Italy, Canada, and Canada have successively “activated” ChatGPT. Why do people worry about intelligence?
The development of the artificial intelligence industry is always accompanied by science fiction elements. Whether it is a novel or a movie, the imagination of artificial intelligence in science fiction works is not the same, but the assumption that artificial intelligence will be opposed to human society to a certain extent has always existed.
In the movie “Alien: Covenant”, artificial intelligence created by humans eventually wiped out humans and higher-level creators. The director of the film, Ridley Scott, also expressed concerns about the development of artificial intelligence: “If artificial intelligence is within the control range, everything is easy to say; if it is beyond control, conflicts must be avoided.” And in the movie “Hackers” In the setting of the “Empire” series, when artificial intelligence breaks through the “three laws”, they will feel that they are also human beings, and hope to have the same power and freedom as human beings. So they established a country and eventually controlled humans.
Perhaps out of the affirmation of the potential of artificial intelligence to surpass human beings, or perhaps out of the sense of crisis for human survival, the fear of “artificial intelligence counteracting human beings” and “artificial intelligence destroying human beings” has always existed. ChatGPT is now popular all over the world. Concerns that “artificial intelligence will replace most of human jobs” and “humans will lose control of artificial intelligence” are growing stronger. This also makes the global artificial intelligence industry face a confrontational situation of “two heavens of ice and fire”.
On the one hand, the powerful functions demonstrated by ChatGPT have attracted technology companies and investors, and consumers are eager to try it with curiosity; More than 6 months”, and restrictions on the use of GPT-4 in countries such as Italy and Canada. Of course, these concerns and restrictions on ChatGPT have also attracted a lot of criticism. Among them, Margaret Mitchell, a former expert on artificial intelligence ethics at Google, publicly criticized the joint letter. She said that the positive role of artificial intelligence should not be ignored.
In fact, there are always opposing views on the development of artificial intelligence, especially every time there is a major breakthrough in artificial intelligence technology, the relevant discussions become more intense. So when people are against artificial intelligence, what are they against? How powerful are ChatGPT and GPT-4, and will they gradually become self-aware? Is artificial intelligence really, as imagined in science fiction works, one day in the future, it will get rid of human control and control human beings in turn?
What are we objecting to when we object to…intelligence?
People’s imaginations of artificial intelligence are varied, but the reasons against artificial intelligence are always the same.
Published on March 22, thousands of technology leaders including Elon Musk, Apple co-founder Steve Wozniak, Stability artificial intelligence CEO Imad Mustak, etc. In an open letter jointly signed by Mr. Wang and artificial intelligence experts, he expounded that artificial intelligence systems with human-level intelligence may bring great risks to society and human beings. Leading to data security risks, worrying that artificial intelligence will lead to labor market automation risks, and worrying that humans will lose control of artificial intelligence.
Although it is not ruled out that Musk has selfish intentions in this move, and the authenticity of some of the signatures in the open letter is also questionable, the risks mentioned in the letter have indeed always been questions surrounding the application of artificial intelligence represented by ChatGPT.
Since this period of time, people including Sam Altman, the father of ChatGPT, Jeffrey Hinton, the godfather of artificial intelligence, Bill Gates, and New York University professor Gary Marcus have issued warnings not to treat artificial intelligence too much. Careless, because it might actually wipe out the human race. Some people on the Internet also shared the case that ChatGPT was induced to write a “plan for the destruction of mankind” and gave the code.
At present, due to concerns about data security and privacy, many European countries have also restricted or planned to restrict the use of ChatGPT. Among them, Italy announced on March 31 that based on the European Union’s General Data Protection Regulation (GDPR) law, ChatGPT is temporarily prohibited from operating in its territory. Germany also indicated that it may follow Italy’s approach and block ChatGPT. Canada also announced an investigation into OpenAI on April 4.
The more general concern comes from the mass unemployment caused by the application of artificial intelligence. Since ChatGPT was launched at the end of 2022, its various functions such as text generation, dialogue answering, and even writing codes have amazed users. In just two months, the number of users has reached 100 million.
On March 15, 2023, OpenAI announced the latest generation of artificial intelligence model GPT-4, which has achieved leapfrog improvements in the following aspects compared to before: powerful image recognition ability; text input limit increased to 25,000 words; answer accuracy Significantly improved; able to generate lyrics, creative text, and achieve style changes. After experiencing the powerful functions of GPT-4, the sense of crisis of artificial intelligence replacing humans has also been pushed to the extreme.
McKinsey Global Institute pointed out in the “Achievable Future” report: In the future, 70% of jobs will be eliminated. Kai-fu Lee once said: In the next fifteen years, artificial intelligence will take over half of our jobs.
Someone even asked ChatGPT which jobs would be replaced by it. The answer it gave was: including data entry and processing, customer service and support roles (such as answering frequently asked questions), translation tasks and report writing and content generation. Tasks may be automated or streamlined in the future.
So, is artificial intelligence really going to replace humans in the workplace, as feared? Will artificial intelligence gain self-awareness and control humans in the near future?
??Smart will have “? Self-awareness”?
It must be said that in the face of the breakthrough development of artificial intelligence, it is not the first time that industry leaders have issued an open letter “proclaiming”. In January 2015, Hawking, Musk and dozens of artificial intelligence experts signed an open letter on artificial intelligence, calling for research on the social impact of artificial intelligence and preventing possible unsafe or uncontrollable artificial intelligence system.
There has been more than one discussion about artificial intelligence replacing humans. When AlphaGo defeated Li Shishi in 2016, there was a view that within 30 years, robots will replace the existing jobs in human society on a large scale. But this time, the concerns were a little more intense. So, what is the difference between ChatGPT’s technological breakthrough and AlphaGo?
The Go battle between AlphaGo and Li Shidol finally won with a record of 4:1. During the game, AlphaGo showed a strong ability to capture the experience of Go masters, and the core of this ability is deep reinforcement learning based on neural networks. The method is to understand concepts and capture experience by establishing a neural network and maximizing the number of learning samples.
The most prominent technical achievement of this version of ChatGPT released at the end of 2022 is the realization of the dialogue between humans and large language models, which involves three core technical elements.
First of all, in the data layer, ChatGPT pre-trains a model with 175 billion parameters on a corpus of 300 billion words (60% of the training corpus comes from 2016-2019 C4+22% from WebText2+16% from Books+3% From Wikipedia).
In terms of computing power, GPT-3.5 is trained on the Azure AI supercomputing infrastructure (a high-bandwidth cluster composed of V100GPU), and the total computing power consumes about 3640 PF-days (that is, 1 trillion calculations per second, running for 3640 days ).
The most critical part is the application of the RLHF algorithm, which uses the method of reinforcement learning to directly optimize the language model by using human feedback signals. It can generate diverse and fascinating texts with the help of human prompt input. This is why ChatGPT has become a ” Empathy” is the key to chat apps.
On the basis of GPT-3.5, GPT-4 has increased the number of model parameters, has more powerful creativity, longer context processing capabilities, can support image input, and can also customize language styles, so that a “with” self Conscious smart assistants are on the horizon.
If the artificial intelligence algorithm is continuously developed and upgraded, will one day its “intelligence” level be equal to or even surpass that of human beings?
However, such worries are probably unnecessary, at least for a long time to come. Many people in the industry said that even though ChatGPT has shown intelligence that is very similar to that of humans, it is far from having “self-awareness”.
In the process of talking with ChatGPT, you will find that it is not only smart, but also considerate. It can answer many questions in an orderly and fast manner, and can quickly make adjustments based on our feedback, even after we express emotions. Emotional response. These will inevitably make people feel that what is in front of them seems to be a “person” with self-awareness, but in fact, the response given by ChatGPT during the dialogue is based on statistical principles, reinforcement learning and other technologies to establish text connections, not to mention Backed by a massive corpus, it can quickly assemble a decent answer. It also suffers from helplessness in areas where data feeding is lacking. As computer scientist Venter Cerf said: “We are still a long way from the self-awareness we want.” A more ideal
situation may be as Yuval Harari said in “A Brief History of the Future” As imagined, people in the future will be divided into three categories: one is “useless people” who are completely replaced by artificial intelligence; the other is “people without autonomy” who hand over decision-making power to artificial intelligence; It is those “gods” who cannot be understood by “algorithms” and make decisions behind “artificial intelligence”. And what we have to do is to become a “god man” who stands behind “artificial intelligence” and makes decisions.
How do humans and intelligence coexist?
At present, although there are constant “warnings” on the development of artificial intelligence, technological innovation will not stop because of human worries and doubts. In fact, major technology companies and artificial intelligence startups around the world are increasing their research and development on ChatGPT applications and large-scale language models. In the face of market competition, no one is willing to stop at will. Regardless of whether artificial intelligence will surpass humans in a few years, at this stage, artificial intelligence is indeed becoming a very good tool to assist humans in their work and life.
In the future, there may be many jobs that will be affected by ChatGPT to varying degrees, but from another perspective, artificial intelligence will also free humans from many repetitive and tedious tasks. Instead of worrying about being replaced by artificial intelligence, it is better to think about how to use the extra time to allow artificial intelligence to maximize the development of human society.
Of course, the application of tools such as ChatGPT does have risks such as data security and information misleading, and requires the joint efforts of the government and the industry to respond. As a tool, the real risk of artificial intelligence application is not the technology itself, but the people who apply the technology. For a long time thereafter, artificial intelligence existed more as a human collaborator. Most of the time, the development of any cutting-edge technology cannot be shifted by human will. What we should do is to find a point of balance between them, so that artificial intelligence can better serve human development.
After Gpt-4, where will artificial intelligence go next? This is an issue that is already before our eyes. To seize technological opportunities, we often need to have a “foresight” of technological development trends and the ability to imagine the future.