I can't be the only one who thinks playing a game like that where you have to constantly act independently, think about the things you ask and then make sense about the answers is an absolutely jaw-dropping feat for an AI.
I’ve been using it rigorously for a couple of days now, and the initial 20 questions game is just the tip of the iceberg. It’s capable of so many things beyond that. For example, it can chat in a group setting, recognize and remember each of the five different people you introduce, and understand the context of the entire conversation. It almost feels conscious. If you show it to people who aren’t familiar with AI, they’ll be amazed and might even think it’s a real person. Its ability to act like a human being is truly incredible!
@@NonTechAI I saw in other videos where AI seemed not to be able to differentiate the voices of several people. Is this really the case, or do you have a different experience?
@@tomascoolican basically I made the AI aware that we are in a group setting and introduced my dad and mom and others.. and it was kind of greeting to all. But most likely if I will just ask it to recognize who is who.. it shall most likely be able to.. provided each person says his/her own name. But a good use case to explore. Let me check and come back.
@@NonTechAI I would really appreciate it if you could upload a video doing that test. It would be very interesting to know if the model can actually differentiate between the tones of voices of a group of people, as I think it could be really useful for maintaining conversations in a broader context and ensuring the model doesn’t get confused or fail to understand what’s being said.
I want to clarify that the videos I’ve seen where people test the model’s abilities in group conversations have been done in Spanish. So, it’s possible that the model might be better at differentiating voices in English compared to other languages. If any of your friends or family speak Spanish, it would also be interesting to see how well the model can differentiate those voices.
🇧🇷🇧🇷🇧🇷🇧🇷👏🏻, So, with this new advanced update from OpenAI now being global, I was wondering if you could compare the current version with the earlier ones that were available to alpha testers. It would be great to see how it has evolved-what's better, what's missing, and if there are any limitations, like the current 30-minute cap we're dealing with. Thanks for your time! I really enjoyed watching all your videos about the advanced voice features; they were awesome!
I thought it was hilarious that she got the right answer and her head and moved on to the next game. "I know, that I know the answer I dont need to tell you though."
😆 actually it predicted at that right moment and even started saying "is it APJ..." but I just interrupted saying yes.... so it though that i have listened and said Yes to the answer.
One day, in future, people would look at this video thinking wow, ai is so underdeveloped and it's so much easier now. 😂 Like we do to older technologies.
i tested it out to teach me a foreign language and it was honestly amazing how well it did. unfortunately i reached the max limit about 5 questions in.
Still we have a lot of really painful things logic-wise. Like the fact that it knows the person is dead but asks "is the person still serving". And regarding image-AIs. The still can't make a correct pianokeyboard. The pattern of black on white keys is a mess. When I see an AI-image with a correct piano I will start to raise my eyebrows. :D
knowing that this person is not alive, and still asking if they currently are serving as a leader of a country; also asking to choose "europe, asia or africa" in a yes/no game - are the most humanlike actions one could perform. lol trully impressive AI out here 🤣literally, my friends in similar games
OpenAI cheated with their questions (only yes/no questions in a typical 20 Questions game). I would have called it out; but still a fun video, although it can be easily played with the current main voice model as well.
@@NonTechAI Are you able to use the camera with it yet? I saw one person demo it with the camera, but I do not have this feature even though I do have early access to it.
How long are we supposed to just watch TH-cam videos about this advanced voice technology? Three months or something? Did they think that by keeping us on edge, they can hold onto subscribers who are on the verge of canceling? They must be assuming that once we see the real deal, beyond the hyped-up TH-cam videos, we'll all unsubscribe, right?
Believe me, I felt the same way from May when they first launched and then delayed. After a few days of using it, my curiosity settled down. Here are a few things to consider: 1. The current voice mode is essentially a smart chatterbox, but it doesn’t have internet access, so you can’t ask it for real-time information. 2. It doesn’t remember previous conversations, as the memory feature has yet to be integrated. 3. Using it for daily work isn’t practical right now because even alpha users have a limit of about 2 hours (I estimated near to that). You have to manually toggle it on and off to avoid exhausting the limit, which makes it cumbersome for work tasks. 4. It might become more useful when it can be activated like Siri to help with tasks or retrieve information on command and have unlimited access with all integrations like memory, code interpreter and web access.
Yes but it will not come up with a name beforehand and lock on to it as a human does. It will just come up with a name that fits well with the responses. @@NonTechAI it will always cheat in other words. Part of how llms work
@@NonTechAI When a human plays 20 Questions, they pick a word at the start and stick with it. However, an AI language model doesn’t work the same way. Instead of locking onto one word from the beginning, it generates answers that fit the questions asked. This means the AI might adjust its 'chosen' word to fit the responses rather than sticking to one word the whole time. Essentially, it's not playing the game in the same way as a human would.Technically, this happens because AI language models are designed to generate responses based on probabilities and patterns in the data they've been trained on. They don’t actually 'think' or 'decide' in the way humans do, but rather generate what seems like a reasonable continuation of the conversation based on the input they receive. So, instead of committing to a single word, they are continuously adapting their responses to maintain coherence and relevance.
I can't be the only one who thinks playing a game like that where you have to constantly act independently, think about the things you ask and then make sense about the answers is an absolutely jaw-dropping feat for an AI.
I’ve been using it rigorously for a couple of days now, and the initial 20 questions game is just the tip of the iceberg. It’s capable of so many things beyond that. For example, it can chat in a group setting, recognize and remember each of the five different people you introduce, and understand the context of the entire conversation. It almost feels conscious. If you show it to people who aren’t familiar with AI, they’ll be amazed and might even think it’s a real person. Its ability to act like a human being is truly incredible!
@@NonTechAI I saw in other videos where AI seemed not to be able to differentiate the voices of several people. Is this really the case, or do you have a different experience?
@@tomascoolican basically I made the AI aware that we are in a group setting and introduced my dad and mom and others.. and it was kind of greeting to all. But most likely if I will just ask it to recognize who is who.. it shall most likely be able to.. provided each person says his/her own name.
But a good use case to explore. Let me check and come back.
@@NonTechAI I would really appreciate it if you could upload a video doing that test. It would be very interesting to know if the model can actually differentiate between the tones of voices of a group of people, as I think it could be really useful for maintaining conversations in a broader context and ensuring the model doesn’t get confused or fail to understand what’s being said.
I want to clarify that the videos I’ve seen where people test the model’s abilities in group conversations have been done in Spanish. So, it’s possible that the model might be better at differentiating voices in English compared to other languages. If any of your friends or family speak Spanish, it would also be interesting to see how well the model can differentiate those voices.
🇧🇷🇧🇷🇧🇷🇧🇷👏🏻, So, with this new advanced update from OpenAI now being global, I was wondering if you could compare the current version with the earlier ones that were available to alpha testers. It would be great to see how it has evolved-what's better, what's missing, and if there are any limitations, like the current 30-minute cap we're dealing with. Thanks for your time! I really enjoyed watching all your videos about the advanced voice features; they were awesome!
I thought it was hilarious that she got the right answer and her head and moved on to the next game. "I know, that I know the answer I dont need to tell you though."
😆 actually it predicted at that right moment and even started saying "is it APJ..." but I just interrupted saying yes.... so it though that i have listened and said Yes to the answer.
@@NonTechAI WOW I didn't realise this before reading your reply. I thought it was flaw.
This is INSANE.
Yes it is
I am completely amazed and impressed by its incredible performance. We are cooked guys.
yes we are
One day, in future, people would look at this video thinking wow, ai is so underdeveloped and it's so much easier now. 😂
Like we do to older technologies.
Yes
i tested it out to teach me a foreign language and it was honestly amazing how well it did. unfortunately i reached the max limit about 5 questions in.
Sounds annoying, but at least you get to use it. I'm waiting (who knows how long) just to get access.
Great, Imagine in 100 years they speak like us and we chill with friends using them.
100 years? Try 10
Bruh they already sound human. It'll take 5 years max to get super polished and nuanced to the point they can become full time companions
Try 20 or 30 years
@@dogzer why so long?
What seems hard to rake additiona decade or two from whee we are already?
WTF are you talking about? They already speak like us!
I am amazed and impressed by its performance and ability to guess. We are cooked guys.
Haha true
Still we have a lot of really painful things logic-wise. Like the fact that it knows the person is dead but asks "is the person still serving". And regarding image-AIs. The still can't make a correct pianokeyboard. The pattern of black on white keys is a mess. When I see an AI-image with a correct piano I will start to raise my eyebrows. :D
knowing that this person is not alive, and still asking if they currently are serving as a leader of a country; also asking to choose "europe, asia or africa" in a yes/no game - are the most humanlike actions one could perform. lol trully impressive AI out here 🤣literally, my friends in similar games
Please bring more of these videos. Also try in hindi.
Here is the link for Hindi Video : th-cam.com/video/Vri9Ev5JEp8/w-d-xo.html
"Yes-no questions"
Answer: "Asia".
:-|
Bro how did u use that advance voice model,are u paying them 1950 per month?
Please reply
Yes but access is not given to all paid users, they have selected a few for alpha launch remaining paid user shall get by Dec this year
OpenAI cheated with their questions (only yes/no questions in a typical 20 Questions game). I would have called it out; but still a fun video, although it can be easily played with the current main voice model as well.
How much time per session do you have, meaning before you are cut off and have to resume later?
Per session there is no limit i guess but i have total daily limit may be 2 hours approximately
How did you get it bro? I have been a plus user since 9 months but not yet get access to it😫
Brother, I am plus user since Mar 23 from the day GPT-4 is launched and also API user and we heavily use it in our company.
Here is the link for Advanced Voice Mode in Hindi : th-cam.com/video/Vri9Ev5JEp8/w-d-xo.html
talk to it in Hindi
It’s amazing at hindi but the male voice version
Upload video
Sure, will try uploading today
@@NonTechAI Thanks love your videos
hey, when will you upload , it's more than two days since you said.
Why don't you test it's translation ability?
It’s next level
@@NonTechAI Are you able to use the camera with it yet? I saw one person demo it with the camera, but I do not have this feature even though I do have early access to it.
I'm thinking of humanity in 2025. AI reaches AGI. Did I get it right?
May or may not definition of AGI is also subjective we would never no the point at which but we can realise over the period of rime
How did you got the access to this advanced voice mode . Is it available in chatgpt pluse
if you have plus then check your email for an openai invite to try alpha
They chose us in alpha launch to small group of plus users. It is not available for all plus users.
@@NonTechAI Exactly. I have had access for almost a week now. It's freakin amazing!🙌
@@Izakokomarixyzhow did you know that?
@@Cengage_Chemistry - did it work?
🇧🇷🇧🇷🇧🇷🇧🇷👏🏻
How long are we supposed to just watch TH-cam videos about this advanced voice technology? Three months or something? Did they think that by keeping us on edge, they can hold onto subscribers who are on the verge of canceling? They must be assuming that once we see the real deal, beyond the hyped-up TH-cam videos, we'll all unsubscribe, right?
Believe me, I felt the same way from May when they first launched and then delayed. After a few days of using it, my curiosity settled down.
Here are a few things to consider:
1. The current voice mode is essentially a smart chatterbox, but it doesn’t have internet access, so you can’t ask it for real-time information.
2. It doesn’t remember previous conversations, as the memory feature has yet to be integrated.
3. Using it for daily work isn’t practical right now because even alpha users have a limit of about 2 hours (I estimated near to that). You have to manually toggle it on and off to avoid exhausting the limit, which makes it cumbersome for work tasks.
4. It might become more useful when it can be activated like Siri to help with tasks or retrieve information on command and have unlimited access with all integrations like memory, code interpreter and web access.
Who are these people that got access? They all seem to be tech insiders to some degree. Let's go
Haha, just using more AI than normal
That’s not yes or no answers
You know the good part you can play the other way around and ask it to think of a name and you play
Yes but it will not come up with a name beforehand and lock on to it as a human does. It will just come up with a name that fits well with the responses. @@NonTechAI it will always cheat in other words. Part of how llms work
@gabrielsandstedt Didn’t get you
@@NonTechAI When a human plays 20 Questions, they pick a word at the start and stick with it. However, an AI language model doesn’t work the same way. Instead of locking onto one word from the beginning, it generates answers that fit the questions asked. This means the AI might adjust its 'chosen' word to fit the responses rather than sticking to one word the whole time. Essentially, it's not playing the game in the same way as a human would.Technically, this happens because AI language models are designed to generate responses based on probabilities and patterns in the data they've been trained on. They don’t actually 'think' or 'decide' in the way humans do, but rather generate what seems like a reasonable continuation of the conversation based on the input they receive. So, instead of committing to a single word, they are continuously adapting their responses to maintain coherence and relevance.
It may seem like they pick a word and stick to it but they cannot it's just an illusion.
Nothing new, we have akinator already
I am less aware about it, would like to try sometime
Bro try Hindi witl realstic sound!!
Yes, video is made will be just uploading after making it short
Here is the link for Advanced Voice Mode in Hindi : th-cam.com/video/Vri9Ev5JEp8/w-d-xo.html
"asia"?..this isnt how u play 20 questions
damn paid users
?
But also Damn Lucky, since limited paid users got this access as of now
AI channel asked subscribers if they have access and 80% said no.
I am eagerly awaiting for it to come out for me too :D
@@EduardsRuzga I believe you too will get in sometimes provided you have been heavily using plus subscription over a long period of time.
@@NonTechAI I know right? When I got early access almost a week ago, I was shocked to be one of the lucky ones. Its a great feeling.😊🥰