Where is the Future of Artificial Intelligence Heading
-
The Forbes biweekly website recently listed five major trends in the development of artificial intelligence (AI) in 2024. The Ministry of Industry and Information Technology issued the 'Guiding Opinions on the Innovative Development of Humanoid Robots,' proposing that by 2025, the innovation system for humanoid robots will be initially established, with whole-machine products reaching international advanced levels and achieving mass production. With the emergence of 'generative AI,' the distance between us and artificial intelligence is gradually narrowing. What new aspects will future AI bring to the world?
Generative AI Opens New Frontiers
In the field of artificial intelligence, a new hotspot emerges every few years. With the launch of ChatGPT, generative AI has made many people new users of AI tools.
Generative AI is a new type of artificial intelligence that learns from large-scale data to generate new original content (text, images, videos, etc.). ChatGPT can not only write essays, novels, and solve math problems but even interpret horoscopes. It is evident that while the generated content is very similar to the training data, it is not a simple replication.
Since the mid-20th century, the development of artificial intelligence has continuously broken through. In the 1970s, the focus was on expert systems, and in the 1980s, it shifted to probabilistic reasoning. Over the past decade, the field has been driven by machine learning, which involves computers learning from data to identify patterns and build models. In recent years, this machine learning has continued to advance, with generative AI introducing deep learning techniques. If previous AI learning relied mainly on human experience to make judgments, generative AI more closely mimics the human brain's thought processes, enabling more abstract and comprehensive judgments.
What are the potential application scenarios for generative AI in the future?
Imagine an AI that, upon hearing descriptive speech, can instantly draft a rich article, paint a picture, and pair it with suitable background music. Forbes predicts that future AI systems will surpass simple applications like chatbots and parody videos, enabling them to write complex narratives, co-author bestsellers, and even compose symphonies.
The key innovation lies in multimodal generative AI, which processes diverse inputs—text, sound, melody, visual signals—and synthesizes them for comprehensive understanding. This will deliver richer, multi-sensory experiences to users.
The rise of generative AI will also drive profound changes in scientific research and industrial design. A recent breakthrough involves AI solving protein-folding problems. Traditional experimental methods took weeks or months to analyze a single protein structure, but AI achieves over 90% accuracy in hours or even minutes.
Starting next year, as multimodal technology advances, AI models will delve into more complex and diverse interactive scenarios, unlocking new applications in smart homes, smart cities, medical diagnostics, autonomous driving, and beyond.
Humanoid Robots Learning to Interact with the Environment
If asked what the most iconic tool of the digital economy era is, robots would undoubtedly be one of the strongest answers. Humanoid robots, in particular, are regarded as the hallmark product of the 'Fourth Industrial Revolution.'
In September this year, Tesla released the latest video of its humanoid robot, Optimus. 'Optimus has improved again,' many commented. Elon Musk stated plans to mass-produce this product within 3 to 5 years. Meanwhile, China's Ministry of Industry and Information Technology recently issued the Guidelines for the Innovative Development of Humanoid Robots (hereinafter referred to as the Guidelines), proposing that by 2025, an innovation system for humanoid robots will be preliminarily established. Breakthroughs will be achieved in key technologies such as the 'brain, cerebellum, and limbs,' with the overall product reaching international advanced levels and entering mass production.
In 1950, Alan Turing, the 'father of artificial intelligence,' first proposed the concept of 'embodied intelligence' (humanoid robots). The core hypothesis is that intelligent behavior can be acquired by an intelligent entity with a corresponding form through adaptation to the environment. In other words, to possess human-like intelligence, one must first have a human-like body.
Lu Cewu, a professor in the Department of Computer Science at Shanghai Jiao Tong University's School of Electronic Information and Electrical Engineering, once described it this way: From a cognitive perspective, humans are 'first-person perspective intelligence,' while machines without bodies are 'third-person perspective intelligence.' Humans can actively learn through interactions with the real world and understand new concepts by expanding their chain of thought. The breakthrough for humanoid robots lies in not just relying on humans to 'feed data' but learning independently through massive data to acquire the ability to interact with the environment.
For robots, the combination of 'brain' and 'cerebellum' is essential to better understand and interact with the world.
The first humanoid robot, 'Expedition A1,' developed by Shanghai-based Agibot, uses large models as its 'brain.' This tests the 'brain's' ability to make decisions and generate solutions for tasks it has never encountered. 'Expedition A1' analyzes the speaker's intent through verbal instructions and executes tasks accordingly. For example, when given commands like 'I want a glass of water' or 'I'm thirsty,' it can accurately 'understand' and proceed to fetch water from the dispenser and hand over the cup.
The Guidance proposes that breakthroughs in humanoid robotics should focus on key technologies such as the 'brain' and 'cerebellum,' as well as the 'limbs,' leveraging advancements in AI like large models. For instance, developing the 'brain' of humanoid robots to enhance environmental perception, behavior control, and human-machine interaction; and refining the 'cerebellum' to improve agile movement. Additionally, it emphasizes systematically deploying key technologies for 'robotic limbs,' such as human-like arms, dexterous hands, and legs, and advancing the integration of humanoid robots with cutting-edge technologies like the metaverse and brain-computer interfaces.
In the view of Gu Jie, Chairman and CEO of Fourier Intelligence, humanoid robots may become the next-generation terminals, akin to cars and smartphones, widely adopted in scenarios like elderly care, rehabilitation, and scientific research.
AI Should Become Humanity's 'Right and Left Hand'
It is foreseeable that with the continuous advancement of technology, artificial intelligence will no longer be just a tool but humanity's "right and left hand." The world is entering an era where humans and AI collaborate synergistically.
For example, surgeons can perform operations with the assistance of AI diagnostics, and some hospitals have already implemented robotic-assisted surgery. Lawyers can access comprehensive reference cases provided by AI during trials. Software developers can also receive real-time coding assistance, significantly improving efficiency.
Additionally, quantum computing will become a crucial development direction in the future of AI. Both emerging startups and established tech giants are allocating substantial resources to develop quantum solutions, as they can accelerate machine learning speeds and optimize algorithms, thereby enabling more efficient and accurate AI applications.
Although artificial intelligence fills the future with hope, it is undeniable that many potential risks coexist. We need to develop corresponding strategies to address these challenges.
For example, generative AI poses challenges to the existing intellectual property system. The main issues include: how to balance the interests between material providers and re-creators, protect innovative achievements, and minimize disruptions to related industries. Addressing this problem requires significant reforms to the current intellectual property framework.
As AI's involvement in daily work and life increases, its ethical dilemmas have drawn considerable attention. Experts believe the current challenges lie not only in designing fair algorithms but also in establishing strict standards to ensure these systems and their designers are accountable for their actions.
According to reports from Forbes' biweekly website, AI is advancing rapidly across various fields. This unprecedented development has captivated technology enthusiasts while drawing close attention from global policymakers.
Major economies including Europe and America are striving to formulate comprehensive AI policies and legislation. In October this year, the United States signed its first executive order on AI. The European Parliament aims to reach an agreement on the EU's AI Act text by June next year. These policies seek to promote technological breakthroughs while protecting the public from the unbridled growth of artificial intelligence.
Industry insiders believe that deliberately not planning artificial intelligence's creativity and leaving some creative space for humans is a highly noteworthy and explorable issue.