AI & Society
Lecture 1: What is algorithmic decision making/ what is AI?
What is AI? Machine learning
It is about machine learning, the only way anything artificial can get intelligent is through
machine learning. It’s learning the artificial method systems. The algorithm is being trained
via the machine learning. This leads to new models, and this eventually leads to the
prediction and answers to questions. Process of fitting some sort of model made up by input
data, leads to new predictions.
We must see the relation between the technological aspect, and when AI is dangerous or
handy.
- Detection of systematic patterns between input and output
- General task: predict output given specific features of the inputs
- Very similar to regular statistical modeling:
o Input features: independent variables
o Output class: dependent variable
o In fact, neural networks can be seen as a form of logistic regression models
- Key difference to statistical modeling:
o We care about predicting something, not about understanding a (causal) process
o Models are highly complex and (multicollinear) and generally seen as ‘black box’
The magic is not how it learns from data, it is about what it does with the data.
Deep learning
- Fancy term for machine learning with very large models
- Based on:
o Very large neural networks
o Trained on enormous amounts of data, e.g., “all of the internet”
o Using massive computing power, especially of GPU’s
,AI & Society
- Key innovations:
o Feature layers find patterns in raw input
o Networks can be (pre-) trained based on unannotated data
o Patterns from (pre-) training are transferred to actual task, and fine-tuned on
annotated data
Marketing term that was applied to the term of artificial intelligence. The process of turning
hundreds of documents in such model is hard computer working. These systems are very
powerful.
The revolution in deep learning is that we don’t have hundreds of documents of
unannotated data, but we have documents that are annotated. Annotated data can lead to
more finetuned new data.
Natural language processing
- Core application area of AI
- Research field of NLP or computational linguistics
- Can the computer understand, generate, or translate text?
- Generally cast as machine learning problems:
o Predict specific meaning given text
o Generate (= predict most likely) newspaper article given prompt
- Recent innovations all based on ‘BERT’: Decoder-encoder architecture that builds layers
of understanding trained using tasks such as ‘predict the next world’
Robotics
- ‘Embodied intelligence’: AI used to navigate/understand and manipulate environment
- Industrial robots
- Automotive AI
- Conversational/ care robots
, AI & Society
AI In a complex world
1. Scale of digitalization (ß digital traces as research objects)
2. AI/ Digitalization build on existing technologies
3. Global reach of AI/ digitization
4. Ubiquity of AI/ digitization
5. Increasing complexity of digitized/ AI systems
6. Intrusiveness of digital/ AI technology
Ethics of AI: main debates
- Privacy
- Manipulation
- Opacity
- Bias
- Autonomy
- The singularity
AI, data, network effects, and the surveillance society
- “For example, facial recognition technology is used in many parts of China nowadays
that tracks the emotions of people, and you can only enter some metro stations if your
past behavior has gotten a high score in the nationwide system of data-sharing”
- What is the role of privacy in the face of machine learning?
- How do “anonymous data” dude from decisions based on personal data?
- Is the GDPR a good and sufficient instrument?
AI & Journalism (JMCQ 2019 Invited Forum)
- Meredith Broussard:
o Center humans, consider economics, explain what AI is(n’t)
o Not: what can AI do, but what should AI do
- Nicholas Diakopoulos:
o AI as a medium to express journalistic values; value-centered design
o Need to study human-centered (“hybridized”) AI
- Andrea Guzman:
o Need to cross divides in research (and applications)
o Disciplinary divides, technological divides, theoretical divides
o Human-Machine Communication as research agenda
- Rediet Abebe:
o Biases/limitations often discriminatory
o Need for inclusion/representation of communities
- Michael Dupagne & Chin-Hua Chan:
o More research into fear of and resistance to technology
o Need to adapt education to changing reality
Conclusions
- AI is a group of techniques clustered around machine learning
- Deep neural networks can achieve spectacular results
o Text/image generation, understanding, translation, decision making
Lecture 1: What is algorithmic decision making/ what is AI?
What is AI? Machine learning
It is about machine learning, the only way anything artificial can get intelligent is through
machine learning. It’s learning the artificial method systems. The algorithm is being trained
via the machine learning. This leads to new models, and this eventually leads to the
prediction and answers to questions. Process of fitting some sort of model made up by input
data, leads to new predictions.
We must see the relation between the technological aspect, and when AI is dangerous or
handy.
- Detection of systematic patterns between input and output
- General task: predict output given specific features of the inputs
- Very similar to regular statistical modeling:
o Input features: independent variables
o Output class: dependent variable
o In fact, neural networks can be seen as a form of logistic regression models
- Key difference to statistical modeling:
o We care about predicting something, not about understanding a (causal) process
o Models are highly complex and (multicollinear) and generally seen as ‘black box’
The magic is not how it learns from data, it is about what it does with the data.
Deep learning
- Fancy term for machine learning with very large models
- Based on:
o Very large neural networks
o Trained on enormous amounts of data, e.g., “all of the internet”
o Using massive computing power, especially of GPU’s
,AI & Society
- Key innovations:
o Feature layers find patterns in raw input
o Networks can be (pre-) trained based on unannotated data
o Patterns from (pre-) training are transferred to actual task, and fine-tuned on
annotated data
Marketing term that was applied to the term of artificial intelligence. The process of turning
hundreds of documents in such model is hard computer working. These systems are very
powerful.
The revolution in deep learning is that we don’t have hundreds of documents of
unannotated data, but we have documents that are annotated. Annotated data can lead to
more finetuned new data.
Natural language processing
- Core application area of AI
- Research field of NLP or computational linguistics
- Can the computer understand, generate, or translate text?
- Generally cast as machine learning problems:
o Predict specific meaning given text
o Generate (= predict most likely) newspaper article given prompt
- Recent innovations all based on ‘BERT’: Decoder-encoder architecture that builds layers
of understanding trained using tasks such as ‘predict the next world’
Robotics
- ‘Embodied intelligence’: AI used to navigate/understand and manipulate environment
- Industrial robots
- Automotive AI
- Conversational/ care robots
, AI & Society
AI In a complex world
1. Scale of digitalization (ß digital traces as research objects)
2. AI/ Digitalization build on existing technologies
3. Global reach of AI/ digitization
4. Ubiquity of AI/ digitization
5. Increasing complexity of digitized/ AI systems
6. Intrusiveness of digital/ AI technology
Ethics of AI: main debates
- Privacy
- Manipulation
- Opacity
- Bias
- Autonomy
- The singularity
AI, data, network effects, and the surveillance society
- “For example, facial recognition technology is used in many parts of China nowadays
that tracks the emotions of people, and you can only enter some metro stations if your
past behavior has gotten a high score in the nationwide system of data-sharing”
- What is the role of privacy in the face of machine learning?
- How do “anonymous data” dude from decisions based on personal data?
- Is the GDPR a good and sufficient instrument?
AI & Journalism (JMCQ 2019 Invited Forum)
- Meredith Broussard:
o Center humans, consider economics, explain what AI is(n’t)
o Not: what can AI do, but what should AI do
- Nicholas Diakopoulos:
o AI as a medium to express journalistic values; value-centered design
o Need to study human-centered (“hybridized”) AI
- Andrea Guzman:
o Need to cross divides in research (and applications)
o Disciplinary divides, technological divides, theoretical divides
o Human-Machine Communication as research agenda
- Rediet Abebe:
o Biases/limitations often discriminatory
o Need for inclusion/representation of communities
- Michael Dupagne & Chin-Hua Chan:
o More research into fear of and resistance to technology
o Need to adapt education to changing reality
Conclusions
- AI is a group of techniques clustered around machine learning
- Deep neural networks can achieve spectacular results
o Text/image generation, understanding, translation, decision making
