"Popular Science: What Level Does AI Need to Reach to Fully Replace Human Work?
[Original Title: Popular Science: What Level Does AI Need to Achieve to Completely Replace Human Work?]
[Image Description: An image showing a person interacting with an AI interface.]
The author of this article, Chen Yunwen, is the CEO of Daguan Data. He holds a Ph.D. in Computer Science from Fudan University and has won championships in international data mining competitions such as those organized by ACM. He has previously held positions such as the Chief Data Officer of Shanda Literature, Senior Director at Tencent Literature Data Center, and core technical engineer at Baidu.
For different application scenarios, the practical level of artificial intelligence in industries varies. According to internationally renowned artificial intelligence expert Professor Sandeep Rajani, in his article "Artificial Intelligence: Man or Machine," he compares AI capabilities horizontally with human abilities and divides them into four distinct levels:
1. Peak Level – Has achieved an unsurpassable optimal capability.
2. Beyond Human Level – Stronger than all humans.
3. Strong Human Level – Stronger than most humans.
4. Weak Human Level – Weaker than most humans.
In the textbook "Intelligent Web Algorithms" (2nd Edition), which introduces computer algorithms and data mining techniques, Chen Yunwen translated and introduced Professor Rajani’s method of categorization. In the current technological landscape, the levels achieved by AI in various existing application areas are as follows:
Peak Level:
In some application scenarios with complete rules and a small strategy space, such as playing backgammon on a 19x19 board, today's computer technology can exhaust all possible game states to ensure optimal solutions in any game. Similarly, in simpler games like Tic-Tac-Toe and checkers, as well as statistical data applications, computers have already reached the peak level.
Beyond Human Level:
The matches between AlphaGo and Ke Jie, and Deep Blue defeating chess champion Garry Kasparov, demonstrated that AI has surpassed human capabilities in complex board games. In specific application areas such as fingerprint and iris recognition, AI has also reached a level beyond human abilities.
Strong Human Level:
Some intellectual activities require deep field experience. Compared to a small number of professionals, computers still lag behind but outperform the general population. For instance, in areas like Texas Hold'em and bridge, computers have excelled in certain specialized fields such as facial recognition under good conditions (no obstructions, lighting issues, or angle problems), and speech recognition without regional accents or noisy environments. In areas like identifying plant species, computer capabilities have also reached a strong human level.
Weak Human Level:
Many skills are easily mastered by most humans, such as driving a car. However, for a computer system, due to the complexity of the signals to be collected and analyzed, it remains challenging to reach the level of ordinary humans, placing it at a weak human level. Common fields include writing articles, reading comprehension, and human language translation.
The varying levels of AI development across different fields are influenced by three main factors:
Factor 1: Rules and Clarity of Evaluation Methods
The simpler and clearer the rules, and the more objective the criteria for evaluating performance, the lower the cost to implement AI. Chess and card games, for example, allow computers to perform exceptionally well.
However, problems with greater uncertainty pose significant challenges for AI. Driving a vehicle, for instance, involves numerous variables, no clear "win/loss" or "good/bad" driving styles, and unpredictable environmental factors. Similarly, games like Mahjong or poker, with their random elements, present additional hurdles.
Thus, clearer rules and more objective evaluation criteria enhance the practical effectiveness of AI technologies. In ambiguous rule-based areas, human approaches and everyday solutions tend to yield better results.
Factor 2: Practical Application Scenarios and Exceptions
Many application problems are handled under typical scenarios but may involve unique exceptions, presenting varying degrees of difficulty.
Take face recognition as an example. Existing machine learning technologies perform well under ideal conditions—correct angles, no occlusions, and good lighting. However, in practical use, it encounters numerous challenges such as variations in lighting, angles, makeup, age changes, and deliberate attempts to deceive the system. These factors significantly impact real-world performance.
Autonomous driving technology also faces similar challenges, with varying difficulties in normal versus adverse weather conditions, and diverse road conditions posing substantial obstacles to practical implementation.
While many AI applications show promising results in lab settings, industrial applications face more complex and harsher conditions, necessitating the management of anomalies and interference factors. Thus, the actual level of many applications remains between the strong human and weak human levels.
Currently, moving AI from lab to practical application presents numerous challenges. A pragmatic approach involves restricting specific scenarios, minimizing uncertainty, and simplifying problems. For example, limiting autonomous driving applications to fixed routes or closed circuits drastically reduces technical complexity, enabling the system to move up one or two levels in terms of practical applicability. In text comprehension, narrowing down industry types and text understanding can significantly improve system accuracy.
Machines also have advantages in terms of "uncertainty." Humans' finite energy leads to declining judgment accuracy over extended periods, affecting work quality. Machines, however, maintain stable performance under prolonged high-pressure environments, avoiding fatigue and emotional influence. This makes machine outputs more deterministic compared to human work.
Factor 3: Data Accumulation and Algorithm Models
As the adage goes, "Big data + algorithm models = artificial intelligence." The key to AI lies in big data. Only by accumulating vast amounts of training data can AI capabilities be enhanced. AlphaGo's defeat of top human players was made possible through tens of millions of Go game data and comprehensive model training.
In the real world, the accumulation of training data has only just begun, especially labeled data for supervised learning, which requires significant manual effort and incurs high costs, restricting AI advancements in related fields. Additionally, some data is limited by policy factors, such as medical data, or monopolized by certain industries, hindering data circulation and slowing AI progress.
Recent reductions in computer hardware storage costs, the growing popularity of cloud computing, and improving data accumulation environments have spurred data collection awareness. It is hoped that with application demands, more data will be digitized and recorded, leading to better-trained algorithm models and enhanced effects.
Starting from the weak human level, progressing to the strong human level, beyond human level, and reaching the peak level, there is still a long way to go. Technological advancement typically begins easily before becoming complex and challenging. With continued accumulation and progress, it is believed that in the future, AI technology can replace humans in performing increasingly valuable tasks.
Chen Yunwen, Founder and CEO of Daguan Data, holds a Ph.D. in Computer Science from Fudan University. He is the Vice President of the Shanghai Computer Society's Multimedia Society, and a senior member of both the ACM and IEEE. He has served as the Chief Data Officer of Shanda Literature, Senior Director of Tencent Literature, Head of Data Center, and core technology R&D engineer at Baidu. He has represented China multiple times in international data mining competitions such as ACM and won championships."
[Image Description: An image of a person working with advanced technology.]
As the demand for electric scooters continues to rise, we understand the importance of providing a reliable and safe power source for these vehicles. Our lithium battery packs are designed to meet these demands, featuring the latest lithium-ion technology and precise engineering to provide optimal performance and longevity. Our E Scooter Lithium Battery Packs are available in various voltage and capacity options, allowing you to choose the perfect fit for your electric scooter. Whether you need a lightweight, compact battery for easy portability or a higher-capacity battery for longer travel distances, we have you covered.
Light Electric Vehicle Battery,E Scooter Lithium Battery Pack,24V 210Ah Lifepo4 Battery Pack,24V Folklift Lifepo4 Battery
JIANGMEN RONDA LITHIUM BATTERY CO., LTD. , https://www.ronda-battery.com