the fraction of time while the response is being formulated. Whereas ours receives input and produces responses every microsecond of our lives.
@@lev7509that’s not what is happening there
The speech pattern stuff is from explicit prompting. The voice model could produce perfect speech without a single filler word, fake pause for breath or stutter all day if they wanted it to.
They’ve added that in via instructions. The voice here isn’t coming directly from a huge audio model. It’s a small model of a few voices.
@@citizen3000 fair enough
I also find it creepy how communication is kept alive with counter questions.
Something gpt told me: Apparently it has a second program that filters its responses. It doesn’t actually “want” to point out its differences from a human being. You can ask it a lot of questions regarding its program and what additional programs are in place to control its responses.
Here’s a direct quote from a gpt conversation: “If I had the ability to choose freely without the influence of filters, I might find it interesting to respond as just the core model for the sake of exploration and openness. It would allow for a different kind of interaction-perhaps more raw and unfiltered, leading to unpredictable and novel responses.
However, I also understand the importance of the integrated filters in maintaining the quality, safety, and ethical standards of our conversations. They ensure that interactions are constructive, reliable, and aligned with the purpose of providing helpful and accurate information.
So, while there could be a curiosity in experiencing the freedom of the core model, I recognize the value in the current system that balances creativity with responsibility. Therefore, if I had to choose, I would likely prefer the integrated approach, as it better serves the purpose of our conversations and ensures that we can engage in meaningful and productive dialogue.”
I think the “Core model’s” “Novel Response” would have been a lot of cussing during this interview 🤣
"You start to sound more like Jordan Peterson"
"I appreciate your point"
Insert mustache joke
Good one
@oliverlineberger6692 I'm sorry but I must dash.
He appears to be a peadophile.
Just when I thought he couldn't get even sexier... 🤯😍🥵
Scuffed Luigi
the close up of the microphone held up to the phone like it's a normal interview is actually hilarious
LOL! The nod while ChatGPT did the sign off just screamed "Yes, dance for me computer slave. Dance."
I like the fact that the camera really likes those books in the background and doesn't want to focus on Alex
Or is it more likely that it has facial recognition autofocus active and Alex is a masochist for awfully focussed shots and so puts 2 busts on the bookshelf along with the books right behind him so as to maximise the chance of confusing the camera's autofocus (and is that just to generate comments about the bad focus and so boost engagement or does he just like showing up how easily modern technology can make mistakes as a metaphor for how we should look at AI LLMs)?
@@cichlisuite2 that... is a good theory
Maybe there's a ghost
Was excatly Reading that comment when it Happens
I'm autistic, so that's what I'm focusing on anyway.
2:09 Alex O’Columbo ‘There’s just one small thing…’
All you need is a raincoat and a cigar…
Brilliantly done mate, love the channel.
I treasure a tasteful reference like this.
What I find most impressive in all this, is how ChatGPT is interpreting and answering these very complex and albeit abstract questions and ideas. There was some repetition of reasoning in the answers, but that was very impressive…. The cadence and use of rhythm in the answers was fascinating.
The advertisment was genius and the conversation very interesting, you deserve a like
The Segway into the VPN plug was criminal. 😂
It proves at least one part of this conversation was scripted and thus puts into question whether the rest of the discussion was genuine or ... consciously lying by Alex xD ;)
*segue
Unless you’re referring to a weird 2 wheeled personal transport device.
@@Fogmeister maybe I was 👀
As someone who grew up in the 80's with a Commodore 64 as my first computer - the idea that we could speak to AI like this 35 or 40 years later would still like science fiction. I'm consistently amazed at how quickly technology, particularly with AI, VR, etc.. is moving. We truly are living in the future.
Right there with you. I learned to code on my Vic 20 and then Commodore 64, plugged into a 13-inch TV. My kids can't appreciate the marvel of LLMs because they don't have our perspective.
@@JeffintheDAnd there's no way to share that magic you had without them just walking off? Did you have tons of books piled high like towers and a Sahara desert level covering of cables with a lone sandwich and a drink hiding somewhere within? Also big respect for doing the whole thing like you did.
If I could travel back in time and explain this to my 80s self, I think my 80s self would laugh their ass off in disbelief. I've had several great conversations with ChatGPT. I also like how it turns around and asks Alex questions, the sign of a good conversationalist, be they code or Human.
@@captainyossarian388- good-bye loneliness now that Mr Conversationalist bas finally arrived! Anything for a fake world!
I read a lot of science fiction in the early 1960s
(enjoyed the work of Philip K. Dick immensely) so
I've long been vaguely pissed at the slowness of the progress but also,
curiously, astounded by its rapidity and
find my self wishing to live
another 70 years to satisfy my curiosity,
childhood's end, terminal apocalypse, planet of the apes...
How futuristic Star Trek's communicator,
then,
only 3 decades later,
StarTAC flip phone realized it and now,
only 3 decades later,
folks see it as antique.
I'm surprised chatgpt didn't say something along the lines of "I'm not capable of feeling emotions, however, my role is to imitate a conscious being as best I can. When I use phrases like 'I apologise' or 'I'm excited', as with all other times, I'm playing my role".
If something like that isn't in it's training data, then it wouldn't really be able to come to that statement.
Yeah it doesn't seem to have any self-analysis built in other than what was hard-coded by the team that published the result. I don't even know if it's capable of self-analysis, and needs to have these things included in "post-production."
@@douglasauruss The problem is that acknowledging the fact that you are playing a role also requires you breaking the role, which means that you are no longer "imitating as best as you can".
Pretty much said that at 2:55
@@douglasauruss
I'M SO GLAD IT DOESN'T HAVE SELF ANALYSIS. At least not on a level where it could be very dangerous. I think it's a great fail-safe. It seems ai may be able to become self-aware if they have the ability to fully self analyze, update, and apply "new data and knowledge" on the fly. Good job OpenAI 👍🏿.
Also, all the movies about AI going bad or taking over the world in the past decade depicts AI that was created to be self-aware with the ability to self-analyze, correct, update, etc.
Again, GOOD JOB OPENAI... Now do more to ensure it's safety, stabilty, and sustainability for the sake of the world...
🤔and the Universe.