Vol 2 Chapter 1: source

　　 Chapter 101 Source
When the sun rose eastward and the sun was spreading the earth, Xiang Kun finally stopped making changes to the PRD. Looking at the plan that was completely different from the original plan, he couldn't help rubbing his eyebrows, as if he was guilty of old problems. Up...
　　 In the end, the AI ​​assistant he envisioned is already a large-scale project from the perspective of numerous functional modules and complexity, and some functions are not yet mature and can be used for reference. Even for giant companies such as Baidu, Google, and Microsoft, it is not an easy task, let alone him alone.
　　 If he got such a demand document while still working more than a month ago, he would pia back to the product manager's face 80% of the time and ask him to rewrite it after sober.
　　 But now this document was written by himself, and this requirement was made by himself...
　　 So he was a little unwilling to give up, he wanted to try it out to see how far he could do it.
　　 But when he restarted to do what he was really good at, he found that he felt more directly the changes in the way his brain thinks and the speed of calculation, and his subjective excitement became higher.
　　 The current AI assistants are actually very smart. For example, if you ask "How is the weather tomorrow?"
　　 AI first converts your voice into text through speech recognition technology, then analyzes your intentions, extracts keywords to "fill in the slot".
There are two "slots" in 　　, one is the time "tomorrow" and the other is the event "weather".
　　 Then execute the command, call the weather forecast program, find the weather tomorrow, and feedback the result.
　　 If some AIs need to continue the next conversation, they will even "talk more": "Are you going to travel tomorrow?"
　　 If you get a new keyword, it will start a new round of slot filling to assist users in processing various affairs.
In the final analysis, it is relying on each keyword to trigger the preset feedback, but the current AI is constantly changing, the success rate of recognizing the true intention is getting higher and higher, and the feedback is becoming more and more comprehensive and humane. It looks more and more "live".
　　 But in fact, the current AI is still far from the autonomous artificial intelligence.
Xiang Kun also did not have the extravagant hope to create an autonomously conscious artificial intelligence on his own. He just wanted to use "reinforcement learning" combined with "deep learning" to allow this AI assistant to continuously help him improve the "vampire mutation". Model", and according to the output of the model, continuously targeted searches for medicine, biology, physics and even news on the Internet, selecting content that may be useful to him, sorting it out for him to browse, and then according to his actual choices , Improve future search rules and decision-making.
　　 In life and training, the AI ​​assistant will also advise him through the "vampire mutation model", and spontaneously raise some questions that he may overlook, propose new directions that can be tried, and assist him in experiments and parameter recording.
　　 And he also hopes that the AI ​​assistant can help him analyze and monitor daily life behaviors and online operations, whether there are abnormal details that will be detected by people, and propose solutions to help him keep the secret of the mutation.
　　 After spending a whole day, Xiang Kun already has a general implementation framework, and also marked the implementation of most of the functions, as well as a large number of technical difficulties that need to be resolved.
　　 But the deeper it goes, the clearer Xiang Kun knows that this project cannot be completed by himself in a short time.
　　 Although many functions are now available on the Internet with open source code and data sets available, it still takes a lot of time to integrate. Besides, there are many functions that must be developed from scratch.
　　 He is naturally much better than his previous self, and even said that one can top 10 uses without much problem, and he can work 24 hours a day.
But many constraints still cannot be solved by him now. Not to mention manpower, his current hardware facilities cannot support the planned development. Not only do computers need to be replaced, but also cloud services with larger bandwidth and higher computing power are needed. It even needs to involve some terminal equipment customization.
　　 "It seems that you can either find a shortcut or become very rich, just like Tony Stark or Bruce Wayne..." Xiang Kun touched his bald head and muttered to himself.
However, he did not intend to go back to the original needs directly, and use it as a very crude version of the AI ​​assistant. It is like having a table full of Han Han, seeing the cold hard steamed buns in front of him, unless he is too hungry, otherwise I can't speak.
　　He is going to start with the speech synthesis function. Anyway, no matter what kind of AI assistant is in the end, this function is a must.
　　 And just like playing a game, players always like to pinch their faces before assigning specific attribute points. For an AI assistant without a physical body, the voice can be said to be her face.
　　 The function of speech synthesis is now relatively mature. In addition to those major manufacturers, there are also many open source toolkits online that can be used directly.
　　 However, Xiang Kun has committed OCD again. One is still dissatisfied with the smoothness of the voice, and another is dissatisfied with the timbre of the existing female voice.
　　……
　　 At 11 o'clock in the morning on September 13, Xiang Kun received the express from JD.com, so he temporarily stopped his entanglement with synthesized speech and unpacked the package expectantly.
　　 The magnetic field detector and the microscope were sent together. After confirming that there was no problem or damage, Xiang Kun prepared to start the test with the magnetic field detector.
　　 This detector can measure low frequency 5Hz~400Khz, high frequency 300MHz~3000MHz, UU reading www.uukanshu.com and can switch the magnetic field test mode, the accuracy is 1v/m and 0.01μ.
　　 He first took the handheld tester and approached various electrical appliances to confirm that its readings were normal and functioning normally.
　　 Then he took the tester and sat in front of the canary's cage. Since opening the cage the day before yesterday, it hasn't come out yet, and he still stays in his own territory very "safely".
　　 Looking at the canary that tweeted twice, Xiang Kun thought for a while, got up and went to the kitchen, and experimented with the remaining two rabbits.
　　 But what makes him more regretful is that in the process that he used "shock" to cause various effects on the rabbit, the reading of the detector did not change significantly.
　　 "Is my previous guess wrong? Or the detector I bought is too cheap and not sensitive enough?" Xiang Kun couldn't help muttering a few words, then put the detector aside and began to tinker with the microscope.
　　 He didn't test his own cells or blood immediately. Instead, he took a bit of rabbit blood with a lancet. He didn't make a smear. He simply dropped the saline solution to complete the push and observe.
　　 The eyepiece of the microscope he bought is connected to a display screen, which can be directly connected to the computer screen, so he directly uses the computer screen to observe.
　　 There is nothing special about rabbit blood, just to familiarize yourself with a microscope that has not been operated since middle school, and to make a simple reference, so he didn't do much observation.
　　 Next, he found several microscopic pictures of human blood on the Internet, and watched videos of other people using microscopes to observe human blood, as a reference.
　　 Finally, he took his own blood and began to observe.
　　 Xiangkun didn’t make a smear because he wanted to observe his blood more naturally. He knew that within ten minutes, his blood would change drastically.