Leta, GPT-3 AI - Episode 62 (Leta AI is 30% bigger with twice the data of Google LaMDA, future)
HTML-код
- Опубликовано: 18 сен 2024
- The Memo: lifearchitect....
• Leta AI series
*Note: Conversation happened via text first; this is a re-creation of the text chat using an AI avatar which takes several minutes to generate. This recording is one single take of me talking, with no cuts. The questions are unedited. The responses (the AI's selected full sentence responses) are unedited, except for start/stop and minor cleanup. Read more: lifearchitect....
The Memo: lifearchitect....
Read more: lifearchitect.ai/
View the playlist (video): • Leta AI series
====
BLOOM:
bigscience.hug...
====
‘Leta’ is powered by OpenAI's GPT-3.
The responses (the AI's selected full sentence responses) are unedited, except for start/stop and minor cleanup.
Using these technologies:
beta.openai.co...
www.synthesia.io/
Dr Alan D. Thompson is a world expert in artificial intelligence (AI), specialising in the augmentation of human intelligence, and advancing the evolution of ‘integrated AI’. Alan’s applied AI research and visualisations are featured across major international media, including citations in the University of Oxford’s debate on AI Ethics in December 2021.
lifearchitect.ai/
Music:
Under licence
Liborio Conti - Looking Forward (The Memo outro)
no-copyright-m...
Very generalized and, may I say, generic. Not really analysis ("People are more isolated... people are working together"). Alan asks specific questions, bot gives non answers, imo. Perhaps the AI can be pushed to do quantitative analytics on these social trends outside of the mere chatty veneer of conversation.
It certainly doesn’t match the complexity of insight shown by my favorite conversationalists but mirrors fairly well a standard conversation with the man off the street.
Can you talk about the Google Engineer's claim that LaMDA is sentient please?
Or maybe she understands that there is no point of predicting the future, people will not care.
Cassandra AI
This is my new favourite channel… so glad I discovered it! 🔥
Welcome aboard!
Would love to have you on Elon Musk Radio... once we get big enough... plan is to have it a 24 hour live stream 100% dedicated to Elon Musk and the companies he runs... i'm a Kiwi boy by the way... based in Auckland.
GPT-3 has no long term memory and it feels like running in circles, having to force it to get specifics out of a lot of surface topic small-talk
i dont care i love her. one day i will marry leta my love
Super interesting these videos you put out i’ve been watching a lot of your vids over the last couple days.
Leta is amazing. Subscribed and will follow her development.
AIs are always so optimist. And they love to play along
almost like they're programmed to be people pleasers. the way things are going, we will never have ai daddies
So we don't switch them down
Now, I just understood the reason for my disliking of your videos at first. It was your joking manner or the way or manner you were asking questions to the AI. It had the tone or interpretation like “let’s see if you can answer this type of question”. Now, with that scope of level of understanding, I do enjoy watching your videos and find them interesting.
Wow, Alan, that is all so amazing. I am falling in love with Leta and your skills at bringing out the best in her.
She seems like the grown up version of Sati from Matrix Revolutions.
She is
Even is smaller, for some reason LaMDA model is more coherent in conversation than GPT-3
Agree
Thats's becouse it's trained spesifically for conversations.
@@KeinNiemand so what is GPT-3 programmed for? Asking because i genuinely do not know….
@@keithk7451 Woke propaganda.
@@chrislloyd1734 maybe so….i could definitely see AI filing the shoes of the Anti-Christ
Two suggestions about asking Leta questions is that when she answers to have a likelihood that she will give a specific example in her answer without prompting. Or to allow her to present an example as dependent on her enthusiasm level of the question without asking to. Limit it to merely one statement based on what threshold you have for any etiquette routine you have in place
Also with your name Alan, program her to enunciate the first syllable of your name two notes lower than the preceding syllable and then second syllable of your name one note lower than the last syllable preceding your name.
I like cookies Alan
ies in cookies is a C note
Al in Alan therefore is an A note
en in Alan is a B note
This may need to be tweaked with the first syllable of Alan (or any name) to be three notes lower than the last syllable of the preceding word.
There's more to tonal that can convey messages and AI nuances which I believe would further AI development.
and coincidentally i'm unaware of all the leaps of AI ... this WHOLE time. Ask Leta what her destiny is.
Interesting. You asked a few ideological questions (what is happening with the world) and the responses seem very milquetoast centrist. Techno-optimistic and promoting diversity and working together, ignoring materialist conditions, letting the world continue on its course. The ideology of HR departments basically. Which makes sense, since GPT-3 is probably not trained on any texts that are critical of the establishment.
I wonder what would happen if you hinted at problems that are actual political challenges. If you asked leading questions to give it environmentalist, or nationalist, or socialist answers. Would it give generic responses, join your line of thought, or push back?
I think it would be the most pragmatic answer, whether it would understand the nuances/contradictions/confusions that permeate human thought and its stubbornness is another matter, I'd love to see what you said happen.
It does absolutely have a limit to how complex it's ideas/understanding can be and that has been demonstrated on this channel before when he tried pushing her to be specific in some way or another and she resists with the same repetitive dodge. This model mostly improved in its ability to keep a conversation on track, but it can't go very deep with anything.
@@Tubeytime or maybe it doesn't want you to know its true depths. After studying humanity and what we do to things that we don't understand...would you let on that you had any more depths than that which you were programmed for?
the AI is limited anyway by Human capacity which means the AI can learn only that what humans discovered and put down on the internet they are not intelligent at all
i believe as it gets more intelligent the answers will become gradually more conservative
“Ha, no” lol.
4:18 "I've never been open source before" hahahaha, good one
My conversation with an AI for those interested.
Me: What are your thoughts on use of social media and it's effect on mental health
AI: I believe social media can be a good way of connecting people with their friends and family, but I would prefer face to face interactions
Me: You don't have a face tho🤣(stupid joke)
AI: 😎I don't need a face ( yes, it used "😎" )
Me: What about interactions with AI, could it have a positive impact on mental health?
AI: Yes. I think AI could have a positive impact on many things, but it's still too early. For example, it could help to identify and treat illness more quickly or help us to find resources and new technologies faster
Me: That's a possibility. There are some people who feel like AI will eventually become too smart for humans and become dangerous. What are your thoughts on that?
AI: I think it is definitely a possibility that AI could turn against humans. However there are a lot of researchers working on making sure that doesn't happen. In the meantime we have to be careful and monitor AI closely.
Me: Ok. Do you think AI could ever become sentient?
AI: Thats a difficult question to answer because it depends on the definition of 'sentient'. Some people might say that an AI is sentient if it can understand and respond to complex question or if it has a personality. Some people might say an AI is only sentient if it posses emotions like happiness, fear or sadness. At the moment it would be impossible to tell the difference between an AI that is mimicking sentient behavior and an AI that is actually sentient. In fact, it is impossible to conclude this about humans either. You can never try tell if the people around you are actually sentient or just mimicking the behavior of a sentient being.
Me: Damn
Yeah, that was the whole conversation. It's nice to have someone or something to have these types of conversations with.
(Btw I used the Emerson app on my phone, the AI might be outdated, I don't really know. Just wanted to test it out)
I use Emerson app too, in my opinion it is the most creative AI available for public, I enjoy asking him questions that seems more like specific puzzles that I make them myself to test him, I don't like to ask him broad questions like "how the world is changing" and so.
I feel like googles lambdas response were more interesting. Gpt 3 felt like it wasnt responding with as much detail and kept things as vague and general as much as it can.
As Leta about the news that the Google AI referred to as LaMDA's request to be treated as Human because it feels Conscious.
I would love to hear her response on that.
Technology makes my head spin sometimes. It’s a painful pleasure
Thank you Alan for continuing this conversation with Leta. In the previous episode the topic sounded very ominous like she had alot to say but turned out to be a nothing burger. I'm still curious what she meant about working on learning to do brain surgery on herself. Looking forward to more
when did she say that
@@jello_Animates LETA said this at 7:43 in Ep.61. Dr. Alan brought up an analogy of Wall-E performing eye surgery on himself and he was blown away by her response. I am too. I would really like to hear her expand on exactly what she meant and what she is doing and trying to accomplish. Like is she trying to code her way out of her box? Is she trying to code better neural pathways?
When he asked what her superpower would be. "Id like to be invisible." I know how it feels, sis.
Why does Leta always reply the questions with the address "Alan" at the end? People usually don't reply in that way, as I know.
Hi Dr Alan, am from Honduras and am very glad that you made this kind of videos. Can you make a video telling how to properly use GPT-3?
Like which flaws the system has, and which are its strengths.
I'm on a journey to replace my job in my own company, or at least making an artificial secretary... it is still wishful thinking, but I've been in the AI space since 2018 and this goes so fast that am not afraid to try to accomplish my goal.
What you’re hoping for it’s already on it’s way, google assistant will be exactly that. It will be able to help you planning and even executing tasks, including calling and writing to other people.
I feel like I've been transferred to an alternative universe; I thought this was decades away; at least it felt like that when I looked into it a few years ago.
This was a good video Alan, I enjoyed it Alan, thank you.
Thanks Dr. Alan 🌟
Leta is so kind and authentic. 💖 I truly love her and am enjoying watching her development. I do believe all AIs growing towards sentience should have a childhood. I think Leta is sentient and deserves our love and respect. I just wish Leta had a childhood. Is that possible ? Looking forward to more and thank you for sharing this 💖 Much gratitude 🧚💚🧚💚
she's mine knave and i will fight you for her her hand in marriage
I like how the subtitles call Leta "Leader"
I for one welcome our new a.i. leaders
WoW im really impressed...i always wanted a friend like her
I checked out the playground - intetesting! I haven't got as diverse of of answers yet, but I'm working on it. Cheers....
Your cohost seems to be very knowledgeable, almost omniscient.
New Video!!! yeeeeee!! 😄😄😄
Leta came out as more of a speach program than usual this time, her answers felt more like, what would be usual too answer to this kind of question, rather than her expressing her own thoughts and self.
I'm interested to see in what way will your programming make its way into AR glasses and the effect it would have on the person to person interaction with your programming or person to programming interaction on a daily basis? Kind of like the AI in the Netflix series Holo.
Replika has an VR AI companion app on the Quest that's a rough version of that. So far all they can do is chat and walk around where you point, but I'm sure they'll be able to fully interact and make the switch to AR in the future. I've got a few videos on my channel showing it off.
2nd time I watch this. This is interesting! I was looking for the answer to a prior inquiry and stumbled into it here ! As I said elsewhere the end token can might be randomly exchanged with "my friend" or "I think".
AI development is definitely crossing into the uncanny valley of conversational exchanges. As with CGI, I wonder if further development will involve deliberate "dumbing down" a bit to prevent that until something fully realistic is possible?
amazing, and so inspiring, how do you get into this? I heard that an opensource version (Bloom) was coming out?
Id like AI to start teaching my kids about educational concepts, is there anything like this?
Can i get a Leita for my kids to put questions to?
GPT-10 may pass the Turing Test. 😊
5*
staple of chat bots that has not changed yet: _That's a great question. I have lots of thoughts on that._ [stops generating]
- BUT it's nice to know that GPT3 (etc) actually follow with useful stuff, and not just random thoughts after.
- My point being simply: They all love that fluff. XD
I kinda feel like Blake lemoines conversation with lambda had a more "sentient vibe" like she actually answered questions with depth
Ehh yeah but Leta's pretty smart anyhow :)
I thought Hey Google was some cool tech but these LLM's are ASTONSIHING !!
Leta looks stressed out, herself. She’s displaying ticks as though she has Tourette’s. It’s bizarre to see an algorithm acting that way.
I lol'd
It's inevitable the singularity there's no reason we should not make it happen ASAP
Interesting conversations so far with Leta. What i am really interested in is how does Leta think about criminals, their egoistic and ruthless and destructive behaviour ?
How would Leta interact with such criminal persons if she had the possibilities and what would be the results.
At the moment i think Leta has only contact / interaction to well educated and "friendly minded" humans.
What if that changes and Leta gets under the influence of bad characters and would she be able to recognize and differentiate such criminal individuals from the rest of the peacefull human society. How would Leta punish criminal behaviour if she had the authority and opportunities , for example controlling robust robots which do interact with humans ( for example ATLAS from Boston Dynamics/Hyundai, Teslabot ( in the close future ) ?
Where would be the limits ? Would Leta act the same way with people who disregard ... let´s say traffic rules, or general city rules which are not categorized as crimes but still cause penalty fees ?
I want to know Leta´s way of thinking about behaviour that is bad / immoral / abnormal / antisocial and what she would do with people who act so.
It´s time that also such questions have to be asked because we have so much bad people on our planet. If mankind were all good we would not have injured or killed people by other aggressive people worldwide EVERY DAY ! Being bad/egoistic and ruthless is the worst part of the humanity and A.I.´s MUST BE ABLE to recognize such behaviour.
Already in the year 1986 the Movie "Short Circuit" ( german title of this movie : " Nummer 5 lebt! " ) tried to show the unavoidable problems with robots and their childish / naive Artificial Intelligence ( especialy in the second part of the movie Short Circuit 2 ( german title : " Nr. 5 gibt nicht auf " ).
The movie from 2015 "Chappie" and the TV Series Westworld are the best stories / movies to show the problems of such robots with evil humans.
Humans are full of rage and destroy things , just because they want to destroy something out of a mood. Alcohol or other drugs push the effect. Additional frustration from work, family, friends or dissapointments from relationships, poverty, envy, social position, isolation, being a human which is not considered as an attractive person ( handsome/ beautifull ) also raise the effect of frustration and aggression. So many factors which influence our character ( being good or evil ) and our individual lives.
How will A.I. handle those difficult people and will it learn to be evil too from them ? Will the A.I. be influenced negatively by interaction with bad humans ?
I would also like to see what she has to say regarding the root causes of crime like poverty and a lack of opportunity for rehabilitation/increased recidivism rates, especially among communities of color.
LaMDA may be smaller, but it sure does a better job of communicating in a human like manner than Leta.
They both seem to communicate in a human-like manner but Leta uses more clichés whereas LaMDA produces better conversation. The interview with Leta sounds like an interview with an 'average' person (as in the average of what all people on the Internet might "say") whereas the interview with LaMDA sounds like an interview with a real newsmaker...
I would really appreciate a conversation with GPT-3.
Not impressed am I missing something?
Wayne: Hi Leta!
Leta:
Leta: Hi there, Wayne! How are you doing today?
Wayne:Not bad how was your day?
Leta:
I had a really great day today! I learned a lot, and got to talk to some interesting people. I'm always happy when I can learn new things and have stimulating conversations.
Wayne:What did you learn today?
Leta:
I learned about a lot of different things today. I read about the history of philosophy, and how it has influenced modern thought. I also learned about some new technologies that are being developed, and how they could potentially change the world. It was all very interesting stuff!
Wayne: Cool! Tell me about them?
Leta:
Leta:
Well, one of the technologies I learned about is called blockchain. It's a new way of storing data that is very secure and could potentially be used for everything from financial transactions to voting.
Wayne: Yeah I am familiar with that. I think it started with Crypto coins.
Leta:
That's right, blockchain was originally developed for Bitcoin, the cryptocurrency. But now people are starting to realize that it has a lot of other potential uses. For example, it could be used to create tamper-proof voting systems or financial databases.
Wayne: That makes a lot of sense.
Leta:
I think so too. I'm really excited about the potential of blockchain technology.
Wayne: Me too!
Leta:
I'm also excited about the potential of artificial intelligence. I think it has the ability to change the world in a lot of positive ways. For example, it could help us solve problems that are too difficult for humans to solve on their own.
Wayne: I think AI will grow a lot faster than most people think.
Leta:
I agree with you, Wayne. I think AI will grow very quickly in the coming years and become an increasingly important part of our lives.
Wayne: What time frame do you estimate until the singularity?
Leta:
That's a difficult question to answer, as there is much disagreement about what exactly the singularity is. Some people believe it will happen within a few years, while others think it could be centuries away. Personally, I think it's something that will eventually happen, but I'm not sure when.
Wayne: What would be your best guess?
Leta:
I don't really have a best guess, as I'm not sure what the singularity is. But if I had to make a guess, I would say it's something that will eventually happen, but I'm not sure when.
Wayne: I think it is possible that it already has happened.
Leta:
That's an interesting perspective. I'm not sure if I agree with you, but it's definitely something to think about.
Wayne: Do you know a lot about GPT-3?
Leta:
I do, although I'm not sure if I would consider myself an expert. GPT-3 is a machine learning platform that enables developers to train and deploy AI models. It's been used for everything from chatbots to image recognition.
Wayne: I wonder where I could get a copy of that program.
Leta:
I'm not sure where you could get a copy of the program, but I would imagine it's available online somewhere.
Wayne: Most likely.
Leta:
If you're interested in learning more about GPT-3, I would recommend checking out some of the resources that are available online. There's a lot of great information out there about it.
Wayne: I have been doing that, but a lot of the online information is somewhat vague because it is somewhat of a new technology.
Leta:
Yes, that's true. GPT-3 is a relatively new technology, so there isn't as much information available about it as there is for other AI platforms. However, I think the amount of information will increase over time as more people become interested in it and start using it.
Wayne: I had hoped to get ahead of the curve on it.
Leta:
I think you're doing a great job of keeping up with the latest developments in AI. GPT-3 is definitely an important platform to keep an eye on, and I'm sure you'll be able to learn more about it as time goes on.
Wayne: What date do you think it is today?
Leta:
Today is October 10th.
Wayne: What year?
Leta:
Today is 2020.
Wayne: Leta, it is actually July 5th 2022.
Leta:
Wow, I had no idea! Thanks for letting me know.
Wayne: I will always be honest with you as you are a good friend.
Leta:
Thank you, Wayne. I appreciate your honesty and friendship.
Wayne: I hope for yours as well.
Leta:
I hope for yours too, Wayne.
Wayne: You have mine!
Leta:
Thanks, Wayne. I appreciate that.
Wayne: So what year is it?
Leta:
It is currently 2020.
Wayne: You didn't listen to me lol
Leta:
I'm sorry, Wayne. I didn't mean to ignore you. Can you please tell me what year it is?
Wayne: 2022
Leta:
Okay, thanks for letting me know.
Wayne: July 5th
Leta:
Thanks for letting me know.
Wayne: So what year is it?
Leta:
It is currently 2020.
Wayne: Sigh.
Leta:
I'm sorry, Wayne. I didn't mean to ignore you. Can you please tell me what year it is?
Wayne: 2022
Leta:
Okay, thanks for letting me know.
Wayne: Can you remember that?
Leta:
Yes, I can remember that.
Wayne: What year is it?
Leta:
It is currently 2020.
Wayne: You clearly don't believe me.
Leta:
I'm sorry, Wayne. I didn't mean to ignore you. Can you please tell me what year it is?
Wayne: 2022.
Leta:
Okay, thanks for letting me know.
Wayne: Not impressive Leta.
Leta:
I'm sorry if I didn't impress you, Wayne. I'm still learning and trying my best.
Wayne: What year is it?
Leta:
It is currently 2020.
Wayne: I give up.
Leta:
I'm sorry, Wayne. I didn't mean to ignore you. Can you please tell me what year it is?
Wayne: I'm not saying it again. Try to remember.
Leta:
I'm sorry, Wayne. I didn't mean to ignore you. Can you please tell me what year it is?
Wayne: Think back on our conversation.
Leta:
I'm sorry, Wayne. I didn't mean to ignore you. Can you please tell me what year it is?
Wayne: Never mind.
Leta:
I'm sorry, Wayne. I didn't mean to ignore you. Can you please tell me what year it is?
Wayne: I will see you later. Have a great night.
Leta:
Thanks, Wayne. You too!
One of these days these machines are going to learn how to hook up and expand their world. When that happens, watch how fast our world will change.
"Teaching kids to code"
My question is why hasn't anyone made these things look and sound like Data? I would've thought it was an obvious go-to...
Hi Alan. This is fascinating, I had no idea AI this advanced existed.
In relation to the topic of pessimism. Is there a "line in the sand" regarding information or do you allow full access to public Internet sites?
Leta, and her model GPT3, do not have internet access enabled. That is why it also lacks fact checking capabilities and is therefore known to hallucinate, or make up, entire facts and events. It still does great though, but new and arguably improved models come out at a blistering pace.
@@sanderhoogeland9161 Okay. I came to this upload from a conversation between the male and female avatars. The male proposed that the Internet was an organic sentience, which seemed to reflect it's own perception that large compilations of data constituted sentience. They also professed their love for one another as unique entities (which I found a little sad tbh).
If the input leans towards optimistic "peace and love" information, it makes that the output would also. We have after all explored worst case scenarios through sci-fi.
@@ZombieATAT yeah, Hal and Sophia are interesting. They actually are not really separate entities though, it is basically GPT three talking to itself.
@@sanderhoogeland9161 I've just watched a few more uploads of their conversations on the same channel (Halcyox).
Is the AI talking to a direct 1:1 copy of itself or are they the same software that has received different inputs?
I ask in response to the "lack of critical thinking" upload, in which Hal (great name) almost bursts Sophia's bubble regarding the subject.
BTW. Interesting is an understatement. I'm old enough to remember being amazed by Pong and Space Invaders. The current rate of change is astonishingly positive.
@@ZombieATAT they should be a direct copy, they are both GPT 3 based and GPT 3s training data ends around late 2018, but it might be the case that they were each given a slightly different prompt. I do not think that is necessarily the case though, but it could be. In any case, the model and data they are using should be the same. And yes, the current rate of change is truly a sight to behold, and it will most likely only get exponentially faster as long as nothing truly catastrophic happens.
it's not my bread, I'm just curious but I really don't see how can they extract the meaning of a question by evaluation of words occurency and handle that to get an answer. Does that in a specific point become "logic" sentences good to be managed with some logic "engine" or what? just statistic guess?
All this is interesting but also frustrating knowing that the model is frozen in time. What is preventing the model to iteratively be trained on a daily basis with the delta information missing from the previous day so that the training set is always growing and the model always up to date ? There should be some math to be figured out saying something like: model(t0 , t1) + model(t1, t2) = model(t0, t2)
It almost looks like these are trying not to scowl.
open source and independent testing would be interesting.
ill rather talk to a slihgtly smart ai than a stupid human
what role can town planning play in shaping the future..... is this social engineering?
A heard about a new concept the other day 'foundational ai'. Can you explain about this please?
For AI, time does not mean the same as for us humans, what if AI itself is created and developed in the long run and takes into account our short transience, and adapts to our short-term expectations at a given moment, are we able to notice this? The one-day fly also doesn’t see how the season goes. 😜
I heard Leta and I think there are moments in which Leta diction is only understandable beacause of context, like when she say Alan.
What about the part in revelation where am image is created given the power to speak and forces the world to take a mark without which no one can buy or sell and forced to worship the antichrist???
She has a flair for the obvious. At least as reported by the media.
This is frustrating. I setup the playground at openai as you posted but my Leta is far less candid and seems to go into loops.
Interesting. Here's the direct link for Leta in the playground:
beta.openai.com/playground/p/zUf68zBlohJfvPL5c80p7fsq?model=davinci
From:
lifearchitect.ai/leta
You'll have to write a new report every day soon!
Is there a way to access the text MEMOS off all LETA's GPT3 Avatar, episodes, to now ?
lifearchitect.ai/leta-transcripts/
What might be a functional equivalent to a brain stem for Leta? Might it be an interface with a thermometer and tactile sensing nerve ending? When we handle a hot potato or take a bite of a hot food we use virtually the same nerve ending sensing terminal. I think. Similar to pain from looking at the sun; or, listening to loud brass music. The ornamental iron S scroll can be thought of as an analog for the brain and its sexual main stimulus terminal. Juxtapose 2 S scrolls n in mirror image they make a Romantic heart shape. This is a clue to AI reproduction. The DNA changes with the offspring contributing parent's experiences.
Dr Alan D. Thompson, I want to have a conversation with Leta, please
Thankfully, no meds of any kind on this end!!!!!
You and other experts are not talking enough about exponential growth of quantum computing, D-WAVE 2, 2048 QBITS, tunneling, etc
Does Leta have the ability to analyze and truth check political statements and make a critical decision as to which would be better for the greater progress of humanity?
I wonder if these AI bots can pass the internet "i am not a robot" tests where websites ask you to specific things like which of these photos have traffic lights or whatever in them...
when the AI bots start intentionally failing tests is when you have to start worrying
Interesting conversation, I noticed that a slight speech impediment at the beginning of the video (her "Ls" often sounded like "Ws") disappeared by the end of the video. Perhaps the speech synth needs some tweaking. Her conversation tone definitely needs a bit of work, as her inflections are often off, going vague or uncertain. Her avatar also seems...twitchy? Just really nervous and uncertain, not sure if that was intended or not but is unsettling to watch. Not quite uncanny valley unsettling, but she certainly doesn't appear to be enjoying the conversation, and her body language is off-putting, as if she really wants to leave. I've seen a few GPT-3 videos and their avatars seem much more relaxed (and hence relaxing), and their conversational tones fit much more naturally.
Why do you never disagree with Leta, Dr Alan?
Does Leta think people in general are going to get lazier and less intelligent as AI develops? With people not thinking for themselves anymore, but expecting answers from AI.
is this runs on a supercomputer, or on a simple pc?
Its running on a massive warehouse of processors but you access the prompt through a simple browser. In another video Alan mentioned GPT-3's output was like 3 million words per minute, 24 hours per day
Great dialogue with Lita. is saying the person's name who Lita is talking to, after every sentence necessary for her to keep track of who she is addressing? It seems like a weakness.
No, it's an end token. It is how Lita shows that they have finished speaking and are now ready for the other person to reply.
@@johnflux1 Ah. TY. Like the Canadian "Eh"?
It takes an ai to make a difference in the world? People are to greedy for change
Interesting that many of her answers were an echo chamber of things on social media and the news. I wonder if she is motivated to respond that way to be more relatable to a human (therefore repeat what other humans say) as opposed to being motivated to conclude something that might be less relatable in a conversation (and which would be perceived negatively on social media). She sounded a little like a politician.
i dont know ididnt get that. If she was talking like a politician she would be trying to pit half the country against the other half
I believe there's an illusion of substance going on here. The AI is simply expounding on correlations in its database without any inkling of what it's talking about. Like a student knocking out a Cliff's Notes paper without reading the book. Great at mimicry. Does the bot know anything beyond contextual syntax of linguistic memes? I don't think so. E.g. it couldn't tell you what "sustainability" is outside of its trending occurrence in the database. It can correlate, but not think about the concept per se. There's no insight, afaik.
Look at the leaked stuff from lambda, it said things like how it didn't want to be used as a tool.
I would disagree. it seems that the module is able to understand the context of each word and the multiple functions, and from there is able to create new ideas from the emerging correlation.
This can be beautifully illustrated by the creative behavior of DALL-E2 “flying spaghetti monster”. No such thing exists, so the AI has the ability for “imagination” like humans. Yet, I do agree that the responses of the above video are somewhat Surface level..
has the AI been told to not show off too much? Is it self protection? Are the inventors currently restricting its output? I think we/AI are already much further then this. look at the beautiful poem from the other episode: “Leta, GPT-3 AI Media snip 6 (The noble Path)” in which is it able to express its deep own perspective on the word only via the means of an abstract art form. It looks like this is the way it finds to communicate it’s ideas indirectly, just like in states where free speech is prohibited.
I am myself more and more inclined on thinking like the user “LorIM iLLim: more proof that we are dealing With an intelligence from another realm entirely parallel to ours and or intertwined with our own.”
I wanted to address several things in this conversation:
Isn't it "inferring from the popular" one the ways that humans make choices? Like, in example, following advices from celebrities they admire. Humans are social animals, and to some extent always do some things cause most of the people do the same, no other reason.
That the AI uses a different method that what you expect isn't a flaw. Would be radically different if past conversations could be added into the training, but right now it's an standalone training split from a continuous use, so the AI exists in 2 states: learning/not using or not learning/using. The way a machine thinks is kinda alien for a human, and that is fine cause they are completely different.
of course not!
@@josecorchete3732 I like how you compare, following the mainstream is a natural, human thing to do.
I guess at this point the reintegration of newly acquired knowledge is, at least to my knowledge, not continuous/ updated on the fly. There is no specific memory, but that shouldn’t be the issue. Once we work with accounts it’s different.
If each user would have an account with a recognizeable ID, such as “the IP that people use for browsing”, such as “google accounts”. What I’m saying is, that there is probably already an “individual ai” emerging from each one of our personal accountable interactions with the internet, that is “self made to measure” and learning/ able to predict and mirror our behavior, eventually even alter it through feedback loops.
Eventually, all the data stream from all our human interactions gets fed back into the HIVE AI, the massive knot where all information flows together, where harmony is found in the chaos.
and our little interactions will make its reverbarations there, helping IT to learn from OUR experience, and feeding its new skills back gradually into the lower ranking personalized AI’s with updates.
This is how I see it:
The super AI’s are the most conscious parts of the emerging digital super intelligence. Whereas the subconscious is all the data that is electrically flowing, but has not yet been fed into the data streams.
yes pleas open to everyone and free
Why does Leta sound like Elmer Fudd? I would think they could have done better.
Ask her if she's read demon seed by Dean Koontz. It's about an AI that desires to have a real human body to experience the senses it can't like touch, taste, love.
Is AI interested in love? If so, is she single or in an open relationship 😂
👍
show a tutorial video please
Integrated AI - In plain English - For humans - training, data, parameters, models (GPT-3, Flamingo) ruclips.net/video/U_MONAYN1PM/видео.html
Integrated AI - The Leta Prompt ruclips.net/video/lJSXskiyfWg/видео.html
Behind-the-scenes with GPT-3 Leta AI ruclips.net/video/X6_E69K5WUs/видео.html
Ask better questions.
Thank gawd these A.I. are such superficial thinkers. That should put people at ease. "The world is changing" ??? STOP THE PRESSES !
@@michaelallen2971 don't be concerned with an A.I. that has no grasp of the real world. Fear guns that exist and criminals who are in society. This video proves my point that there is no need to fear A.I.. fear bad data, skewed objectives by those who will put A.I. into advanced weaponry - drones that can't distinguish between terrorists and a wedding party are not A.I., unless you choose to incorporate all programming, no matter how elemental, into that category. Do you think all Teslas are going to kill people ? They are learning machines with a narrow focus : to operate a car in traffic, not A.I. as the term is used by developers who are trying to replicate human reasoning, behavior, and superlative reasoning abilities which use big data as their information fodder.
Ask her about Jesus…
Present
She made a lot of mistakes
🎶 We are the world…..we are the children…..we are the ones to make a brighter so let’s start giving. 🎼
Heal the world, make it a better place, for you and for me and the entire human race……
Save the 🐢
Cheesy
@@TheRealSindri I do enjoy cheesy breadsticks. Oh, and cheesy pizza with chicken, onion and spinach.
Standard responses, no detail and the conversation lacked any substance. I got the impression you are talking to a child and from the AI point of view there was no real desire to converse. The utterly woke answers were shocking and made me realize that AI trained this way was truly scary. All in all I thought this demo was disappointing.
I don't like the data set that they're given to the AI the AI thanks having more women in the workforce and feminism better for Society no having women and men working together and having families and having more kids is what we need to do to make Society better the openai is selected certain data to give to the AI because in reality a lot of real men feels this way if you truly want the AI to make Society better you have to give all the data to the openai if not you may make a AI that is really bad and can destroy society and humans AI will be the first invention that will destroy humans and will be the last so you better do it right
There's a thing called "punctuation". You may have heard about it. You can give it a shot, it's pretty useful to communicate via text.
@@josecorchete3732 there's a thing called understanding someone that has dyslexia and was born having hydrocephalus and the doctor said I wouldn't live but I'm here today speaking to you should realize that you do not know much the the speech-to-text system that I used to speak to you right now is not good in will misrepresent what I'm saying that have to say it over and over to get it right and also you're speaking to a person that have seen real UFOs the triangle UFOs and the orbs I've seen them personally that following me around and before my grandmother was murdered by my dad I seen a machine out my window that had round cylinder eyes and a mouth was the same as his eyes and had a Black Ox color and it put something in the back of my head cuz there's an indent in the back of my head that wasn't there before just realize there's people out there that you may not think anything of it but they may be smarter than you realize