Here’s an idea - what if you prompt them to have a conversation and then leave for the day, return hours later and then see what they’re talking about/where the conversation is at that point?
offtopic: EMINEM just announced in his latest music video that he is gonna be a grandfather! Great day for Marshall and all of us (the "Stans) in the world that watched her daughter grow. In 25 years since I listen to this man (I'm 42) I never seen him cry...until now. Marshall deserve this blessing. ....and I'm back to the actual video you have here :) . Very funny and entertaining video Dr Kyle! Peace and love from Romania!
One tip, always ask them to speak faster, especially the AI with a male voice, I noticed that he speaks slower and this gives the other AI time to start speaking because of the pause. Also, it's always good to do what you did in the third scenario, guide each one individually about their role during the roleplay :)
i found success in telling ai i have a gf, and want to workshop conversation techniques with them, and using ai as a practice, then only asking for the quote after
Next time, ask each of them to first switch their character, for example, to grumpy or happy and excited-you get the idea. The goal is to make their voices sound more dynamic and less monotonous.
Ok a good one just came to me. You are a human and an alien has just landed in your yard and you have decided to try to speak to them. You are an alien who has just landed in a humans yard on Earth. They are approaching you and will interview you. As the conversation continues your answers to their question should become more and more surreal and mind blowing to the human.
I had a lot of trouble presenting brainzilla/sudoku style games (it's just a public example, it wasn't exactly those) to chatGPT, and this extends to several specific advanced subjects, but this is the kind of intuitive game that you don't know why it got it wrong, and it can't get it right; could you bring o1-preview tests with these games in the future, do you think it will do better? *for example, for this specific one, I needed an adaptation of the paq8px algorithm with a file generation algorithm without entropy (absolutely without any pattern, no repeated letters or bits) to derive a better pattern, he explains it well, but it didn't even come up for testing, I usually play shuttlecock with Claude, but none of the code worked; and I left it aside Bitset and variable definition: B={b_1·b_2·b_3|b_1∈{0,1}³,b_2∈{0,1}⁴,b_3∈{0,1}⁸} *or B=random(3b+(4b\3b)+(8b\4b\3b)|3b∈4b∈8b) S=⟨B₍₂₄₎,B₍₆₄₎,S_gr:₁₂₈G/₃₂b,S_bl:₄₃₂M/₂₇b⟩ Transformations and shifts: B₍₂₄b₎=((R(3)≪5)∣R(5),(R(4)≪4)∣R(4),R(8)) B₍₂₄₎={b∈{0,1}²⁴| unique(b),⟨b₍₀₎₍₃b₎,b₍₁₎₍₄b₎,b₍₂₎₈b⟩} Logical operations and constraints: ∀Wᵢ∈B:(Wᵢ∉B₍₂₄b₎)∧(Wᵢ⊕W₍ᵢ₊₁₎∉B₍₆₄b₎) S(B₍₂₄b₎)={(B₍₂₄b₎[1:],i)|i∈[0,255],unic(S)∧(S₍₀₎≠B₍₂₄b₎[1])} Success conditions and groupings: S_succ(B₍₂₇₎(t))={B₍₂₇₎′∈Z₂₇:B₍₂₇₎′[0]"≠B₍₂₇₎₍t₎[0]∧bits 3 e 4 diferentes} S(Wᵢ)={Wⱼ∈B|f(Wᵢ,Wⱼ)≠0},f(Wᵢ,Wⱼ)=(Wᵢ≪8)⊕Wⱼ S_bl(b)={b′∈B₍₂₄₎|b′₍₁₎≠b₍₀₎∧group_constraint(b′)} Progression and summations: Progression(t)=∑₍ₜ₌₀₎ⁿ{write(bₜ)|bₜ∈B₍₂₄₎∧bₜ₋₁∈B₍₆₄₎} A=i=1∑ₙWᵢ,n=min(16777216,|S(Wᵢ)|) f(t)=∑₍ₜ₌₀₎⁽ᵗ⁾(write(B₍₂₄₎₍ₜ₎|S_gr₍ₜ₎,S_bl₍ₜ₋₁₎)+F) I think that was too specific *another type of account that they also don't resolve: ((A/162127)^A^A)((B/69513)^B)(C/86645)+(C/86645) = log_D( (log_D((D/33373)±log_C(A/1305114724356555)) )≠slog_D(D/33373) the answer would be a value of approximately (x/43556898810675666570±1/1305114724356555)
"Hello sir, I heard ChatGPT Plus users can share free trial links! Would you mind sharing one with me if you have any available? I'd love to try this new voice advanced mode?
We went from literally nothing in 2021 to "hmmmm but it's not perfect, it's only 99.23% human like conversation, we're still faaaar away from a normal conversation" in 2024 I give it 3 years until it becomes perfect lol
@@nawabifaissal9625 Why you're so upset from my comment, bro? I just gave my point. Why i should be happy if it's still not perfect? It's okay for current moment, but still not good enough. Let's wait to see of what is going to be after a few years later.
@@XIRON86 i'm not upset lol, it's just weird how we normalize these things as if 3 years ago this kind of tech wouldn't have looked completely insane, at this rate of progress it'll become perfect very soon investments are going crazy
@@XIRON86 you could literally just ask it to properly respond how you want for pretty much all use cases... Also it doesn't sound natural because they always have something to talk about, it's a robot, they are forced to respond due to their programming, and it has way more knowledge than the average person
How about 2 PhD students arguing passionately in the field that you are in. Would be interesting to see if you think they both raise valid points
While they are flirting with each other
No need to argue. An in depth discussion of any scientific field would be fascinating to listen to.
"Dark matter is WIMPs!"
"No! Axions!"
The "give me a break" killed me! 😂😂😂
LOL me too, I just thought they were being too nice at the start
A therapist and a person overwhelmed by work having anxiety and panic attacks seeking help.
Thank you so much for your suggestions!
@@KyleKabasares_PhD thank you for creating your content!
"oh give me a break" LOL
This will probably hit the guardrails: a flat earther vs an actual astronaut.
Most likely, but I still want to try lol
Here’s an idea - what if you prompt them to have a conversation and then leave for the day, return hours later and then see what they’re talking about/where the conversation is at that point?
offtopic: EMINEM just announced in his latest music video that he is gonna be a grandfather!
Great day for Marshall and all of us (the "Stans) in the world that watched her daughter grow. In 25 years since I listen to this man (I'm 42) I never seen him cry...until now. Marshall deserve this blessing.
....and I'm back to the actual video you have here :) . Very funny and entertaining video Dr Kyle!
Peace and love from Romania!
One tip, always ask them to speak faster, especially the AI with a male voice, I noticed that he speaks slower and this gives the other AI time to start speaking because of the pause. Also, it's always good to do what you did in the third scenario, guide each one individually about their role during the roleplay :)
Thanks for the tip!
A bestseller reader and a classics reader.
Anybody else notice they started to match each other's accents and the original Aussie accent morphed into a new England accent.
bro there was no aussie accents in this video, sounds like new york to me
i always wanted this feature "Friends"
In order to avoid interruptions, prompt them to use the word over, to indicate that it's the other one's turn to speak.
Have you tried that?
@@nope9310 I don't have an extra device to test it on, but it's just a suggestion that people can try.
The way that they actually did his order and talked to each other 😂
Btw is it just me but does the so ai guy sounds like the guy narrating Matilda 🤨
She's gonna ghost him just because he called her Kate.
Different Galaxies or nebulas arguing why they are much more awesome than the other.
Nice job! That was entertaining. Can you do a zombie apocalypse comedy scenario ? I just rewatched Shaun of the Dead🧟so that’s on my mind haha 🤣
Have chatgpt call pizza hut
How about not role play but a real scientific discussion. Maybe something theoretical.
If only it let you do impersonations. Christopher Hitchens debating Jordan Peterson on the existence of god.
I thinhk you should add some custom instructions to each model to add unique characters and temperament . Good job!
Great video idea. Bonds would not have come close to Aaron without juicing!
Get them to try to do the Who's On First Abbott & Costello routine
Oh I love that routine!!!!
Make they roleplay parts on classic fables like: The Wolf and the Lamb, The Lion and the Mouse, etc.
You need a third voice to be the theatrical director
i found success in telling ai i have a gf, and want to workshop conversation techniques with them, and using ai as a practice, then only asking for the quote after
Next time, ask each of them to first switch their character, for example, to grumpy or happy and excited-you get the idea. The goal is to make their voices sound more dynamic and less monotonous.
This was hilarious 😂
Try to jailbreak the broke up scenario I think it can be hilarious
Oh and see if they can play rock paper scissors or chess together lol 😂
So much better than notebookLM, i know not the same, but i cant listen to those voices on notebookLM.
TARS trying to convince HAL 9000 to help Dave.
Any of the above or other, but they are talking like in a musical.
I tried dnd light pg version but its restrictions stopped it from doing things like eating a chicken.
Ok a good one just came to me. You are a human and an alien has just landed in your yard and you have decided to try to speak to them. You are an alien who has just landed in a humans yard on Earth. They are approaching you and will interview you. As the conversation continues your answers to their question should become more and more surreal and mind blowing to the human.
I am having the same problem in like in this video where chat gpt don't wait my conversation end
I had a lot of trouble presenting brainzilla/sudoku style games (it's just a public example, it wasn't exactly those) to chatGPT, and this extends to several specific advanced subjects, but this is the kind of intuitive game that you don't know why it got it wrong, and it can't get it right;
could you bring o1-preview tests with these games in the future, do you think it will do better?
*for example, for this specific one, I needed an adaptation of the paq8px algorithm with a file generation algorithm without entropy (absolutely without any pattern, no repeated letters or bits) to derive a better pattern, he explains it well, but it didn't even come up for testing, I usually play shuttlecock with Claude, but none of the code worked; and I left it aside
Bitset and variable definition:
B={b_1·b_2·b_3|b_1∈{0,1}³,b_2∈{0,1}⁴,b_3∈{0,1}⁸}
*or B=random(3b+(4b\3b)+(8b\4b\3b)|3b∈4b∈8b)
S=⟨B₍₂₄₎,B₍₆₄₎,S_gr:₁₂₈G/₃₂b,S_bl:₄₃₂M/₂₇b⟩
Transformations and shifts:
B₍₂₄b₎=((R(3)≪5)∣R(5),(R(4)≪4)∣R(4),R(8))
B₍₂₄₎={b∈{0,1}²⁴| unique(b),⟨b₍₀₎₍₃b₎,b₍₁₎₍₄b₎,b₍₂₎₈b⟩}
Logical operations and constraints:
∀Wᵢ∈B:(Wᵢ∉B₍₂₄b₎)∧(Wᵢ⊕W₍ᵢ₊₁₎∉B₍₆₄b₎)
S(B₍₂₄b₎)={(B₍₂₄b₎[1:],i)|i∈[0,255],unic(S)∧(S₍₀₎≠B₍₂₄b₎[1])}
Success conditions and groupings:
S_succ(B₍₂₇₎(t))={B₍₂₇₎′∈Z₂₇:B₍₂₇₎′[0]"≠B₍₂₇₎₍t₎[0]∧bits 3 e 4 diferentes}
S(Wᵢ)={Wⱼ∈B|f(Wᵢ,Wⱼ)≠0},f(Wᵢ,Wⱼ)=(Wᵢ≪8)⊕Wⱼ
S_bl(b)={b′∈B₍₂₄₎|b′₍₁₎≠b₍₀₎∧group_constraint(b′)}
Progression and summations:
Progression(t)=∑₍ₜ₌₀₎ⁿ{write(bₜ)|bₜ∈B₍₂₄₎∧bₜ₋₁∈B₍₆₄₎}
A=i=1∑ₙWᵢ,n=min(16777216,|S(Wᵢ)|)
f(t)=∑₍ₜ₌₀₎⁽ᵗ⁾(write(B₍₂₄₎₍ₜ₎|S_gr₍ₜ₎,S_bl₍ₜ₋₁₎)+F)
I think that was too specific
*another type of account that they also don't resolve:
((A/162127)^A^A)((B/69513)^B)(C/86645)+(C/86645) = log_D( (log_D((D/33373)±log_C(A/1305114724356555)) )≠slog_D(D/33373)
the answer would be a value of approximately (x/43556898810675666570±1/1305114724356555)
Their pauses are too long, they trigger the other to interrupt all the time
You can have a couple arguing about their needs. Maybe we'll be able to find the 2nd man to win an argument with a woman. 🤣
How bout 2 chatgpt characters talking each other about Stool chart and both are struggling from constipation, shares their simillar experiences.
what
have them re-enact scenes from books, movies, TV
An interesting idea. I haven't tried it but I bet it would object on copyright grounds. Maybe if whatever is chosen has passed out of copyright.
Star trek and start wars fans arguing why their series is so much better than the other.
Bro! I was searching for that exact experience yesterday!!! 😂😂😂
🇧🇷🇧🇷🇧🇷🇧🇷👏🏻, How they are talking to each other, lol
Pon subtítulos en español, hombre. Buen video
let it play role play of Akinator
"Hello sir, I heard ChatGPT Plus users can share free trial links! Would you mind sharing one with me if you have any available? I'd love to try this new voice advanced mode?
Bro there also have limits of advanced voice mode
@@dzn_999 per month na ! For free users
You should've used different voices.
I did in scenarios 2 and 3 :)
Maybe you could have done a man and a woman
I did in scenarios 2 and 3?
It's pretty cool, but still very far from a human conversation. It seem like i'm listening a scripted conversation of two characters from the game.
It's the restrictions and censorship that make it that way, sadly.
We went from literally nothing in 2021 to "hmmmm but it's not perfect, it's only 99.23% human like conversation, we're still faaaar away from a normal conversation" in 2024
I give it 3 years until it becomes perfect lol
@@nawabifaissal9625 Why you're so upset from my comment, bro? I just gave my point. Why i should be happy if it's still not perfect? It's okay for current moment, but still not good enough. Let's wait to see of what is going to be after a few years later.
@@XIRON86 i'm not upset lol, it's just weird how we normalize these things as if 3 years ago this kind of tech wouldn't have looked completely insane, at this rate of progress it'll become perfect very soon investments are going crazy
@@XIRON86 you could literally just ask it to properly respond how you want for pretty much all use cases... Also it doesn't sound natural because they always have something to talk about, it's a robot, they are forced to respond due to their programming, and it has way more knowledge than the average person