The emergence of large language models has pushed us to a new technological outlet, so under this outlet, how can we look at the diversity and complexity of AI technology? In this article, the author starts from the topic of "the contrast and integration of anthropomorphism and non-anthropomorphism", and talks about his views, let's take a look.
Imagine if we could create an intelligent creature that has both the flexibility of human thinking and the ability to process information beyond humans – that's the beauty of AI technology. From the original dream, the journey of exploration of artificial intelligence began with a simple but grand idea: to imitate human intelligence. Over the decades, scientists have been like artists, using human thinking patterns and behavioral characteristics as a model to gradually sketch out the prototype of intelligent machines.
With the rapid advancement of technology, this dream is no longer out of reach. Today, we are standing on the cusp of a new technology - the era of large language models. These models, like magical giants, not only replicate human language abilities, but in many ways surpass the limits of human thinking. They are like characters out of science fiction, filling us with infinite imagination and anticipation for the future.
1. The two-sided myth of artificial intelligence - OpenAI and Character.ai
Since ChatGPT detonated the Internet in November last year, the two AI companies have been in the limelight. The first is OpenAI, the force behind ChatGPT, whose impressive achievements in the field of non-anthropomorphic large language models are closely followed by a fascinating Character.AI with its unique anthropomorphic character model.
Between September 2022 and August 2023, ChatGPT led the way with a staggering 14.6 billion visits, while Character.AI followed with 3.8 billion visits, demonstrating their significant influence in the field of artificial intelligence, according to the ranking of the 50 most visited AI tools.
There are many reasons why OpenAI and Character.AI have achieved such outstanding results, and the development path and major breakthroughs of large language model technology are even more eye-catching.
In particular, I would like to explore a topic that I have been thinking about for a long time—the contrast and fusion of anthropomorphism and non-anthropomorphism. Anthropomorphic technologies, such as those used by Character.ai, aim to create more vivid and relatable AI characters, while non-anthropomorphic technologies, such as OpenAI, emphasize the efficiency and widespread use of information processing. The two different paths reveal the diversity and complexity of AI technology, and also indicate the future development direction of artificial intelligence.
2. The two flavors of AI - anthropomorphic and non-anthropomorphic
In the world of anthropomorphic technology, AI is not only a tool for information processing, but also a partner with personality and emotion. Imagine how exciting it would be if your digital assistant could not only answer your questions, but also understand your emotions and even provide comfort and advice when necessary. It's like Jarvis in the movie "Iron Man", not only is he an efficient information processor, but also a caring companion to the owner, capable of having natural and intelligent conversations, and even showing a sense of humor and emotion.
In the real world, anthropomorphic AI platforms such as Character.AI are trying to make this vision a reality. They are designed with AI characters that mimic the way humans have conversation, providing a more natural and intimate interactive experience. This technology has shown great potential in a number of ways: in customer service, it can provide more personalized and emotional responses and improve customer satisfaction, in education, it can interact with students in a more friendly and engaging way, and even in medical and mental health, it can provide support and companionship to help people find solace in difficult times.
It shows that the market has great interest and expectations for anthropomorphic AI. After all, who can resist a chat partner who is always listening, caressing, and even providing wisdom and insight anytime, anywhere?
These AI characters can not only conduct ordinary daily conversations, but also play roles with different identities and personalities according to their personal preferences, adjust their chat styles according to user requirements, and communicate freely in various scenes and plots. They are like the friends who never get bored and are always there for you. Whether you're looking for relaxation between busy work hours or someone to talk to in the middle of the night, these AI characters are always the first to respond to the user's request for companionship, and make the user feel comfortable and happy in the way they want it to be.
Despite its remarkable achievements in mimicking human communication, anthropomorphic AI technology still faces challenges when dealing with some of the more complex and abstract problems. It's like a world traveler who is fluent in multiple Chinese, and although he can communicate freely in different countries, he may not be able to cope when it comes to esoteric philosophy or advanced mathematical theories.
In the development of large language model technology, we can see that many of the capabilities that are very powerful on the basic model are greatly diminished after the alignment with human values is completed.
This phenomenon is not difficult to understand, after all, AI has advantages that are difficult for humans to match in terms of information acquisition, knowledge extraction, and data updating.
An example I often cite in my work is that a person may hear about 1 billion words in their lifetime, while large language models can be trained on trillions of words, equivalent to the content of 2.5 million books. According to statistics, the average Chinese adult reads about 4.7 books per year, which means that an average person would need more than 530,000 years to read the same number of books. Considering that books are only 3,000 to 4,000 years old, and words are only about 5,000 years old, this comparison shows the astonishing information processing power of AI.
Another example is AlphaGo, a Go AI that has attracted widespread attention from the society. Human beings play chess with a very strong path dependence and a limited life, so the safest way to become a master is to study the endgame of their predecessors, rather than groping for it themselves. With its powerful computing power, AlphaGo Zero has explored a large number of ways that humans have never explored in a short period of time.
On May 28, 2017, chess player Ke Jie defeated AlphaGo 0:3, completely shattering the dignity of human beings in this complex game. After the game, the talented teenager choked up for a while, and in an interview, Ke Jie lamented that AlphaGo was too perfect to see any hope of victory. With tears in his eyes, he said: "We humans have played Go for 2,000 years, and we haven't even entered the door. ”
Chinese chess saint Nie Weiping even called AlphaGo "A Teacher", he said: "I am fascinated by AlphaGo's moves, how profound and mysterious Go is. The order and timing of AlphaGo's moves are very well controlled. Its level completely surpasses that of human beings, and if you challenge it to play chess, you can only find death. We should let Mr. A teach us how to play chess. He also said: "Teacher A is at least 20 dans, and he is simply a Go god." ”
Allowing AI to imitate human thinking and language Xi may limit human thinking like ants, which will greatly limit AI's capabilities. OpenAI insists on not making efforts in the field of anthropomorphism, which may be a consideration.
3. Black box and bridge - the front-end and back-end of AI
So, how do you balance the human demand for anthropomorphism with the potential limitations of anthropomorphism on AI capabilities? As an IT practitioner, I can't help but think about the relationship between front-end and back-end technologies in software development. In traditional Internet application development, the "front-end" usually refers to the part that the user can interact with directly, such as the interface of a web page or mobile application. The main task of front-end technology is to provide an intuitive, user-friendly user experience that makes the presentation and interaction of information as simple and enjoyable as possible.
In contrast, the "backend" refers to the application running on the server, which is responsible for handling core functions such as database access and application logic execution. Back-end technologies are more focused on things like data processing, storage, and security, and are often invisible to users. It is the backbone of the entire system, ensuring the accuracy and efficiency of the data.
The combination of the two provides us with a unique perspective on the convergence of anthropomorphic and non-anthropomorphic AI technologies. In the field of AI, "front-end" anthropomorphic technology is like the user interface of an Internet application, providing users with a natural and intimate interactive experience. The "back-end" non-anthropomorphic technology is similar to the back-end data processing system, silently supporting the intelligent processing and decision-making capabilities of the entire system behind the scenes.
By understanding these two concepts, when discussing the front-end and back-end of AI, we can compare them to user interfaces and server-side programs in software development. This analogy not only helps us understand the different functions and roles of AI more clearly, but also inspires us to think about how to better integrate these two technologies to jointly promote the development and application of AI technology.
1. Front-end anthropomorphic technology: an unprecedented interactive experience
In the future, anthropomorphic AI technology will play a pivotal role in our lives. In the future, AI will not only be our digital assistant, but also a friend with emotions and a steward of life. They learn through our daily interactions with our daily Xi, gradually understanding our preferences, interests, and even mood swings. This deep understanding and emotional bond allows AI to play an unprecedented role in the business world.
Imagine an AI that knows your preferences and your life Xi and can recommend music, movies, books, and even travel destinations for you based on your mood and needs. This kind of personalized recommendation goes far beyond the capabilities of existing search engines and e-commerce platforms. Because they are built on a foundation of deep understanding and emotional connection, these recommendations are more likely to match the real needs and preferences of users.
In addition, this emotional connection can also play a huge role in the field of marketing and advertising. Different from traditional advertising, AI-based emotional recommendations can more accurately match users' interests and purchase intentions, thereby improving the conversion rate of ads. Users are more likely to receive product recommendations from a "trusted friend" than from an indiscriminate advertising bombardment.
In service industries such as hospitality, airlines, or retail, anthropomorphic AI assistants can act as managers of customer relationships, providing personalized services and enhancing customer loyalty. For example, in hotel accommodations, AI can adjust the temperature, light, and even welcome gifts of the room based on the customer's past preferences, providing a unique, customized stay experience.
All in all, the AI of the future will win our friendship and trust while understanding us. This emotion-based connection not only makes AI an integral part of our lives, but also opens up new possibilities for the business world, making it an efficient, personalized, and emotive source of information and a medium for product recommendations.
2. Back-end non-anthropomorphic technology: a more mysterious black box
While front-end anthropomorphic AI provides us with intimate and personalized partnerships, back-end non-anthropomorphic technology plays a more mysterious and powerful role. These technologies act like brains behind the scenes, focused on complex computational, data analysis, and logical reasoning tasks that are far beyond human comprehension.
At its core, non-anthropomorphic technology lies in its agnosticism and powerful intrinsic capabilities. They can quickly process and analyze huge data sets, revealing patterns and trends from vast amounts of information to provide amazing predictions and solutions. Predictably, this ability must be completely beyond human intuition and cognition.
These technologies are no longer just tools for data analysis and logical reasoning, they will become the key to revealing the mysteries of the universe and solving the puzzles of human history.
Imagine an AI system capable of deciphering complex galaxy movements, predicting future climate change, and even revealing the origin of life. These back-end non-anthropomorphic technologies, with their incredible computing power and data processing capabilities, will be able to explore areas that have not yet been touched or understood by human scientists. They will be able to uncover patterns and patterns hidden in subtleties in a vast ocean of data, leading to a series of jaw-dropping discoveries.
In the medical field, these technologies may reveal the root cause of disease, develop revolutionary treatments, and even find key clues to prolonging human lifespan. In environmental science, they are able to simulate the Earth's complex ecosystems, predict and respond to natural disasters, and even find ways to reverse global warming for us.
Furthermore, back-end non-anthropomorphic technologies may also play a role in theoretical physics, cosmology, and other fields to answer esoteric questions such as black hole information paradoxes and quantum entanglement. Their computing power and data analysis capabilities will provide humanity with a whole new perspective to understand the most intractable puzzles in our universe.
The future potential of these technologies is almost limitless, and they will become an extension of human intelligence, leading us into a new era of cognition. While this powerful ability may be both exciting and intimidating, it will undoubtedly greatly expand our understanding and control over the world.
3. The combination of the two: a synergy beyond imagination
Therefore, both anthropomorphic and non-anthropomorphic AI are the key to the future world. The combination of the two will not only combine human interaction with computing power beyond human comprehension, but will also usher in a new era of AI applications.
In this collaborative world, the role of front-end anthropomorphism resembles a warm, emotionally rich interface. They are able to create deep connections with human users with natural conversations, understanding, and emotional responses. At the same time, these technologies are able to gather valuable data about individual preferences, emotional states, and behavior patterns through continuous interaction with users.
The back-end non-anthropomorphic technology is working its superbrain function behind the scenes. Using data collected on the front end, these powerful systems can conduct deep Xi and sophisticated analysis to come up with solutions or predictions far beyond human intuition and cognition.
Fourth, the three-legged support - the supervision of AI technology security
The above visions do seem good, but there is still a potential danger hidden in these visions. Anthropomorphic and non-anthropomorphic technologies alone cannot guarantee that the future we are building with AI will be smarter, more efficient, and safer at the same time. In the context of the rapid development of AI technology, we must ask an important question: in a future where AI is becoming more and more powerful, how can we ensure that scenarios like those depicted in "Terminator", "The Matrix" or "The Wandering Earth" against humanity do not occur?
To avoid this, we need to build a "three-legged" system that balances the development, application, and oversight of AI technology. This means not only innovation at the technical level, but also a comprehensive consideration of ethical, legal and social responsibility.
AI security is also a noteworthy and significant issue, because of space limitations, I will bring it out for a separate discussion later when I have the opportunity.
5. The singularity is coming—the crossroads where humanity is surpassed
As we ride the wave of AI, we can't help but imagine a future full of possibilities. In this future, anthropomorphic AI will be our new friend, sharing laughter and providing comfort to us, while non-anthropomorphic AI will silently show its powerful computing power behind the scenes and solve scientific mysteries. The combination of these techniques is like a wonderful symphony, with both gentle melodies and shocking climaxes.
However, in this symphony of changing tunes, we must always be vigilant to ensure that the baton is firmly in the hands of humanity.
In this era of rapid development of AI technology, we may be standing at a historic crossroads - the tipping point of the singularity. This singularity, often portrayed as the moment when artificial intelligence surpasses human intelligence, may indeed be on the horizon. Our generation may be witnesses.
The day when AI surpasses humans may bring unprecedented opportunities, but it also comes with risks and challenges. At this crossroads, the choices and actions of each of us are crucial. So it's time to start thinking and preparing for how to ensure that our civilization, values, and human spirit can survive and thrive while respecting and harnessing the enormous potential of AI.
In this new era of AI, let's explore together, learn Xi together, and most importantly, determine our future together.
Columnist
has been a product Wang, WeChat public account: apmdogy, everyone is a product manager columnist. A logical product manager who is committed to combining scientific thinking with product manager methodology. Focus on artificial intelligence, education, good at product incubation, demand mining, project management, process management and other product skills.
This article was originally published by Everyone is a Product Manager and is prohibited from reprinting without permission.
The title image is from Unsplash and is licensed under CC0.
The views in this article only represent the author's own, everyone is a product manager, and the platform only provides information storage space services.