Seeing the bewildered looks on the faces of the people present, Gao Nian continued, "Remember when we talked about the emergence of artificial intelligence?
When the parameters reach a quantity of 10 billion to 100 billion, artificial intelligence will enter a state of emergence similar to enlightenment.
The result is that the artificial intelligence, originally as stupid as a pig, suddenly learns, just like a baby suddenly being able to express what they want with vague words one day.
And the more parameters there are, the more it knows, and the final result is that it becomes smarter!"
Hearing Gao Nian's words, the people at the scene seemed to understand the meaning of his words.
Based on the minimum of 10 billion parameters required to enter the enlightened state of emergence, the current model has trillion parameters.
That is equivalent to a 180-fold difference in parameters. With a 180-fold difference, how smart is this third-generation artificial intelligence assistant?
Thinking of this question, the people at the scene instantly became excited.
After all, a 10 billion parameter model can enter a state of enlightenment, so with a 180-fold difference in parameters, the third-generation artificial intelligence assistant will definitely be smarter, right?
However, Gao Nian's words did not stop. He continued:
" trillion parameters are not actually the limit of our third-generation artificial intelligence assistant.
Theoretically, the normal number of parameters for our third-generation artificial intelligence assistant should be around 100 trillion.
The reason it can only reach trillion parameters currently is because our third-generation artificial intelligence assistant has too little knowledge and skills.
The current third-generation artificial intelligence assistant only excels in two fields: text and voice chat, and game generation.
Other functions that it should theoretically also be proficient in, such as image and video voice recognition, image generation, video generation, article generation, and programming code,
are being developed one by one due to our limited energy and funds. After all, the funds and energy required to train it are quite astonishing.
You may not know, but in order to train the third-generation artificial intelligence assistant, we used a full 300,000 Phoenix 1090 graphics cards!
It can be said that in order to train the third-generation artificial intelligence assistant and give it the current text and voice chat function and game generation function,
we used 300,000 Phoenix 1090 graphics cards for several months of training."
Hearing Gao Nian's words, the people at the scene were instantly dumbfounded.
Because Gao Nian actually said that in order to train the third-generation artificial intelligence assistant, they used 300,000 Phoenix 1090 graphics cards for several months of training!
If this is true, then the cost of training this third-generation artificial intelligence assistant is too great.
After all, this is 300,000 Phoenix 1090 graphics cards, and the computing power of one Phoenix 1090 graphics card is TFLOPS.
Three hundred thousand Phoenix 1090 graphics cards continuously calculated for several months, and finally allowed the third-generation artificial intelligence assistant to learn text and voice chat and game generation.
This Geek Technology company is playing too big. It's too exaggerated for a third-generation artificial intelligence assistant.
Of course, that's what they said, but at this time, they couldn't help but feel a great sense of curiosity in their hearts.
How powerful is the third-generation artificial intelligence assistant that Geek Technology has spent so much effort to develop and train?
Judging from the fact that Geek Technology dares to convene mainstream media from all over the world to participate in the product launch conference, this third-generation artificial intelligence assistant will definitely not disappoint.
"Next, let's demonstrate the text and voice chat function.
First of all, our third-generation artificial intelligence assistant supports 50 mainstream languages, including Chinese, English, Spanish, French, German, Italian, Japanese, Korean, Dutch, and Russian.
Subsequently, whether it is voice chat or text chat, they are essentially the same.
Voice chat is that it will recognize your voice, then convert it into text to understand what the user is saying,
and finally give a reply in voice form.
So the essence of voice chat is still text chat, but it can save the time of entering text and watching text."
Hearing Gao Nian's words, the people at the scene nodded slightly, and then continued to look at Gao Nian.
"As for how to use it? The way to use it is actually the same as chatting with a friend.
You can freely ask it all kinds of questions and talk to it, and then it will answer you. At the bottom of the answer, it will also attach the source of the information.
This seems similar to the second-generation Geek smart search engine on the surface, but the actual experience is completely different.
Simply put, it's like a 'know-it-all librarian' versus a 'librarian who can only help you find knowledge answers in the library'.
When you asked the previous second-generation artificial intelligence search engine something,
it would actually search for corresponding knowledge answers from the 'vast library' based on its own experience, saving you the time of using the search engine.
If there is no information answer on the Internet, then the second-generation artificial intelligence assistant will not be able to answer, and will directly say no, please search by yourself.
The current third-generation artificial intelligence assistant is completely different. It is actually a 'know-it-all librarian' mode.
Because the operating mode of the third-generation artificial intelligence assistant is to directly instill the books and knowledge of the library into the librarian's mind.
So when we ask questions, we are actually asking the librarian himself, instead of letting it help us search for knowledge answers in the library.
These two seem to be no different on the surface. They both get answers from the librarian, but the actual process and results are completely different.
The so-called 'reading a book a hundred times will reveal its meaning'.
If we read a book a hundred times, we can deeply understand the content and ideas of the book and gain some insights.
After that, the more books we read, the more insights we will gain, and then we will have our own different views on the world and on everything. This is the reason why reading makes people wise.
And this phenomenon of reading making people wise is actually feasible when applied to generative artificial intelligence.
When we let generative artificial intelligence read a book, it is just memorizing it, and it doesn't understand what you are talking about at all. But when we let it memorize another ten thousand books or even millions of books, it reads more books and knows more things, then the 'emergence' state we mentioned earlier will appear.
From then on, when it answers user questions or handles things, it will have logic, as if it can suddenly understand what you are saying."
Hearing Gao Nian's words, the faces of many people who understood artificial intelligence knowledge suddenly changed.
Because listening to Gao Nian's words, how does it feel that the third-generation artificial intelligence developed by Geek Technology seems to have self-awareness?
Because this is how we humans have come to be. Newborn babies don't understand anything.
But through the words and deeds of their parents, babies gradually learn to speak. As the content they learn to speak becomes more and more, their self-awareness eventually becomes stronger and stronger, and they become a 'person'.
Looking at the emergence phenomenon in the third-generation artificial intelligence assistant at this time, how does it feel that it seems to have self-awareness?!
So, in the VIP seats in the front row, Elon Musk, the president of Tesla Motors, suddenly stood up and raised his hand.
"Mr. Musk, do you have anything you want to ask?"
Gao Nian asked Musk, who suddenly stood up and raised his hand to ask a question, with a hint of incomprehension in his eyes, wondering why the other party suddenly interrupted his words.
