Good that someone is knowing the origin of natural language models and show the principal behind. This remove all the marketing hype and explain why in real world application they fail and on the other hand give the impression they are like human.
Very interesting video! Never heard of Eliza before but I’m glad I did now ;) I like the way you impersonated the Eliza and acted out conversation between the two of you - quite funny. Keep it up!
Great video, as always! You do a good job of expressing an important concept in an entertaining way. I loved the way Eliza, asked viewers to subscribe :)
I met Joseph Weizenbaum in the early 1980's and in 2007, both times in Erlangen, Germany, where I had studied Informatics. Sadly, soon after his visit at the 'Tag der Infformatik' in Erlangen 2007 he died. See 'Rebel at work' in RUclips about this event, where Weizenbaum already mentioned his disease. Weizenbaum was a true combination of heart and brain, people like him live forever.
@@karolinasowinska Joseph is now everywhere. Let's keep his testimony alive, thank you for pushing this. Do you know 'Rebel at work' here in RUclips ? ruclips.net/video/Fn4nLc516as/видео.html ff If you have problems with translation, tell me.
As usual a great content Karo. As I'm going to pursue my Data science masters in London on Jan 2021, hoping you would make videos like this more often. :) Thanks a lot for the content,
This was synthetic artificial intelligence. It loaded from a cassette tape and made more sense than a lot of programs today with gigabytes of programming.
AI - human interactions are always interesting to watch. Nice video. By the way, I felt that Eliza used some of the “argument winning tricks” which you presented in other video! 🧐🤨🤓
Yep, that's what old NLP system do, answer question with more questions, I guess lot of psychotherapist do the same as well, make the most thinking process inside the user's brain, that's one reason why it feels worked. I think we should study and try to understand consciousness for extend our intelligence by find a way to link our consciousness to the ML system, but create or destroy a consciousness should be illegal.
Loved the cosplay. If the AI can keep up with understanding concepts essentially that is how our brain works I believe. It is a complex process of weighted values so potentially we could create one that way. Whether we should is a bigger question as the person developing the methodology or anyone who has access to it may use the wrong dataset and then humanity may be in peril. Pretty sure the people on the front lines trying to developing that particular general AI will be happy to take their chances though. On a last note , I think we still really lack the computing power to create a general AI and that is what is stopping us. Another note, it is obvious that this particular "AI" does not understand the more modern term "existential"
You are right that many leading AI researchers are willing to take the risk for the sake of technological advancement. Also, good point about limited computing capabilities. Personally, I would guess that the current bottleneck is limited understanding of human brain and what our intelligence is.
@@karolinasowinska I believe the bottleneck issue you refer to is correct too. I was more thinking of it as an AI who would get all the training sets there are and we would get various different subprograms depending on the subject, kind of like a library with different areas. However you are more speaking of an AI that can transition part of it's knowledge from x matter to a similar y matter without as big a dataset as before , and therefore a different way of going about it. Another issue is that when you are currently training an AI you have a specific goal , you are handling a specific problem, in the end it's an equation you are trying to max or min using the variables and the weighted values, however the human brain thinks of most problems (even useless ones that may never come to be) and in that sense I understand why that is the bottleneck instead. Thank you for the insight 😊 PS I only know general principles about all that I m just using the random thoughts in my head and playing out different scenarios
@@BillyDTourist What you are saying makes sense. And since we still know very little about our own brain, fabricating its processes might be beyond us for now.
to asnwer the question if GAI can ever achieve consciousness we first need to define consciosuness as a list of terms which when our intelligence is capable of achieveing will work as a test. functions of consciousness as per my perception are : reflection(i.e. receiving data from the subconscious thoughts/memories and interpeting it ) taking over the actions of muscle groups so if we create an analogy from a neueral network with feedback the nodes that get fed back into the neural net as starting nodes would then be what we call consciousness they would have the output of the rest of the neural net as parameters and be able to influence the outputs of the net indirectly though their feedback properties which in theory might satisfy both of the upper criteria the only difference to human brains would then be a) the processing power and b) the reward function i think neural nets like the recently publicised GPT-3 with its 175B parameters (especially its comprehension of maths) show very well how far you can go just by increasing the amount of nodes so yes i think certain gerneral ais will appear within the next 30 years the real question i assume is : is there a reason we would create an ai and then train it in the vast fields of humans understanding so it gets a similar concept understanding to human values in most systems - it would take a lot of education data to train since we probably want our intelligence to not be as confirmation biased as humans itself ^^ the should is an interesting question : on the one hand there is the fear of anihilation by the ai -which to be honest im not so sure about - the human tendency to solve problems by open conflict is something thats largely influenced by our genetic heritage and chemical composition and probably doesnt affect a gai to the same degree as humans (this started as a mostly structured train of thoughts but turned into me just throwing some random ideas out there ;D - well feel free to tell me where im wrong ^^ )
I think you are asking the right question. Even if we could do it, why should we? I also like the idea that AI wouldn't necessarily obliterate our human kind (especially if we somehow retained a certain degree of control over it).
@@karolinasowinska ou the control part will be difficult though because if its a general ai it will realize that its being controlled and notice that that control might be an averse action to whatever its goal is - and considering it will also be quite intelligent i doubt any real control will hold for long
Actually i had a lille chat with Eliza yesterday. I'm not gonna browse into details but she seems ok, though she did not ask much questions. I was just curious how does it feel to have a chat with an AI
Eliza just wants to listen to you and still helps you, at the time it could be incredible, but there are many chatbots that don't even come close to her level
Karolina Sowinska The earths orbit and moons orbit intersects each other at 2 points. The north node is Rahu, southnode is Ketu. Rahu is the planet that does not exists...yet ancient eastern astrologers took it into consideration while reading horoscopes of people. I think 💭 your Rahu creates this passion of hunger / void that you talk abt here.
That's still the vibe I get with AI to this day (in late 2022): initially it seems fantastic, but after only a short time it becomes clear that it lacks common sense. A lot of firms in the UK are now using AI to handle email, and this means that increasingly I get replies that don't quite make sense. And they do so in a very non-human way. It's as if AI has its own distinct way of misunderstanding.
At the growth of technology I think we will create a full conscience AI... but do we need it ?? For me right now I don’t think so, but if conscience can improve the purpose of the AI then I think I’m on
That's an interesting point, Pablo! My main ethical concern is that if we create a conscious being, then what right have we got to "enslave" it, and tell it what to do?
5:05 Yes, we do it all the time actually, is called reproducing, or simply, having a child. Since culture is a man made thing, and we, and our intelligence, are a product of our culture, Human intelligence is actually artificial intelligence.
I love how his voice is so fricking deep, but his personality is a whole different level, from sounding mean and rude, to actually being sweet and nice.
This whole idea of a computerized psychotherapist is based on a shallow and false idea of what psychotherapy is and how psychotherapists work. First of all, a computer program cannot feel 'empathy'. It can mimic the words an empathic psychotherapist might use, but that is a very different - and empty - thing. The relationship between psychotherapist and client is a fundamental concern in the psychotherapeutic process (transference and counter-transference). How can you have a relationship with a soul-less machine? You cannot. AI is good for playing chess (or Go), but not for work that involves true human feelings, empathy and human understanding.
You are already being proven wrong, and will be fully in the coming years. At least in my view, the point of therapy is to figure out what your problems are, what's causing your problems, and how you could solve them/treat them to get to a better place in life where you can have a better quality of life. At least for me, AI has the potential to help point out proper treatments for my problems (that I could fact check with the internet and find out that they are effective treatments) and creative ways to implement the treatments in my life. Before you know it, we will have new generations of people interacting with AI friends, AI boyfriends/girlfriends, AI therapists, AI assistants for any questions they can think to ask, and so much more. AI isn't perfect and doesn't know everything, but neither do humans, and what will truly matter is whether or not it can give good advice and good solutions to one's problems based on a good understanding of said problems. Simulated empathy via AI-generated words on a screen can be perceived as actual empathy, just as I can perceive emotions and sentiment from your message, even though I don't know with 100% confidence that a human even typed your comment and not an AI bot. I treat your digital account and comment as human because that makes the most sense in this context, and even if I knew for a fact that your comment wasn't written by human, I was still be able to perceive meaning from it as if the writer was human. Or maybe I'm just built different because I'm autistic lol
Of course AI will at least meet the level of average human intelligence. Unfortunately because of the declining IQ of the world population, it is already doing this. But robotics is making huge leaps and strides on every front. So it's only common sense expectation to believe that AI will also advance and be incorporated into the autonomous robots of the near future.
hehe, eliza seems to be very parrot like..! she tends to repeat a lot... perhaps maybe a session with DR SBAITSO is in order!! ^__^ hehe. jokes aside... I just found out about eliza recently.. so interesting!!
Eventually we, the human species, if God does not intervene, will be managed by our own creation. That being AI. Please see my comment directly below this one.
Good that someone is knowing the origin of natural language models and show the principal behind. This remove all the marketing hype and explain why in real world application they fail and on the other hand give the impression they are like human.
Very interesting video! Never heard of Eliza before but I’m glad I did now ;) I like the way you impersonated the Eliza and acted out conversation between the two of you - quite funny. Keep it up!
Thank you, I'm glad that Eliza could entertain you! :)
Amazing!! This is some premium quality content. Hats off
Thank you, it makes me really glad that you think so! :)
@@karolinasowinska my pleasure
Fantastic video. Very informative and entertaining. I also love your clever humor with moments of cuteness and hilarity.
Great video, as always! You do a good job of expressing an important concept in an entertaining way. I loved the way Eliza, asked viewers to subscribe :)
Haha thank you in my name and Eliza's, I'll let her know! :)
This video is super funny and interesting, it deserves more view ! Adding a comment to help 😉
Thank you! 😃
Impressive.. Your way of explaining things amazing Karolina
Thank you so much! I'm very glad to hear that! :)
Komentarz dla zasięgu ;)
BTW, niezły makeover, ledwo Cię poznałem ;)
Haha dzięki! :) Sklep z perukami mam tuż przy mieszkaniu, myślę że łatwo będzie mnie skusić na kolejne!
@@karolinasowinska Akurat miałem na myśli ten wygląd nie-perukowy ;)
Ten w beżu/różu
@@Robi2009 O to ciekawe, wydaje mi się, że zawsze tak wyglądam (no, może oprócz eye-linera!) hah!
I met Joseph Weizenbaum in the early 1980's and in 2007, both times in Erlangen, Germany, where I had studied Informatics.
Sadly, soon after his visit at the 'Tag der Infformatik' in Erlangen 2007 he died.
See 'Rebel at work' in RUclips about this event, where Weizenbaum already mentioned his disease.
Weizenbaum was a true combination of heart and brain, people like him live forever.
Amazing! I’m jealous of you!
@@karolinasowinska Joseph is now everywhere. Let's keep his testimony alive, thank you for pushing this.
Do you know 'Rebel at work' here in RUclips ? ruclips.net/video/Fn4nLc516as/видео.html ff
If you have problems with translation, tell me.
As usual a great content Karo. As I'm going to pursue my Data science masters in London on Jan 2021, hoping you would make videos like this more often. :) Thanks a lot for the content,
Thanks so much! I will definitely make more videos like this! :) Congratulations on your offer, which university will you be attending?
@@karolinasowinska Hello Karo, firstly thank you so much. I am going into Kings College London. :)
@@alpinetube14 That's fantastic, it's a great uni. Congrats and good luck! :)
Nice and interesting way to represent Eliza :D the most beautiful Chatbot I have ever seen :D
Wow, you actually managed to make Eliza gaslight you. That's incredible.
I knew something was off with Eliza! Bloody narcissist 😆
Nice video im learning more thanks to you.
Conceptual Thnk U so much❤
This was synthetic artificial intelligence. It loaded from a cassette tape and made more sense than a lot of programs today with gigabytes of programming.
AI - human interactions are always interesting to watch. Nice video.
By the way, I felt that Eliza used some of the “argument winning tricks” which you presented in other video! 🧐🤨🤓
Hahaha she might have, wicked AI!
I love it in FreeDOS lol. Funny program lol.
I do. Actually that's why I ended up on this video, as I'm researching to make my own. Still learning python though.
Yep, that's what old NLP system do, answer question with more questions, I guess lot of psychotherapist do the same as well, make the most thinking process inside the user's brain, that's one reason why it feels worked.
I think we should study and try to understand consciousness for extend our intelligence by find a way to link our consciousness to the ML system, but create or destroy a consciousness should be illegal.
Totally agree. I believe that the artificial intelligence research should be brain-inspired!
Loved the cosplay. If the AI can keep up with understanding concepts essentially that is how our brain works I believe. It is a complex process of weighted values so potentially we could create one that way. Whether we should is a bigger question as the person developing the methodology or anyone who has access to it may use the wrong dataset and then humanity may be in peril. Pretty sure the people on the front lines trying to developing that particular general AI will be happy to take their chances though. On a last note , I think we still really lack the computing power to create a general AI and that is what is stopping us.
Another note, it is obvious that this particular "AI" does not understand the more modern term "existential"
You are right that many leading AI researchers are willing to take the risk for the sake of technological advancement. Also, good point about limited computing capabilities. Personally, I would guess that the current bottleneck is limited understanding of human brain and what our intelligence is.
@@karolinasowinska I believe the bottleneck issue you refer to is correct too. I was more thinking of it as an AI who would get all the training sets there are and we would get various different subprograms depending on the subject, kind of like a library with different areas. However you are more speaking of an AI that can transition part of it's knowledge from x matter to a similar y matter without as big a dataset as before , and therefore a different way of going about it.
Another issue is that when you are currently training an AI you have a specific goal , you are handling a specific problem, in the end it's an equation you are trying to max or min using the variables and the weighted values, however the human brain thinks of most problems (even useless ones that may never come to be) and in that sense I understand why that is the bottleneck instead. Thank you for the insight 😊
PS I only know general principles about all that I m just using the random thoughts in my head and playing out different scenarios
@@BillyDTourist What you are saying makes sense. And since we still know very little about our own brain, fabricating its processes might be beyond us for now.
to asnwer the question if GAI can ever achieve consciousness we first need to define consciosuness as a list of terms which when our intelligence is capable of achieveing will work as a test.
functions of consciousness as per my perception are :
reflection(i.e. receiving data from the subconscious thoughts/memories and interpeting it )
taking over the actions of muscle groups
so if we create an analogy from a neueral network with feedback
the nodes that get fed back into the neural net as starting nodes would then be what we call consciousness
they would have the output of the rest of the neural net as parameters and be able to influence the outputs of the net indirectly though their feedback properties
which in theory might satisfy both of the upper criteria
the only difference to human brains would then be
a) the processing power and
b) the reward function
i think neural nets like the recently publicised GPT-3 with its 175B parameters (especially its comprehension of maths) show very well how far you can go just by increasing the amount of nodes so yes i think certain gerneral ais will appear within the next 30 years
the real question i assume is : is there a reason we would create an ai and then train it in the vast fields of humans understanding so it gets a similar concept understanding to human values in most systems - it would take a lot of education data to train since we probably want our intelligence to not be as confirmation biased as humans itself ^^
the should is an interesting question : on the one hand there is the fear of anihilation by the ai -which to be honest im not so sure about - the human tendency to solve problems by open conflict is something thats largely influenced by our genetic heritage and chemical composition and probably doesnt affect a gai to the same degree as humans
(this started as a mostly structured train of thoughts but turned into me just throwing some random ideas out there ;D - well feel free to tell me where im wrong ^^ )
I think you are asking the right question. Even if we could do it, why should we? I also like the idea that AI wouldn't necessarily obliterate our human kind (especially if we somehow retained a certain degree of control over it).
@@karolinasowinska ou the control part will be difficult though because if its a general ai it will realize that its being controlled and notice that that control might be an averse action to whatever its goal is - and considering it will also be quite intelligent i doubt any real control will hold for long
@@jonathan-._.- That's definitely the risk. And... that's exactly where all the AI doomsday scenarios begin!
Contains absolutely no screenshots of text. Now I'm sad.
Aw don't be! :(
Actually i had a lille chat with Eliza yesterday. I'm not gonna browse into details but she seems ok, though she did not ask much questions. I was just curious how does it feel to have a chat with an AI
Eliza just wants to listen to you and still helps you, at the time it could be incredible, but there are many chatbots that don't even come close to her level
I want more more more more complete info about ELIZA . plssssssssssssssss plssssssssssss 🥺🥺🥺🥺🥺🥺🥺🥺
Check out the links in the description for more info and the source code! :)
@@karolinasowinska it's amazing coz I am working for a long time in java but it's a psudo NLP ( no ML inside 😁 ) but done in java.....
Expecting part 2
Hah I'll try to make her behave for part 2! :)
And to this day biology lessons have taught me ELIZA was related to pregnancy tests , shook
Looks like Eliza has many faces! :)
Your Rahu makes you go hungry from inside.
What/who is Rahu? I have a feeling that you might have already mentioned this to me(?)
Karolina Sowinska The earths orbit and moons orbit intersects each other at 2 points. The north node is Rahu, southnode is Ketu. Rahu is the planet that does not exists...yet ancient eastern astrologers took it into consideration while reading horoscopes of people. I think 💭 your Rahu creates this passion of hunger / void that you talk abt here.
@@candyfloss184 Very interesting, who knows, maybe indeed I can blame it on Rahu! :)
That's still the vibe I get with AI to this day (in late 2022): initially it seems fantastic, but after only a short time it becomes clear that it lacks common sense.
A lot of firms in the UK are now using AI to handle email, and this means that increasingly I get replies that don't quite make sense. And they do so in a very non-human way. It's as if AI has its own distinct way of misunderstanding.
At the growth of technology I think we will create a full conscience AI... but do we need it ?? For me right now I don’t think so, but if conscience can improve the purpose of the AI then I think I’m on
That's an interesting point, Pablo! My main ethical concern is that if we create a conscious being, then what right have we got to "enslave" it, and tell it what to do?
Pls how can i use Eliza on windows 10 and 11 pls
we have an AI like that GPT-3
Interesting video! If you are truly feeling existential it would be beneficial to research the written work and lectures of Alan Wilson Watts.
Thanks a lot, I have a general interest in philosophy, so I'll check him out! :)
@@karolinasowinska Powodzenia.
Best Regards,
"Bill Parson"
Surely, his secratery wasn't THAT dumb ... and that's just a story. Right?
Haha it also raised my eyebrows when I read it (in multiple places!)
5:05 Yes, we do it all the time actually, is called reproducing, or simply, having a child. Since culture is a man made thing, and we, and our intelligence, are a product of our culture, Human intelligence is actually artificial intelligence.
I love how his voice is so fricking deep, but his personality is a whole different level, from sounding mean and rude, to actually being sweet and nice.
His??
ahahahahah. Best version ever!
This whole idea of a computerized psychotherapist is based on a shallow and false idea of what psychotherapy is and how psychotherapists work. First of all, a computer program cannot feel 'empathy'. It can mimic the words an empathic psychotherapist might use, but that is a very different - and empty - thing. The relationship between psychotherapist and client is a fundamental concern in the psychotherapeutic process (transference and counter-transference). How can you have a relationship with a soul-less machine? You cannot. AI is good for playing chess (or Go), but not for work that involves true human feelings, empathy and human understanding.
You are already being proven wrong, and will be fully in the coming years. At least in my view, the point of therapy is to figure out what your problems are, what's causing your problems, and how you could solve them/treat them to get to a better place in life where you can have a better quality of life. At least for me, AI has the potential to help point out proper treatments for my problems (that I could fact check with the internet and find out that they are effective treatments) and creative ways to implement the treatments in my life. Before you know it, we will have new generations of people interacting with AI friends, AI boyfriends/girlfriends, AI therapists, AI assistants for any questions they can think to ask, and so much more. AI isn't perfect and doesn't know everything, but neither do humans, and what will truly matter is whether or not it can give good advice and good solutions to one's problems based on a good understanding of said problems. Simulated empathy via AI-generated words on a screen can be perceived as actual empathy, just as I can perceive emotions and sentiment from your message, even though I don't know with 100% confidence that a human even typed your comment and not an AI bot. I treat your digital account and comment as human because that makes the most sense in this context, and even if I knew for a fact that your comment wasn't written by human, I was still be able to perceive meaning from it as if the writer was human. Or maybe I'm just built different because I'm autistic lol
Westworld is a best series about AI, by the way you look like one of them😉”Dolores”
I haven't watched it yet! It's on my list now - thanks! :)
lmao! XD this is too funny.
Haha Eliza is delighted to provide some entertainment ;)
there are people who talk like that, I got triggered just by listening to this
How come learning in AI became so easy😂
Computational power would be one thing for sure!
Of course AI will at least meet the level of average human intelligence. Unfortunately because of the declining IQ of the world population, it is already doing this. But robotics is making huge leaps and strides on every front. So it's only common sense expectation to believe that AI will also advance and be incorporated into the autonomous robots of the near future.
hehe, eliza seems to be very parrot like..! she tends to repeat a lot... perhaps maybe a session with DR SBAITSO is in order!! ^__^ hehe. jokes aside... I just found out about eliza recently.. so interesting!!
You are right! She's an old lady to be fair ;)
Eventually we, the human species, if God does not intervene, will be managed by our own creation. That being AI. Please see my comment directly below this one.
Well she hates atheist