From: aminer68@gmail.com   
      
   Hello,   
      
      
      
      
   More of my philosophy about the mechanisms of attention and self-attention of   
   Transformers AI models and more of my thoughts..   
      
   I am a white arab from Morocco, and i think i am smart since i have also   
   invented many scalable algorithms and algorithms..   
      
      
   I think i am highly smart since I have passed two certified IQ tests and i   
   have scored "above" 115 IQ,  and i mean that it is "above" 115 IQ, i think i   
   am understanding deep learning, but i say that Transformers are deep learning   
   + self-attention and    
   attention, and this attention and self-attention permit to grasp "context" and   
   "antecedents", for example when you say the following sentence:   
      
      
   "The animal didn't cross the street because it was too tired"    
        
      
   So we can ask how the artificial intelligence of ChatGPT that uses    
   Generative Pre-trained Transformer will understand that the "it" in    
   the above sentence is not the street but the animal, so i say that    
   it is with self-attention and attention mechanisms of artificial intelligence   
   and with the training with more and more data that the transformer  can   
   "detect" the pattern of the "it" refers to the "animal" in the above sentence,   
   so self-attention and    
   attention of the artificial intelligence of ChatGPT that we call Generative   
   Pre-trained Transformer permit to grasp "context" and "antecedents" too, and   
   the "generative" word of the Generative Pre-trained Transformer refers to the   
   model's ability to    
   generate text, and of course we are now noticing that it is making ChatGPT   
   really useful and powerful, and of course i say that ChatGPT will still much   
   more improve , and read my following previous thoughts so that to understand   
   my views about it:    
      
      
   More of my philosophy about transformers and about the next GPT-4 and about   
   ChatGPT and more of my thoughts..   
      
      
      
   The capabilities of transformer architectures, as in GPT of ChatGPT that is   
   called Generative Pre-trained Transformer, are truly remarkable, as they allow   
   machine learning models to surpass human reading comprehension and cognitive   
   abilities in many ways.   
    These models are trained on massive amounts of text data, including entire   
   corpora such as the English Wikipedia or the entire internet, which enables   
   them to become highly advanced language models (LMs) with a deep understanding   
   of language and the    
   ability to perform complex predictive analytics based on text analysis. The   
   result is a model that is able to approximate human-level text cognition, or   
   reading, to an exceptional degree - not just simple comprehension, but also   
   the ability to make    
   sophisticated connections and interpretations about the text, because  the   
   network of Transformer pay “attention” to multiple sentences, enabling it   
   to grasp "context" and "antecedents".  These transformer models represent a   
   significant advancement    
   in the field of natural language processing and have the potential to   
   revolutionize how we interact with and understand language.    
      
   GPT-4 is significantly larger and more powerful than GPT-3, with 170 trillion   
   parameters compared to GPT-3’s 175 billion parameters(and even GPT-3.5 of   
   the new ChatGPT has 175 billion parameters). This allows GPT-4 to process and   
   generate text with    
   greater accuracy and fluency, so with feedback from users and a more powerful   
   GPT-4 model coming up and by being trained on a substantially larger amount of   
   data , ChatGPT that will use GPT-4 may "significantly" improve in the future.   
   So i think ChatGPT    
   will still become much more powerful. And i invite you to read my previous   
   thoughts about my experience with the new ChatGPT:   
      
      
   More of my philosophy about my experience with ChatGPT and about artificial   
   intelligence and more of my thoughts..   
      
      
   I think i am highly smart since I have passed two certified IQ tests and i   
   have scored "above" 115 IQ,  and i mean that it is "above" 115 IQ,    
   so in those two last days i have just tested ChatGPT so that to see if    
   this new artificial intelligence launched by OpenAI in November 2022 is   
   efficient, and i think that it is really useful, since i think by testing it   
   that it can score well on the human average smartness, but if you want it to   
   be highly smart by inventing    
   highly smart things , it will not be able to do it, but if you want ChatGPT to   
   be highly smart on what it has learned from the existing smartness of the   
   human knowledge that it has been trained on, i think it can also score high in   
   many times of it, also    
   ChatGPT can in many times make much less errors than humans, so i think that   
   ChatGPT is really useful, and i think that    
   ChatGPT will be improved much more by increasing the size of    
   its transformer  (A transformer is a deep learning model that adopts the   
   mechanism of self-attention) , and i also think that ChatGPT will be    
   improved much more when it will be trained on a substantially larger amount of   
   data, considering an article that DeepMind just published a few days ago   
   demonstrating that the performance of these models can be drastically improved   
   by scaling data more    
   aggressively than parameters ( Read it here: https://arxiv.org/p   
   f/2203.15556.pdf ), and it is    
   why i am optimistic about the performance of ChatGPT and i think that it will   
   be much more improved.    
      
      
   Thank you,   
   Amine Moulay Ramdane.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)