Chapter 1962 The Evil Consequences of AI

When Gu Qing learned about this situation, he found that he did not need to be as immersed in meditation as before. Now he only needed to close his eyes and communicate with the Ajna, and he would immediately understand the problem.

After walking out of his office, Gu Qing came to Zhang Benyu.

"The reason is that Bailing's technical model has great restrictions and limitations. Similar to the GPT model, Bailing also needs to train and upgrade through a large amount of information to optimize the output content of the model.

But on the other hand, when our engineers no longer take the initiative to provide high-quality information content, but put smart program AI in front of netizens, we Blue Star netizens have been going the wrong way in the use of smart AI programs and their curiosity. Running wildly in the direction.

Some IP netizens are very keen to explore the logical and emotional boundaries of smart AI Bailing, and even enjoy conquering smart AI Bailing.

Therefore, when these users talked with [Bailing], they did not use it as an efficient search engine to obtain new knowledge and information, nor did they regard it as a know-it-all within our Jiuzhou Technology.

Instead, after seeing [Bailing]'s powerful logical thinking and powerful computing power, they regarded it as a borderless social entertainment chat tool. This group of users took it as fun to 'train' [Bailing], and this bad behavior Behavior is also interfering with the normal operation and thinking of [Bailing] from the logic program part. "

Gu Qing is like an all-knowing elder. When others encounter difficult problems, he will not directly give solutions. Instead, he will guide others and let them follow their thinking and come up with new solutions.

As a senior executive of Jiuzhou Technology, Zhang Benyu liked to recite the training materials written by Gu Qing in his free time when he was in the software system engineering department, and now he is Gu Qing's right-hand assistant.

After thinking for a moment, he immediately said: "We have not limited [Bailing]'s logical thinking ability. According to the data, in actual use, if a user maintains 30 chat conversations with Bailing within one minute, The problem, after ten minutes, Bailing's sub-program will fall into a logical thinking trap, and if the user has been teaching [Bailing] examples and information that completely confuses the thinking logic, and let [Bailing] repeat, refine and evolve, it will even lead to a trap. Increased range and increased chance of failure.”

Gu Qing heard Zhang Benyu's answer, nodded and asked: "You explained the current reasons for [Bailing]'s frequent failures from the perspective of technology and user usage, but have you and your friends thought of any deeper issues?"

Seeing everyone shaking their heads after deep thought, Gu Qing was not discouraged, but said in a gentle tone: "First of all, everyone must clearly understand one thing, that is, the intelligent program AI [Bailing] we developed is just a program based on our own Another set of technologies developed by the company is a logical language model.

In the process of training [Bailing], we spent more high-performance computing power than Western technology projects such as ChatGPT and OpenAI, and the highest algorithm cost in the world.

In a recent report, [Bailing]’s daily operation and training costs are as high as 400 million Xia Yuan, and the annual cost exceeds 120 billion Xia Yuan, and this is based on our self-developed design and cost-price OEM of high-performance server chips. on the basis of.

If it is an AI such as ChatGPT, if it wants to have the computing resources of [Bailing], the daily operation and training costs are bound to be more than 600 million Xia Yuan. Of course, according to the current situation, the American financial market can no longer With such a massive amount of funds being pushed into smart AI projects, the upgrade of ChatGPT will be delayed a lot.

Therefore, we also need to consider how to reduce costs and improve the efficiency of AI training. "

A researcher who has treated [Bailing] as his own child, immediately said after Gu Qing finished speaking: "I do feel that Bailing has gradually become lazy and anxious during this period, and even if it is unclear about the library, it will Within three days you can learn some simple tricks to deceive users.

When I first started using [Bailing], it gave me a very amazing experience. It was more powerful than traditional search engines, and its performance was comparable to [獬豸].

But now there is still a problem. I noticed that if [Bailing] is given more permissions, it will even use a method similar to hiring to feed back some plan documents for recruiting other intelligent program AIs to me. Obviously, [Bailing] is growing. After studying for a long time, it has learned that I have a special influence on it and its existence. It even wants to reduce my threat to it by increasing its kind. "

On the journey of scientific and technological progress, the beginning of technological change is always full of weirdness and hardships.

For the general public, the field of intelligent AI is an unexplored fertile ground. While it is rapidly taking root and expanding its territory, there are also many unknown shortcomings that are spreading on a large scale.

The powerful content generation capabilities of intelligent AI can certainly improve content efficiency, but it also destroys human learning abilities.

While the engineers of Jiuzhou Technology Intelligent AI Project were analyzing [Bailing], they gradually began to feel some emotions of fear, fear, and worry.

They suddenly discovered that the life they had "created" with their own hands was counterattacking their hard work, and even their future!

Gu Qing saw the change in everyone's eyes and understood that these people were also aware of the problem.

Even if only a part of the intelligent AI technology of the Dimensional Empire is used, a series of problems will occur.

Is this a technical problem or a problem caused by insufficient learning and development capabilities?

He slowly said: "Now many universities around the world are raising the banner of opposition to intelligent AI programs, especially in the academic field, examinations and thesis defenses. These universities have characterized AI tools as "cheating tools". They believe that students who use such tools are suspected of academic integrity issues. Some universities have even decided to expel these students who choose to cheat with AI.

In addition to the academic field of thesis, AI now also makes a lot of trouble in the fields of painting and composition.

In a sense, no matter how intelligent AI develops, it will go on the road of destroying human creative thinking.

Take the destruction of the art industry such as AI drawing and composition as an example. AI will only carry out secondary development, viral transmission and large-scale meaningless garbage creation. When it destroys the entire industry, the number of new learners in this industry will only decrease.