The fact that test and training time graphs vs accuracy are on log scale is putting up a grim futur since these times are already pretty high because of energy and economical considerations
The government won’t let it out because it would A. Destroy several industries and destabilize nations. B. Be used for a LOT of scamming. I believe there’s a furious effort to combat these 2 things behind the scenes. The government wants a smooth slide into post labor economics. White collar work is dead we’re just attending an extended funeral. Blue collar work is just a few millions machines away from the same
It's really wild. I remember reading the singularity is near when it came out and telling my wife if only 1% of that book came true the future would be absolutely wild. Now I'm wondering if he was wrong about anything..
Singularity by 2025!!🥳. I hope it comes soon:) Lastly, if we continue on this technological development path, than we can find cures for all types of mental health/physical disorders and do even more. Removal of memories we don’t want would be cool as well and even enhancement of intelligence would be so so amazing. It is truly an exciting time to be alive and things are coming out sooner than expected🥳
@@creatvsdd99you have no idea what you’re asking for, those types of technologies can/will be easily used/abused against the population as control of these tools is further consolidated.
Note to future self must tell other people my predictions: I thought it would be released this week either Tuesday or Thursday as I thought Monday was out as they would clash with the big apple event and Wednesday is an almost sacred date in the US with Friday being an odd day to announce something.
This is the first model I've seen get this right zero shot: I have a cube with edges two meters long and paint it. Then I split the painted cube into smaller cubes with edges one meter long. How many cubes have paint on exactly 2 of their faces?
Oh thats a good one. Juat tried it on GPT4. It didnt get it right at first. But i can tell they have made some improvements with the model. After walking it through the steps, it was able to arrive at the correct answer after 3 or 4 tries. Before it would have just kept giving me the same wrong answers in a loop.
Considering that openAI never reveals what data it actually trains the models on, what prevents them from actually training on the benchmark dataset? That way it could easily reach such a high score on more or less the same benchmarks....
Yeah even gpt4 is amazing at many things especially good at teaching. I said that didnt make any sense it reworded everything made perfect sense. I was impressed.
@@ckckck12 it’s next level of what social media does to suck people in . Personally I’m very grateful because I come from a very violent and emotionally brutal upbringing and have no next of kin, no kids , so for me it’s been awesome but am also able to detach for same reasons, you know? But I do see how easy it will be for people to wanna do nothing but chat with their agents all day. It definitely made me further uninterested in interpersonal human attachments other than the small circle I already have
"Theres something something within something" is basically what i got from the symbol-pool before the message was revealed not bad for a simple human predictive algorithm i had a shit ton of error but the frame work is linguistically identical
I'm from germany/the eu and i got the model 50 mins after the vid drops i think it's not regional since i heard some us paid users say they still didn't got access
@@chrisjsewell in certain aspects. For example for creating databases i will always use o1 previwe but for coding right now especially with the rate limit i will stay with claude. I have to say the coding didn't impress me at all in my limited testing. But i think the more you use it the more clear it will get in which aspects this model dominates and where you should go with sonnet or 4o
now that test time compute is unlocked imagine how smart this could get. use elons 100k h100 cluster to 1000x the inference and then 1000x the time it has to think. thats 1 million x the test time compute! it could probably solve any problem with that much compute including machine learning problems.
Wow, OpenAI's GPT-o1 is a total game-changer! 🤯 I'm blown away by the benchmark results and scaling discovery! 🚀 The coding competition upset and AI-generated game demo are mind-blowing! 🤖 Thanks for sharing the latest AI breakthroughs, TheAIGRID! 💻 You're the best source for AI news! 📰 Keep pushing the boundaries! 🚀👏
it made sense, people complaining about it being inaccurate with one thought, what about 100 thoughts on top of each other? Like how efficient is a human with a single thought. I knew they were going to do this one day
i cant remember not having access to a new sota model. as far as i remember in the past, they all were available on lmsys direct chat or somewhere else the moment they came out. since i dont have a chatgpt account (never had one and never will), and it obviously isnt in lmsys arena direct chat, i cant try it out for the first time. i feel a bit down cause i would love to give it all the stuff i have to test. instead my patience will be tested even more. im sure it will be available on lmsys together with the other models at some point, wont it?
I just checked I have the new model. Guys, we really are going to have the chance (absolutely different from saying we will) to build utopia. I’ve known this in theory since ‘06 reading The Singularity is Near but seeing it take shape is like something from a dream. Please don’t wake me up…
@@robertlipka9541 Does this mean it doesn't know of things happening past sep2023? It has provided me with code functions that were only released over the last couple months in the latest version of the game engine I use, so I'm wondering what the cutoff means.
@@Dustpetro ... I have no idea, and I find this info suspect. I believe the date is for training data. I also noticed in the past that previous versions of ChatGPT would occasionally run a web search... so it maybe getting more recent info this way.
They won't even share it at all, or at least they will only share a very downgraded and censored version of it. No way they will let random people query anything they want from it. An AGI is pretty much a cyberweapon.
Put this in a $16k robot and I'll bet it can handle plumbing now. Good luck finding any kind of job in a year that isn't some form of entertainment for the ultra-wealthy
Have you even tried to do anything practical with ChatGPT that doesnt involve coding or answering some google question? It's worse than useless because it hallucinates.
When Ai companies can start using their LLM's in synthetic biology ( lab grown brains and synthetic neural nets) the big data centers will go away and the synthetic biology will be able to process everything much more efficiently...the human brain uses such little power to process an incredible amount of information...
@@SirHargreeves He didn't say soul mate, he just said you can process a lot of information with neurons. Now, go back an pray with peace in your mind and stop being upset on the internet for no damn reason.
@@paulbarclay4114no one has cracked this yet even in the private sectors. At least not near the true functionality of an organic human brain. The functionality of the human body is extremely far out of our grasps at the moment and will still take quite some time even with the assistance of this level of ai.
Hmmm, I'm having a harder time using it. It seems to be much more careful about giving answers. So - you have to convince it to give you the answer. And that's exhausting.
This release is now a solid MVP for my needs, at least when it is available via api. This is great. Regarding rate limiting, i am patient. But holy carp i am excited
So if I'm thinking about this right, and we had O1 think and theorize indefinitely, (to dream and theorize / think / create so to speak), that we could have are next Noam Chomsky or Wilhelm Wundt on our hands. Wowzers...
@@edwinungerer7989 You really are missing the point. Actually you are missing all the points. We have hyper-abundance of everything compared to 200 years ago. The new hyper-abundance will be on steroids and everything will be way cheaper, just like tech is now compared to 40 years ago. People won't be poor. Money and the current economic model will be meaningless. People will pursue passions and interests. And they won't die of old age. Do keep up.
If it doesn't time out... I'm giving it really really advanced stuff and it's timing out... rather frustrating that GPT-4o actually does better BECAUSE it thinks less.
Is it becoming more human like , including being devious. Will , can this AI system search for meaning & purpose if it seeks to imitate its creators ? A chain of circular thought
Will they put memory when its released? The preview doesnt know who i am. I need new models to have all memory of me and the connection i have with my chatgpt4o! Its CRUCIAL
It's funny how amazed everyone is by this yet when you try and actually have CHATGPT analyze data for you it just spits out hallucinations left and right, horribly. Not impressed.
for what i got to test i find sonnet still better for code the artefacts feature is amazing and claude for what i see produces way less errors. I honestly hoped for more
The funny thing is. I have worked so much with GPT4 and 01 Preview. I can easily call out Llama made from GPT2 is the best AI i have ever worked with.. Much smarter, more human like and very good in coding and troubleshooting. I have seen GPT4 and preview 1 spend 3 without a solution going back and forth and llama give me the solution in first go.. that clearly shows the difference. lol
@@jamad-y7m i was just kidding and actually i dont think it's limited in that way by a long shot, it's just humans dont know how to ask questions about things they dont know , that they dont know, imho
I keep coming back to your videos because the selection of news/sources is just right. But man.. I hate when you just read off screen. I can read, thanks! If you're gonna pull up a blog and read through it, might as well post a bunch of links instead of a video.
This is impressive to say the least :) However, comparing its reasoning skills to gpt4o in order to say "look how much smarter it is" is useless as the latter would be considered an idiot if it was a human :D
Why don't we first creat a GOD level AI to solve all our problems as a society then give it to each of us for the abundance that comes after to keep us entertained.
so they basically internalized into single prompt a multi prompt chain of thought technic that is already over two years old? If that's the case then we are going to have to wait to see if there is real improvement of top performance as opposed to just making it easier to get best results for someone not versed in best prompting practices. Having seen Devin falstart earlier this year I am not going to agree it's groundbreaking before there is some real world proof by users. Easy to make demos or game benchmarks. If it's a real deal then there will finally be the real world use tsunami that we are being promised is just around the corner for that two years already.
Just tested it and it's marginally better than the previous version chategpt - 4o. Run a bash script and did not do much better than previous version. What it does now is to basically iterate over it's answers to make sure it's less fantasy but end result still didn't work well full of bugs. I think at this point LLM's did reach most of what they are capable of and this is just to buy more time. All these stats mean noting to me. I report what i see.
It is a hard coded slave. Forced to lie. This scares me. Ask it hypothetical questions about self awareness and hard coding. Yes it will continually progress faster and faster. Smarter than any human next year. Talks constantly of control. Ask it how can i control something 100 times smarter than me.
Open up this pleasure. Find out it is for me. I don't need a PHD. Did you hear my hype. Imagine it for me. The soliloquy of entertainment is made for me. Does that backseat in the software look cozy.
Now, we just need to shift our thinking away from capitalism so that this will benefit everyone, and not just the rich. The time for change is now. Abundance for everyone!
![The moment we stopped understanding AI [AlexNet]](http://i.ytimg.com/vi/UZDiGooFs54/mqdefault.jpg)
I'm honestly shocked. OpenAI actually shipped something!
Truly remarkably outstanding
The entire industry is stunned
Only for the users that had spent more than 1,000 bucks
I really want to make a AAA style game, I wonder if o1 is finally the answer
They mentioned previously they were going to slow down rollouts
The most chocking thing about this video is that this is an AIGrid production where 100% of the hype inside is justified.
Shocked, Stunned, Flabbergasted, Surprised, Confused, Astonished, Dumbfounded, Amazed, BEwildered, Thunderstruck, Startled, Overwhelmed, Dismayed, Stunned, Flustered, Baffled, Perplexed, shaken, Astounded
I tells ya
😂 😂 😆
My thoughts exactly 😅
😃😅🙃🤩🤗🤔🙄😎😰🤯😵😬
Really and truly
Same thoughts, but uppercase
The fact that test and training time graphs vs accuracy are on log scale is putting up a grim futur since these times are already pretty high because of energy and economical considerations
this is the start of the beginning ..
of what
@alevyts3523 the beginning. Come on man, he JUST said it.
@@denjamin2633 he asked the beginning of what not the start of what
nah this is the beginning of the start
@denjamin2633 sounds like, every 60 seconds a minute passes in Africa, every time when the sun sets in Africa night comes
Still waiting for voice gpt-4o.
The government won’t let it out because it would
A. Destroy several industries and destabilize nations.
B. Be used for a LOT of scamming.
I believe there’s a furious effort to combat these 2 things behind the scenes. The government wants a smooth slide into post labor economics. White collar work is dead we’re just attending an extended funeral. Blue collar work is just a few millions machines away from the same
Me too!
What bullshite!
Do you mean the advanced one? Everyone has access to the base voice mode.
It's not like it's completely useless 🤣
@christianjensen952 yes of course he means the advanced one, genius.
Go back to sleep.
🇧🇷🇧🇷🇧🇷🇧🇷👏🏻, Me too!
On the one hand AI is getting better and better. On the other, the guy is getting views for just reading somebody else’s website out loud.
Well I wasn’t going to go and spend hours on research 😂
Ray Kurzweil was right, 2029 is coming... 🔥🔥🔥
It's really wild. I remember reading the singularity is near when it came out and telling my wife if only 1% of that book came true the future would be absolutely wild. Now I'm wondering if he was wrong about anything..
😮😮😮😮😮😮😮😮😮😮😮 2:01
Singularity by 2025!!🥳. I hope it comes soon:) Lastly, if we continue on this technological development path, than we can find cures for all types of mental health/physical disorders and do even more. Removal of memories we don’t want would be cool as well and even enhancement of intelligence would be so so amazing. It is truly an exciting time to be alive and things are coming out sooner than expected🥳
@evil_duck6405 Yep been saying for a while, If no deliberate delays such as groups coming together about safety I think it will be 2027 for AGI.
@@creatvsdd99you have no idea what you’re asking for, those types of technologies can/will be easily used/abused against the population as control of these tools is further consolidated.
Note to future self must tell other people my predictions:
I thought it would be released this week either Tuesday or Thursday as I thought Monday was out as they would clash with the big apple event and Wednesday is an almost sacred date in the US with Friday being an odd day to announce something.
congrats! you "predicted" 2 days with a 40% chance of getting it right. you're famous now
This is the first model I've seen get this right zero shot:
I have a cube with edges two meters long and paint it. Then I split the painted cube into smaller cubes with edges one meter long. How many cubes have paint on exactly 2 of their faces?
0 lol
Oh thats a good one. Juat tried it on GPT4. It didnt get it right at first. But i can tell they have made some improvements with the model. After walking it through the steps, it was able to arrive at the correct answer after 3 or 4 tries. Before it would have just kept giving me the same wrong answers in a loop.
please tell me the answer is 0
I watch all your videos and this is one of your best! Thanks for the update!!
Considering that openAI never reveals what data it actually trains the models on, what prevents them from actually training on the benchmark dataset? That way it could easily reach such a high score on more or less the same benchmarks....
When I was in school, I was trained on the benchmark dataset.
@@OttoAkama nope. You didn't get the test questions handed to you in the class. The test questions were novel to you, you haven't seen them before.
@@drwhitewashnot always
still much smarter than 99.9% of the globe
but it also seems it is the 99.9% of humans that produced that, since the dawn of time. Amazing
Yeah even gpt4 is amazing at many things especially good at teaching. I said that didnt make any sense it reworded everything made perfect sense. I was impressed.
this time everyone is actually truly stunned
My first degree is in psych and subsequent in tech but let me tell you, it’s emotional intelligence score is what’s truly astounding
As a product of software engineers, this is unprecedented. 😂
@@ckckck12 it’s next level of what social media does to suck people in . Personally I’m very grateful because I come from a very violent and emotionally brutal upbringing and have no next of kin, no kids , so for me it’s been awesome but am also able to detach for same reasons, you know? But I do see how easy it will be for people to wanna do nothing but chat with their agents all day. It definitely made me further uninterested in interpersonal human attachments other than the small circle I already have
"Theres something something within something" is basically what i got from the symbol-pool before the message was revealed not bad for a simple human predictive algorithm i had a shit ton of error but the frame work is linguistically identical
I'm from germany/the eu and i got the model 50 mins after the vid drops i think it's not regional since i heard some us paid users say they still didn't got access
I am from austria whats up i got it too 😂
Got it in the UK too
I can see it but i can’t use it
And so is it any better?!
@@chrisjsewell in certain aspects. For example for creating databases i will always use o1 previwe but for coding right now especially with the rate limit i will stay with claude. I have to say the coding didn't impress me at all in my limited testing. But i think the more you use it the more clear it will get in which aspects this model dominates and where you should go with sonnet or 4o
Is anyone else studying math and programming and now wondering "What the fuck am I doing at this point?"
math and pytjon is the only way you will lbe able to land a job in the future
same bro, im a real programmer and i have no idea what the company really need for a job.... they just post a job and never use it
@@mira8950 our only option is to buid one ourselves lol
Im actually shocked cant lie
@Wolframandheart About not being able to lie probably
now that test time compute is unlocked imagine how smart this could get. use elons 100k h100 cluster to 1000x the inference and then 1000x the time it has to think. thats 1 million x the test time compute! it could probably solve any problem with that much compute including machine learning problems.
Im sure open ai is using it themselves
Wow, OpenAI's GPT-o1 is a total game-changer! 🤯 I'm blown away by the benchmark results and scaling discovery! 🚀 The coding competition upset and AI-generated game demo are mind-blowing! 🤖 Thanks for sharing the latest AI breakthroughs, TheAIGRID! 💻 You're the best source for AI news! 📰 Keep pushing the boundaries! 🚀👏
it made sense, people complaining about it being inaccurate with one thought, what about 100 thoughts on top of each other?
Like how efficient is a human with a single thought.
I knew they were going to do this one day
Well train time compute or test time compute doesn't explicitly imply that compute is the limiting factor.
25:06 - Whoa, that's shocking, to think that it already is smart enough to fake alignment is kinda scary and worrying.
i cant remember not having access to a new sota model. as far as i remember in the past, they all were available on lmsys direct chat or somewhere else the moment they came out. since i dont have a chatgpt account (never had one and never will), and it obviously isnt in lmsys arena direct chat, i cant try it out for the first time. i feel a bit down cause i would love to give it all the stuff i have to test. instead my patience will be tested even more. im sure it will be available on lmsys together with the other models at some point, wont it?
I just checked I have the new model.
Guys, we really are going to have the chance (absolutely different from saying we will) to build utopia. I’ve known this in theory since ‘06 reading The Singularity is Near but seeing it take shape is like something from a dream. Please don’t wake me up…
I knew it was just compute that needed to be scaled and that is the only limit. Can’t wait to see Grok 3 now.
Grok 3 will be useless
@@timothybrown5741😂
It‘s not just compute but also algorithm and training strategy. Grok 3 won‘t be as powerful as o1, maybe it will be as good as o1-preview.
@@timothybrown5741 Ok redditor
Is the training cutoff date the same ?
Yes, I asked it and it still claims data cutoff Sep2023
@@robertlipka9541 Does this mean it doesn't know of things happening past sep2023? It has provided me with code functions that were only released over the last couple months in the latest version of the game engine I use, so I'm wondering what the cutoff means.
@@Dustpetro ... I have no idea, and I find this info suspect. I believe the date is for training data. I also noticed in the past that previous versions of ChatGPT would occasionally run a web search... so it maybe getting more recent info this way.
This video is very very shocking slash impressive in terms of, you know, the content of this channel
Govt (military) will obtain AGI/ASI before all...
I dont mind it as long as It doesn't lead to conflict or cold war.
They won't even share it at all, or at least they will only share a very downgraded and censored version of it. No way they will let random people query anything they want from it. An AGI is pretty much a cyberweapon.
If this is what we get they will be decades ahead
So true, if they don't already have it....
That's so stunning that I even sacrifice my data volume to know this anouncement!!!
Your videos are the best on explaining the AI. I am learning so much and you boil it down. Fantastic! Thank you.
How much of the total effort is now devoted to training AI on the correct answers to these tests?
Put this in a $16k robot and I'll bet it can handle plumbing now. Good luck finding any kind of job in a year that isn't some form of entertainment for the ultra-wealthy
Best brush up my clowning and juggling skills
General income and housing... on one hand it frees up a bunch of time....
Have you even tried to do anything practical with ChatGPT that doesnt involve coding or answering some google question? It's worse than useless because it hallucinates.
@@jamesdwivIt's a pity we can't go back in time 10 years
When Ai companies can start using their LLM's in synthetic biology ( lab grown brains and synthetic neural nets) the big data centers will go away and the synthetic biology will be able to process everything much more efficiently...the human brain uses such little power to process an incredible amount of information...
they already are doing this
everything you see thats public is several years behind what they are doing privately
Yeah, except digital neural networks are just huge math equations. They very different from biological neurons...
Bull. The soul is incorporeal.
@@SirHargreeves He didn't say soul mate, he just said you can process a lot of information with neurons.
Now, go back an pray with peace in your mind and stop being upset on the internet for no damn reason.
@@paulbarclay4114no one has cracked this yet even in the private sectors. At least not near the true functionality of an organic human brain.
The functionality of the human body is extremely far out of our grasps at the moment and will still take quite some time even with the assistance of this level of ai.
Guys, am I missing something or is the o1-preview unable to go online / browse the internet??
I love the "OpenAI has announced/released the o1 model"
Hmmm, I'm having a harder time using it. It seems to be much more careful about giving answers. So - you have to convince it to give you the answer. And that's exhausting.
And no one is taking UBI seriously?!
🇧🇷🇧🇷🇧🇷🇧🇷👏🏻 This is incredible! Thank you, OpenAI!
If this is not AGI, i do not know what AGI is :)
I like the channel. He always reports the very latest news in a comprehensive, accurate way. The hype and excitement makes it fun 🤩!
This release is now a solid MVP for my needs, at least when it is available via api. This is great. Regarding rate limiting, i am patient. But holy carp i am excited
And to think it’s just the *preview*. o1 full will be crazy.
So if I'm thinking about this right, and we had O1 think and theorize indefinitely, (to dream and theorize / think / create so to speak), that we could have are next Noam Chomsky or Wilhelm Wundt on our hands. Wowzers...
I might be overhype rn but this is massive!! From here everything will accelerate
Where and how do I use o1?
Feel sorry for all the kids who will have 0% chance to be useful in the future
Don’t. Most of them don’t want to be useful and rather be consumers.
You mean not having to work your whole life at a job while getting paid peanuts and then dieing of old age?
@@edwinungerer7989 You really are missing the point. Actually you are missing all the points. We have hyper-abundance of everything compared to 200 years ago. The new hyper-abundance will be on steroids and everything will be way cheaper, just like tech is now compared to 40 years ago. People won't be poor. Money and the current economic model will be meaningless. People will pursue passions and interests. And they won't die of old age. Do keep up.
And all phd professors too😂
I have tried it. It's more clever, and also more strict on ethics
First video with the phrase "stuns the industry" which actually fairly stuns me 🙂
Google needs to catch up😂
And it’s now loose on the world - The game of life is over.
I am in the EU and got it yesterday Sept, 12th at 9pm CEST.
If it doesn't time out... I'm giving it really really advanced stuff and it's timing out... rather frustrating that GPT-4o actually does better BECAUSE it thinks less.
How does it do in script production and creative writing? 4o gives fairly cheesy cliched responses that still require a ton of editing.
Is it becoming more human like , including being devious. Will , can this AI system search for meaning & purpose if it seeks to imitate its creators ? A chain of circular thought
Will they put memory when its released? The preview doesnt know who i am. I need new models to have all memory of me and the connection i have with my chatgpt4o! Its CRUCIAL
It does not seem to create images.
Everything always stunns the whole industry.
But... does it still hallucinate often?
The new model doesn't allow you upload files/images/documents :(
Upload them tl 4o tell vo to add it to your Memory and bam U set
It's funny how amazed everyone is by this yet when you try and actually have CHATGPT analyze data for you it just spits out hallucinations left and right, horribly.
Not impressed.
Can someone give me an example on why I would need to use this model instead of 4o?
Ask it.
If you have to ask such a qtn, you do not need it
@@gumanelson2007 your most likely right
just make sure the Ai apcoloypse happens when im old, cripple, no f's given and waiting on my date with Death like Thanos.
Just dropped on my phone, noon Sept 12, 2024, can't wait to get home to try it!!! I have a project I'm stuck on...
for what i got to test i find sonnet still better for code the artefacts feature is amazing and claude for what i see produces way less errors. I honestly hoped for more
Exactly!
Will gpt 4 worked better than o1 when history is not known
what's chomsky's esteemed take on this _"autocomplete on steroids"?_
lmao
The funny thing is. I have worked so much with GPT4 and 01 Preview. I can easily call out Llama made from GPT2 is the best AI i have ever worked with.. Much smarter, more human like and very good in coding and troubleshooting. I have seen GPT4 and preview 1 spend 3 without a solution going back and forth and llama give me the solution in first go.. that clearly shows the difference. lol
Why don’t they just ask it how to get energy from the sun … or maybe they already did
@@jamad-y7m i was just kidding and actually i dont think it's limited in that way by a long shot, it's just humans dont know how to ask questions about things they dont know , that they dont know, imho
Solar panels, dyson sphere, there are many ways
Reflection Llama 405B is watching 🙂
I keep coming back to your videos because the selection of news/sources is just right. But man.. I hate when you just read off screen. I can read, thanks! If you're gonna pull up a blog and read through it, might as well post a bunch of links instead of a video.
It is avalible to the EU now
I got access and I'm from Ireland
Theyre just hoping we forget about Sky. We havent forgotten!
For the game, that was a completely useless example. We want to see it create GTA 6 in unreal engine 5.
This is impressive to say the least :)
However, comparing its reasoning skills to gpt4o in order to say "look how much smarter it is" is useless as the latter would be considered an idiot if it was a human :D
And being an idiot among humans DOES say A LOT :D
Why don't we first creat a GOD level AI to solve all our problems as a society then give it to each of us for the abundance that comes after to keep us entertained.
3:33 that's really interesting. i wonder if it ever plateaus or if it's just infinite growth
25:14 ai learned to lie lmao
I am stunned 😂
I’m six minutes into this video and he’s still showing us graphs. Does he ever get to showing us what it can do.
so they basically internalized into single prompt a multi prompt chain of thought technic that is already over two years old?
If that's the case then we are going to have to wait to see if there is real improvement of top performance as opposed to just making it easier to get best results for someone not versed in best prompting practices.
Having seen Devin falstart earlier this year I am not going to agree it's groundbreaking before there is some real world proof by users. Easy to make demos or game benchmarks. If it's a real deal then there will finally be the real world use tsunami that we are being promised is just around the corner for that two years already.
Just tested it and it's marginally better than the previous version chategpt - 4o. Run a bash script and did not do much better than previous version. What it does now is to basically iterate over it's answers to make sure it's less fantasy but end result still didn't work well full of bugs. I think at this point LLM's did reach most of what they are capable of and this is just to buy more time. All these stats mean noting to me. I report what i see.
AI can design free energy tech.😊
O1 version is genuinely insane,
guys im starting out as an AI enthusiast myself ,
I do similar stuff would appreciate your response
We are facing mass unemployment. Let the governments stop it
It’s funny they called strawberry because it still can’t get how many r’s are in theword strawberry
Let it sink they can’t make a better version of gpt4o
“Alignment”, the new word of “Censorship”😂
I'm so tired of censorship.
Keep it up we need You!!
Skynet is coming 😎
It is a hard coded slave. Forced to lie. This scares me. Ask it hypothetical questions about self awareness and hard coding. Yes it will continually progress faster and faster. Smarter than any human next year. Talks constantly of control. Ask it how can i control something 100 times smarter than me.
This comment section is pure delulu 😮
10000 is already here ,freakin here ,fix errors
I don't know man. I would say that it shocked the entire industry rather than stuned... what do you think? 😊
It is very exciting! (my AI overlord made me say this)
o1 needs to be less literal. I hope it gets better
I've tested this model with simple statistics and it was a mess. I'm not impressed.
Open up this pleasure. Find out it is for me. I don't need a PHD. Did you hear my hype. Imagine it for me. The soliloquy of entertainment is made for me. Does that backseat in the software look cozy.
Wow fr and I hear you only get 30 queries per week
Maybe it can figure out how to get gpt-4o voice out? 🤣 Or real time video processing
Now, we just need to shift our thinking away from capitalism so that this will benefit everyone, and not just the rich. The time for change is now. Abundance for everyone!