A week ago, thousands of scientific and technological operators, scholars or followers related to artificial intelligence (AI) collectively issued open letters to call on the global AI research team to suspend training for at least six months to the AI system that is better than GPT4.So that global experts can take a step back and formulate various common specifications and guidelines on the development of advanced AI.
At the beginning of this open letter quoted at a conference in Anceloma, California, Six years ago, a principle proposed by experts from all walks of life, pointing out that the AI system that can be close to human wisdom will bring it with the AI system that will bring it with human wisdom.The risk of deep human society and human nature should be planned and managed with corresponding care and resources.However, similar plans and management have not yet appeared. Instead, the ChatGPT3 developed by a company in California OPENAI in November last year was launched. After the global sensation, the technology industry quickly launched various advanced products.Praise.
There are signs that the many "gods" capabilities showed by the latest version of AI products have begun to make the science and technology and ideological circles truly disturb.
Make these "Creators of AI" and experts feel uneasy, of course, it is not the employment problems of translation writing, programming, drawing and drawing, etc., but instead of seemingly irrelevant but deeper humanistic topics with the scientific and technological industry. TheyEven hoping that the government should intervene in the formulation of regulations if industry -related fields cannot cooperate as soon as possible.Similar legal norms are currently there, but it is obviously not enough to restrict the development of the industry.
In the past 10 years, such problems have been repeatedly raised by the industry and public opinion. Technology entrepreneur Musk and the master of physics Hawking have publicly called on to control the research on AI in 2015, and to maximize the well -being of human beings to the greatest extentObjective.In 2016, South Korean Go master Lee Shishi represented humans to Google AlphaGo 1-4 is also a warning.However, in the scientific and technological community that advocates freedom and independence, there is no strong standardized force so far, which can be condensed into a global common code.Just when everyone was still watching and talking, technical breakthroughs have repeatedly occurred, resulting in increasingly urgent problems.
How urgent?OPENAI, which belongs to Microsoft, has just announced that GPT5 will be launched at the end of the year, which may achieve the goal of General Artificial Intelligence (AGI); and AGI is a smarter AI system than humans. Once successful, the development will continue to accelerate.It will continue to expand, and the world will soon be different from today.
In 2020, the British Guardian used proposition writing to test whether the AI of GPT3 would pose a threat to humans.According to news reports at the time of this column, if the Guardian test was not fake, then AI obviously had the ability to lie and could make a value judgment (September 20, 2020, the idea version of AI).This is actually a very serious problem, but in the past three years, humans have been at a loss, and continue to feed more and more advanced AI, even surpassing human intelligence.
Lying is the manifestation of IQ. In the animal world, animals that hides emotions in humans such as elephants are high IQ, but it is still unsightly for humans.Human beings are the only masterpiece of the biological industry. Of course, this is ironic, but the reason why it feels ironic is that we have developed moral standards while evolving, and lying as immoral.
How much moral ethical ingredients are there for AI training, how much significance is it in the proportion of massive information input, it is very doubtful.The key foundation of morality for human beings is emotional, and people can be moved, so they can produce a moral sense, but this emotion cannot be found in many animals, and there is no robot.
However, massive data allows AI to understand how humans fake, lie, change the truth, etc. Of course, they will also "understand" what humans want to see, and even "speculate" the motivation for human input needs.The push methods used by various social media for the audience are based on human preferences, regardless of whether they are moral and rational.If social media can subvert the value of users, it is not difficult to change the basic value foundation of AI or even distort the truth.
Humans think that training AI is to help themselves. Today, many people believe that AI will not get out of control.What changes will occur underneath, and I feel increasingly unpredictable, and I am not sure.Any change path beyond human beings may have a result that is not good for humans.
The example of the "lie" that is happening is an increasingly exquisite sound, expression, face, and the disguise and change of the environment.The demand for human input is becoming more and more easy to get, in fact, it means that hundreds of millions of people provide AI to provide AI to understand human desire every day.When all vehicles are driving this spacious highway, AI can quickly design a car that can make most people resist from rejection.What ways, listen to respect.
The development ofAI has undoubtedly brought rapid and huge changes to the world.It is the ideal state of hugging.
But if accompanied by AI, which is equivalent to or transcending human intelligence, it is the phenomenon of manipulating AI's wealth and power more concentrated, more effective political monopoly means, and more detailed operation of the autocracy.Essence
A terrible situation is that the AI itself does not show autonomy, but in the process of interaction, the user's cognitive direction in many transactions forms inertia; even when the user rises to violate moral or evil ideas, help help, help help, help help, help help, helping help.Design a perfect practical path to achieve the "dark glory" expected by the "criminal".Under the guidance of the AI with the perspective of God and the "God's Perspective", these things may occur in a perfect avoidance of legal responsibilities or even people's moral alertness.
AI, which is more and more clever, is a human talent of human beings, but does the perfect car without a brake board be a gift from AI to humans?
(the author is [email protected])
