对现阶段AI技术的基本认识
文章目录
- 对现阶段AI技术的基本认识
- 一、What is AI
- 二、Building AI Project
- 三、Building AI in your Company
- 四、AI and Society
- 总结
一、What is AI
AI :ANI(artificial narrow intelligence,弱人工智能 self-driving car)AGI(artificial general intelligence 强人工智能 Do anything a human can do)
1、What is machine learning
machine learning:input(A) -> output(B)
2、What is data?
Acquiring data:
Data is messy:
3、The terminology of AI
Machine learning & Data Science
4、What make an AI company
tasks which AI company should have.
AI transformation(AI转型手册)
5、What machine learning today can and cannot do
为了增强自己或团队的直觉,判断是否是机器学习问题,做出可行性分析。
5.1 Strengths and weaknesses of machine learning
5.2、What is neural network
预测需求
face recognition
二、Building AI Project
1、workflow of machine learning project
2、workflow of data science project
3、the different workflow between data science and machine learning.
4、How to choose an worthy AI project
5、Due Diligence on project.
Due Diligence:
Buy or Build:
6、technical tools for AI
三、Building AI in your Company
1、case1:smart speaker
2、case2:self-drivering car
3、example roles of AI company
4、AI transformation PlayBook
转型AI公司手册(主要给CEO看的)下载链接https://landing.ai/ai-transformation-playbook/
4.1 Execute Pilot projects to gain momentum
4.2 Build an in-house AI team
4.3 provide broad AI training
4.4 Develop an AI strategy
4.5 Develop internal and external communication
5、AI pitfalls to awoid
6、Some initial steps you can take
7、AI in professional field
7.1 Computer Vision
7.2 Natural Language Processing
7.3 Robotics
8、Survey of major AI techniques
8.1 Unsupervised learning
8.2 Trasfer learning
8.3 Reinforcement learning
8.4 GANS (Generative Adversarial Network)
生成对抗网络(无中生有,制造新内容)
8.5 Knowledge Graph
知识图谱对建立数据库很重要
四、AI and Society
1、realistic view of AI
AI and Hype(大肆宣传)
2、Limitations of AI
2.1 The reasoning and anology of AI is biased
2.2 Why bias matters
为什么偏差问题这么重要
2.3 Combating bias
减少AI的偏见比减少人类的偏见更加简单
3、 Adversarial attacks on AI
对抗性攻击
计算机和人看图片的方式是不同的,有些图片稍加修改就可以愚弄AI
对抗性防御
4、Adverse use of AI
5、How developing economies can build AI
现在AI发展还是不成熟
用AI来加强国家的垂直行业
6、AI and jobs
总结
https://www.bilibili.com/video/BV1gE411M7Eg?p=35
备注:
- 该课程主要是以工程化的思想(任务流程分工)来向我们介绍AI技术。让我了解到
- AI是什么(ANI,AGI),数据是什么(结构化,非结构化的数据)
- machine learning的大致过程是什么(其实就是一个算法,输入和输出的映射)
- AI中的术语(神经网络:多个输入通过神经元建立联系)
- AI实际上能做和不能做的事
- AI公司的构成
- AI项目如何立项(技术评估:是否可行,商业价值评估)
- 机器学习和数据科学的工作流程比较。
- 而且从AI和社会的角度,让我重新用理性态度去认识AI
- 认识到AI系统的局限性:在推理类比上是存在偏见的(也许会存在对一部分人的歧视)
- AI与安全
- AI与经济发展
- AI与未来职位变化