**Foreword:**
This article presents the latest research from the Future Intelligent Lab on artificial intelligence (AI) and its potential IQ in the future. The study introduces the concept that an intelligent system's intelligence level can be categorized into three distinct AI IQ types, each tailored for different testing purposes. For these three AI IQs, the paper also proposes specific testing methods and mathematical formulas to evaluate their performance.
In our research, we observed that when people discuss the development level of AI, their needs and goals vary significantly, leading to different evaluations of AI IQ. The first purpose is to assess whether an AI system or robot surpasses human intelligence. The second goal is to understand how smart a smart product is when serving humans and what it costs. Based on this distinction, the Future Intelligent Lab suggests that there should be three types of AI IQ: General IQ, Service IQ, and Value IQ.
**0. Background**
Since 2016, with AlphaGo defeating the human Go champion, the rapid development of artificial intelligence around the world, and the spread of AI threat theories, the need to measure the intelligence level of smart products has become more pressing. Can AI surpass humans? What is the current level of AI intelligence? Answering these questions requires a quantitative approach to evaluating the development of intelligent systems.
The Turing test, introduced in 1950, has been widely used as a method to assess AI intelligence. However, it only determines whether a machine can mimic human intelligence, not its actual developmental level. It is also heavily influenced by human judgment, which makes it subjective and often unreliable.
In 2015, a paper published in PNAS introduced the "Visual Turing Test," focusing on computer image recognition. In 2014, Professor Mark O. Riedl proposed the Lovelace 2.0 test, emphasizing creativity in AI, including tasks like writing novels, composing poetry, painting, and creating music.
Despite these efforts, existing tests lack a unified model for AI intelligence and fail to provide comprehensive quantitative analysis. They often focus on isolated aspects of intelligence without considering how close an AI system is to human-level intelligence or how fast it is developing compared to humans.
To address these issues, the research team proposed three AI IQs: General IQ, Service IQ, and Value IQ. This paper elaborates on the theoretical basis, definitions, and evaluation methods for each.
**1. Theoretical Basis: Standard Intelligent Systems and Extended von Neumann Architecture**
Evaluating AI intelligence faces two major challenges: the lack of a unified model for AI systems and the absence of a common framework between AI and human intelligence.
In 2014, the Future Intelligent Lab proposed a "Standard Intelligent Model" based on the von Neumann architecture, integrating concepts like knowledge acquisition, mastery, innovation, and feedback. This model allows for a unified description of both AI and human intelligence.
Building upon this, the extended von Neumann architecture adds creative capabilities, enabling AI to discover new knowledge and rules, store them, and interact with the environment through input/output systems. It also includes external knowledge bases, allowing for shared learning across systems.
**2. Definition of Three Different AI IQs**
**2.1 AI General IQ**
Based on the standard intelligent model, the research team developed the AI IQ test scale. Testing over 50 AI systems, including Google, Siri, Baidu, and children aged 6, 12, and 18, revealed that while AI has improved, it still lags behind young children in certain areas.
The AI General IQ measures the overall intelligence level of an AI system, treating it as an equal agent in interactions with other agents. It provides a universal score to evaluate AI development.
**2.2 AI Service IQ**
Most AI systems are designed to serve humans, and their intelligence is measured by how well they perform in service. The Service IQ evaluates how effectively an AI serves humans, incorporating ethical and legal constraints that may not be part of general IQ.
**2.3 AI Value IQ**
Value IQ considers the economic cost of an AI system relative to its performance. It helps users determine whether a smart product offers good value for money, based on its service IQ and price.
**3. Design of AI IQ Test Scales**
**3.1 General IQ Test Scale**
The AI General IQ test divides intelligence into four categories: knowledge acquisition, mastery, innovation, and feedback. These are further divided into 15 subcategories, such as image recognition, translation, and creation.
In 2017, the test was updated to include dynamic image recognition, emotional expression, and world transformation abilities, among others.
The formula for General IQ is:
**IQAIG = Σ(FGi × WGI)**
Where FGi is the score for each subcategory, and WGI is its weight.
**3.2 Service IQ Test Scale**
Service IQ focuses on an AI’s ability to sense user identity, interact with cloud systems, protect users, and comply with local laws. It also includes self-energy management and real-time status updates.
The formula for Service IQ is:
**IQAIS = Σ(FSi × WSi)**
**3.3 Calculating AI Value IQ**
Value IQ is calculated by dividing Service IQ by the product price:
**IQAIV = (IQAIS / P) × 100**
**4. Conclusion**
With the three AI IQs—General, Service, and Value—intelligent systems can now be evaluated from multiple perspectives. These IQs reflect different usage scenarios and goals, offering a more comprehensive understanding of AI intelligence.
Future work will involve applying these IQs to specific smart products, such as smart speakers, cars, and refrigerators, to better assess their performance.
The Future Intelligent Lab is a collaborative research institute combining AI, the Internet, and brain science. It aims to develop AI IQ evaluation systems, conduct global AI IQ assessments, and enhance city and enterprise intelligence through cloud-based technologies.
Tier 1 Solar Panel
Chinese Tier 1 brand Solar Panel (Longi, Jinko, Trina, Canadian, Ja etc.)
This type of product is mainly Tier 1 solar module product in China, include Longi solar panel. Jinko solar panel, Canadian Solar panel, ja solar panel etc.all of which are among the top 10 world-renowned brands. The original A grade product is mainly mono crystalline and half cut cell solar modules, with a power range of 400watts to a maximum of 700watts.quality warranty is 25 years.
A grade, this type of film is characterized by a neat and uniform appearance, printed electrode grid lines, no broken grid, no surface stains, no leakage, no aluminum package, no hidden cracks, no scratches, no black core, no reverse current, stable electrical performance, and high conversion efficiency
Using solar cell testing equipment to measure and screen, the conversion efficiency, filling factor, and appearance that meet certain standards are first level. Generally, the conversion efficiency, filling factor, and appearance of first level are better than those of level B
The level of solar panels can be divided into first level, second level, third level, and fourth level, and the first level components can be divided into first level+and first level - two levels. The second level is also the same, and the cost difference between different levels of solar panels is also very large.
Chinese tier 1 brand solar panel
|
TIER 1 brand include
|
Longi, Jinko, Trina, JA, Canadian, znshine, etc.
|
|
quality warranty
|
25 years
|
|
grade
|
original A grade (each solar panel can be checked on official website with barcode)
|
|
power range
|
400watt to max 700watt
|
Tier 1 Solar Panel,Mono Solar Panel N Type,Solar Panels Bifacial 555 Watt,Solar Panel Mono Topcon Cell
PLIER(Suzhou) Photovoltaic Technology Co., Ltd. , https://www.pliersolar.com