Even Google's AI dont always work =>
https://www.forbes.com/sites/jonath...bot-gives-wrong-answer-in-ad/?sh=3ffd468355ce
^ They lost 100 billion because their AI chatbot didnt work lol
As I said, AI is not magic. Its in fact worse than other analytical solutions because the AI solutions are useless without good training data.
https://www.forbes.com/sites/jonath...bot-gives-wrong-answer-in-ad/?sh=3ffd468355ce
^ They lost 100 billion because their AI chatbot didnt work lol
As I said, AI is not magic. Its in fact worse than other analytical solutions because the AI solutions are useless without good training data.
Yeah i watched that live presentation and it was a bit embarasing 
They were forced to do it though cuz Microsoft announced the ChatGPT-Bing/Edge integration and even since ChatGPT 3 was made public they started to get pressured by the media......
That's true, AI has 3 limits for now
- ) crap data -> crap results ( though this applies to us as well, if i try to teach you incorrect things about a topic you'll produce crap things as well xD )
- ) our own ability to create advanced enough models
- ) computers are sure getting fast, but not fast enough for very large AIs like ChatGPT ( meaning that it still take a little to see how the algorithm/models can be improved )
( People will solve all of these with time ofc )
They were forced to do it though cuz Microsoft announced the ChatGPT-Bing/Edge integration and even since ChatGPT 3 was made public they started to get pressured by the media......
That's true, AI has 3 limits for now
- ) crap data -> crap results ( though this applies to us as well, if i try to teach you incorrect things about a topic you'll produce crap things as well xD )
- ) our own ability to create advanced enough models
- ) computers are sure getting fast, but not fast enough for very large AIs like ChatGPT ( meaning that it still take a little to see how the algorithm/models can be improved )
( People will solve all of these with time ofc )
No I don't think it's a tragedy. It's a comedy show.
You can do the same shit by photoshopping people on porn pics. It's nothing new.
What's funny is pussy knights reacting to it all as if she got raped and then beaten to near death, when all it was just a dude fapping to fake pictures/vids of her, in his house privately
You can do the same shit by photoshopping people on porn pics. It's nothing new.
What's funny is pussy knights reacting to it all as if she got raped and then beaten to near death, when all it was just a dude fapping to fake pictures/vids of her, in his house privately
Even if it was for private, it's still a disgusting and disrespectful act. You obviously got something wrong in your brain, maybe quit obsessing with a fictional character and go touch some grass
Depends on the analytical solution. Complex, real problem is almost impossible to solve analytically. Analytical solution is best if you can simplify some things. AI is also the most useful in engineering as it can significantly reduce the time and cost in numerical modelling as well as design topology optimisation
Yeah, your brain can't read the right way
How else is he supposed to share them? Print them and mail?
Oh shit, I didn't know I was talking to Jesus' mother Mary.
Good luck on giving virgin birth again
the bimbo turned him on to jack off, I don't see how that's disrespectful, if anything it's the opposite, she should be glad that her existence did something good for once
Why are you here, wagie? Aren't you supposed to be outside chewing grass? Why are you on a cartoon forum getting mad over a dude jacking off to a bimbo?
How else is he supposed to share them? Print them and mail?
Oh shit, I didn't know I was talking to Jesus' mother Mary.
Good luck on giving virgin birth again
the bimbo turned him on to jack off, I don't see how that's disrespectful, if anything it's the opposite, she should be glad that her existence did something good for once
Why are you here, wagie? Aren't you supposed to be outside chewing grass? Why are you on a cartoon forum getting mad over a dude jacking off to a bimbo?
@Nidai_Kitetsu
These "suboptimal solutions" are not enough to pass the bar for todays standard. You cant design a skycrapper structure analytically, you cant design an aerodynamic model of a modern aircraft analytically, you cant design an F1 frame analytically. Analytical approach only works at the preliminary design. With AI, a finite modeling of a structure that consists a matrix with millions of cells that can take up months to finish may be reduced in half in time and by extension cost. AI may be inaccurate for social needs, but it definitely is an amazing tool for science/engineering
These "suboptimal solutions" are not enough to pass the bar for todays standard. You cant design a skycrapper structure analytically, you cant design an aerodynamic model of a modern aircraft analytically, you cant design an F1 frame analytically. Analytical approach only works at the preliminary design. With AI, a finite modeling of a structure that consists a matrix with millions of cells that can take up months to finish may be reduced in half in time and by extension cost. AI may be inaccurate for social needs, but it definitely is an amazing tool for science/engineering
Last edited by a moderator:
That's a very generic statement and not true always. Take estimation theory for example. The optimal maximum likelihood (ML) algorithm is almost never used because it's too complex and slow. People are doing quite well with suboptimal least square solutions or some improved variants.
You could achieve optimal performance with AI but you will face the same complexity issue like ML, because to achieve such performance you would probably require a very complex AI method, lets say, a 6-7 layer deep neural network.
Now if you cant model something analytically, then AI could be very useful. Again he issue will be the quality of your training dataset, computational platforms etc.
It's an amazing tool but its not a magic wand. I am still not buying all these hype from media.
You could achieve optimal performance with AI but you will face the same complexity issue like ML, because to achieve such performance you would probably require a very complex AI method, lets say, a 6-7 layer deep neural network.
Now if you cant model something analytically, then AI could be very useful. Again he issue will be the quality of your training dataset, computational platforms etc.
It's an amazing tool but its not a magic wand. I am still not buying all these hype from media.
Then that's not on the AI. As you said, these methods are merely the things you want the AI to understand and implement. You can also use least square solution in AI. I agree though what im saying is generic, and it's based on the environment of my workplace.
If you ever aboard a plane, you probably have experienced small disturbance (oscillation) during take off and landing. This is called Pilot-induced oscillation, an oscillation caused by the pilot typically occur during high tasking. That kind of PIO is considered mild and is common in aircraft. It sometimes however can develop into catastrophic one, like the one happened in space shuttle enterprise.
It most often caused by the incapability of the actuator. Basically, when pilot gives high input during high tasking, it exceeds the mechanical capability of the actuator. Since it is hard for pilot to recognize, we commonly use neural network to assess the situation.
Yea that is the thing, most engineering problem today are not enough to be solved with analytical solution. They only employ this approach in conceptual and preliminary design to at least depict the general problem.
Moreover, in engineering you're working with human life, so a complex method is a must if it is the only thing that can guarantee safety.
At the end i agree with you that it is all up to the people making it as well as the methodology they want to train the AI with. What i am saying is the potential as well as the hype is real if it is done correctly
If you ever aboard a plane, you probably have experienced small disturbance (oscillation) during take off and landing. This is called Pilot-induced oscillation, an oscillation caused by the pilot typically occur during high tasking. That kind of PIO is considered mild and is common in aircraft. It sometimes however can develop into catastrophic one, like the one happened in space shuttle enterprise.
It most often caused by the incapability of the actuator. Basically, when pilot gives high input during high tasking, it exceeds the mechanical capability of the actuator. Since it is hard for pilot to recognize, we commonly use neural network to assess the situation.
Yea that is the thing, most engineering problem today are not enough to be solved with analytical solution. They only employ this approach in conceptual and preliminary design to at least depict the general problem.
Moreover, in engineering you're working with human life, so a complex method is a must if it is the only thing that can guarantee safety.
At the end i agree with you that it is all up to the people making it as well as the methodology they want to train the AI with. What i am saying is the potential as well as the hype is real if it is done correctly
Sure, you can use least square solutions in AI, but that brings me to my first point where I said AI is nothing new. Why is there suddenly such a big hype over regression analysis or decision trees? These models have been there for ages, they suddenly didnt get a magical improvement to provide you with optimal solutions.
The truly fascinating development that happened in the AI field over the last decade is deep neural networks. People used to think adding more layers to a neural network doesnt provide performance improvement, but Hinton disproved that. As I said, we could reach optimal performance by adding many layers, but that will again introduce complexity.
I see. This sounds similar to the transient behavior of a filter or power amplifier before it reaches the steady state. Those transients are always nearly impossible to model and can vary between devices, i.e. can be device specific.
Of course, there is tremendous potential only if we know when and how to properly use it. It is going to solve some problems that we cant solve right now, but at the same time, it wont be easy to replace an established, well-defined and well-understood analytical solution with AI.
The truly fascinating development that happened in the AI field over the last decade is deep neural networks. People used to think adding more layers to a neural network doesnt provide performance improvement, but Hinton disproved that. As I said, we could reach optimal performance by adding many layers, but that will again introduce complexity.
I see. This sounds similar to the transient behavior of a filter or power amplifier before it reaches the steady state. Those transients are always nearly impossible to model and can vary between devices, i.e. can be device specific.
Of course, there is tremendous potential only if we know when and how to properly use it. It is going to solve some problems that we cant solve right now, but at the same time, it wont be easy to replace an established, well-defined and well-understood analytical solution with AI.
Last edited:
AI will create deep fakes and try to make fake news, AI will be created to detect deep fakes and never be revealed to the public. I built an auto moderator for discord server once. Nothing got pass it, it never slept and understood meme, it could not be tricked, I did this for a kaggle event because I was a nerd.
Man, ChatGPT's a life savior for me. You can use it to practice new things you learn; prepare for exams or interviews; learn complex topics in an easy way; get opinions(to brainstorm); plan your days and do all kinds of productive stuff. My only suggestion to the young ones would be to use it to improve themselves and be more productive.