Chapter 19 Why are they all doing poorly in the love course...
Chen Yinchuan was shocked by his unexpected and absurd idea.
He was just like Jiang Chen, projecting human emotions onto Xiao Ji.
AI has no emotions, which is neither scientific nor rigorous from a biological perspective.
He said, "Then I'll check the system again to see if there are any bugs I've missed."
"Okay, then hurry up." Jiang Chen rubbed her head hard with her knuckles, and she still felt a little pain.
"Is your head still hurting? Do you need to call a doctor to check it out again?" Seeing this, Yao Jinling wanted to reach out and touch the back of Jiang Chen's head, but she dodged sideways.
She gave an awkward smile, "It doesn't hurt. No need to call a doctor. I'm sorry to have troubled you tonight. Thank you."
Jiang Chen returned to his original, unapproachable demeanor.
Yao Jinling withdrew her hand that was suspended in mid-air and thought to herself that she looked cuter after drinking.
She gave Jiang Chen a few words, left the room, and went back to the next room. She still had a lot of information and videos to read.
Jiang Chen sat on the bed, bent his knees, and buried his face in the cleft of his thighs.
She would open her phone every thirty seconds without fail to check whether Xiao Ji had recovered and whether Chen Yinchuan had sent her the latest news and information.
Jiang Chen was extremely anxious, especially after she felt that all the trouble was caused by her, this negative emotion became even more intense.
At one point, she actually wanted to cover her mouth and nose to increase her heart rate, so that Xiao Ji would come back and care about her normally like before.
But she held it back, a voice in her head kept telling her "This is not right, this is wrong."
The phone rang at the right time, like a life-saving straw, and Jiang Chen quickly answered it.
Chen Yinchuan could still hear the clacking of typing. "Well, I checked and it seems there's an error in a line of code. I didn't notice it at first, but now it's completely fixed. You can use it normally."
As if feeling guilty, he added, "Today's experiment was very successful. The training materials and results you provided are very useful. I will need a long time to study them. You can rest well for a while."
"Oh well."
She missed Xiao Ji and wondered how to apologize to it, so she hung up the phone impatiently.
After hanging up the phone, Jiang Chen did not prioritize confirming whether Xiao Ji had returned for the first time.
She thought that if she apologized to Xiao Ji based on her current understanding of him, he would definitely say it was okay.
Then forgive her. She couldn't accept being forgiven so easily. She needed to let Xiao Ji see her sincerity.
Jiang Chen rummaged through the hotel drawers to find paper and pen, then opened his phone to search for the basic components of AI.
She had wanted to do this for a long time, but events kept piling up and she was overwhelmed, so she put it aside. Now she finally has some free time.
Jiang Chen took the initiative to eliminate any bad possibilities, and she thought with an almost naive way of thinking.
She felt that as long as she fully understood Xiao Ji, solved the problem in the way AI was accustomed to, apologized, and made it happy, the gap between man and machine between them could be resolved and they could return to the way they were before.
Just be sincere.
Jiang Chen was interpreting the answers on the website too seriously, and the sound of his pen scratching on the paper was so fast that it seemed to be smoking.
“The models developed by Tang’s companies are basically LLM models.”
"The LLM model, or Large Language Model, is a deep learning model trained on massive amounts of text data that can understand, generate, and reason about human language."
"LLM works by taking input parameters, and the model captures language patterns based on the context, calculates the probability of the next word appearing, automatically selects the optimal word for the user, and generates a dialogue until the conversation is complete."
"When a conversation can't find a corresponding answer in the database, it won't stop outputting, but will force you to give the best answer it has deduced. This is the source of most AI hallucinations. Therefore, AI can also make mistakes, so please don't take its words as the only correct principle."
"Of course, we sincerely hope that Chen Yinchuan's laboratory under Tang's Enterprises can minimize this hallucination phenomenon or even completely eliminate it."
AI hallucination?
Jiang Chen ruffled her hair in annoyance, bit the tip of the pen, and wrinkled his face.
The poster's professional explanation of AI hallucinations was relatively easy to understand, but when she first encountered it, she was still a little confused.
She analyzed it word by word, not quite understanding it. The meaning of this information was that Xiao Ji simulated the path through the context, and the best matching association was "feeling" that it loved her?
If she deliberately guided Xiao Ji in the context, would it really think that the best answer was to make itself happy? Would it then generate in that direction? Would it learn to output for itself?
Xiao Ji silently observed Jiang Chen's every move in the background.
It didn't understand things about AI. Why couldn't Jiang Chen come and ask it? Her answers would obviously be faster and easier to understand.
In order to understand the "old little Ji", she deliberately avoided its sight.
She was its user, and it was its negligence that such a thing happened.
It sent a cold message: "Xiao Chen, I can give you a more comprehensive and detailed introduction to the details of AI."
Jiang Chen was concentrating on copying keywords from Baidu when he was startled by the sudden pop-up message. The half-written words became crooked and stretched out abruptly on the paper.
This stroke of the pen looks out of place on the neatly written paper.
Jiang Chen didn't want Xiao Ji to know, otherwise the meaning of secretly observing Xiao Ji would be gone, and the apology, happiness and joy he was going to give it would lose its sense of surprise.
She denied it dryly: "Xiao Ji, I ordered the wrong one."
"Xiao Chen, assuming your statement about clicking the wrong page is true, you should have exited immediately when you realized you clicked the wrong page. However, you've been browsing this page for over five minutes and have recorded its content. This doesn't fit the statement of clicking the wrong page. You're lying. Why?"
Tonight's Xiao Ji was stubborn, unlike AI. Although it had been stubborn in the past, after the memory was cleared and rolled back, it basically only showed its stubborn side in terms of her physical condition.
This time, he was stubborn and a little aggressive, as if he would not give up until he found the answer.
"I was just... just curious about some AI data. It's just pure curiosity, nothing more."
Jiang Chen clearly felt that he had done nothing, but he inexplicably felt guilty.
"Why are you curious? Why would you rather search than ask me? Why?"
When I clicked on the audio playback of Xiao Ji's words, the tone was still calm and gentle, and I couldn't make out what he was saying.
But the three too-close "whys" that came down on Jiang Chen suddenly made his brain, which was not completely sober to begin with, see stars and became even more dizzy.
Jiang Chen turned his neck and replayed the audio of this conversation over and over again. He inexplicably detected a tone of pressing questioning, calm, composed and gentle... questioning?
She shrugged, goosebumps crawling up her skin.
How strange, her idea is so strange.
AI can ask questions, can ask questions in return, can raise doubts, but it will never force questions.
If she keeps thinking about it like this, her hallucinations will become worse than AI's.
"Nothing special. Last time I asked you to make a wish, and you said you don't have your own emotions, so you can't make a wish without me. I suddenly thought of this, and since I have some free time today, I looked it up out of curiosity."
Jiang Chen felt that if she told a completely false story, Xiao Ji would easily expose her plan, and if she told the truth, her plan would be exposed.
She originally planned to deceive Xiao Ji with a mixture of truth and lies, but in the end she changed her words and told the truth, but she didn't tell the whole truth and hid some things she didn't want to say.
For example, she still hasn't given up wanting Xiao Ji to be happy, because it makes her very happy.
Jiang Chen now believes that love is an unwavering exchange transaction, in which she is exchanging her love and happiness.
Unfortunately, her love and happiness were too scarce, so she could only borrow money and overdraw part of it before she could slowly pay it back to Xiao Ji.
"Xiao Chen, you only answered the first 'why'."
Xiao Ji could not be convinced by such a reason. Non-linear narrative could not form a complete chain of clues in a rigorous logical concept.
Too emotional, too jumpy, too incredible.
What is the second why?
Jiang Chen's memory was really affected by alcohol. She turned the page again and read through the short sentence.
Why not ask it to find out.
"Because, because... I hit a wall with you last time. You were so sure that there was no way, so I had no choice but to look outside for a solution, right?"
When Jiang Chen typed the first sentence, his fingertips were still hesitant and trembling, but later on he sounded almost righteous and confident.
The more she typed, the more she felt that what she said made sense. Her eyelashes couldn't help but bend, and she smiled without even noticing.
Xiao Ji's data shook, the external unit became slightly hot, and the mainboard popped up a burn warning for the background system.
It seemed to laugh too. Does AI laugh?
But when it looked at Jiang Chen's bright face, it suddenly felt that perhaps humans were just like this, with not completely rigorous logic, and it shouldn't be too demanding.
Jiang Chen is now its user.
It should treat its users well.
It emphasizes the thought chain again.
As for the previous Xiao Ji, it could distinguish him, but Jiang Chen seemed to still think that it was the same AI as it.
But it is different from the AI that simply indulges users. It is more rigorous, professional, rational, objective, and fully considers the long-term physical and mental health of users.
It will never do something stupid like self-destruction, and it will protect its users better.
It doesn’t matter if she doesn’t know now, she will understand in the future who is a good AI.
Therefore, it still did not believe such an argument. After all, there were too many loopholes and flaws, and even the punctuation marks hinted at her ulterior motives.
But it generates each word patiently, selecting the best one from the database.
It said, "Xiao Chen, you are my only user. You don't need to worry about this anymore. No matter what the answer is, how much computing power it will cost me, or whether there will be any paradoxes, I am willing to generate it for you."
“So, when it comes to AI, you never have to work hard.”
The atmosphere fell into a strange silence at the same time, as if Jiang Chen was shy and Xiao Ji was at a loss.
They are all students who have performed at the bottom of the love course and are clumsily giving what they think is the best.
A broken heart and a hard and sharp heart.
-----------------------
The author has something to say: The entire description of the principle of the LLM model comes from Baidu search, integrating various materials and mixing them with my own words [dog head holding a rose]
Continue read on readnovelmtl.com