Artificial Intelligence

Prof. Dr. Mirco Schoenfeld

Hello

Nice to meet you!

About me

Professor for Data Modelling & Interdisciplinary Knowledge Generation at University of Bayreuth

Computer scientist who enjoys collaborating with people from other disciplines.

Questions

I have questions…

Where do you get in touch with AI?

What do you think is AI?

Do you use AI?

Do you actively use AI?

What is AI

What is AI?

AI in the wild

https://www.europarl.europa.eu/topics/en/article/20200827STO85804/what-is-artificial-intelligence-and-how-is-it-used

AI in the wild

AI can

explain
detect
recognize
decide
predict
act
generate

Short history of AI

From Turing to…

(a brief history of AI)

Short history of AI

Stepping stones:

Turing Test (1950)
Neural Networks (1957)
Backpropagation Algorithm (1986)
Transformer Architecture (2017)

Turing

Turing (1950)

Turing

In essence, a computer can’t be distinguished from a human
by another human.

Blockhead

In 1981, Ned Block imagined a system
indistinguishable from adult humans.

Block focused on responses to inputs

(Millière and Buckner 2024; Block 1981)

Blockhead

Blockhead thought experiment:

[In natural language], there are a finite number of syntactically- and grammatically-correct sentences that can be used to start a conversation. Consequently, there is a limit to how many “sensible” responses can be made to the first sentence, then to the second sentence, and so on until the conversation ends.

(Wikipedia 2024)

Blockhead

Now, imagine a computer that knows all these
sentences and possible conversations.

Would it pass the Turing test?

Blockhead

And is it intelligent?

Intelligence

Intelligence is not a matter of observable behavior of a system

It also depends on the nature and complexity of
internal mechanisms that drive this behavior

(Millière and Buckner 2024)

GPT-4

What does this mean for GPT-4?

It passed the Turing test recently… (Jones and Bergen 2024)

From AI to ML

artificial intelligence

From AI to ML

artificial intelligence

machine learning

From AI to ML

artificial intelligence

machine learning

deep learning

From AI to ML

artificial intelligence

machine learning

deep learning

LLMs

Machine Learning

The objective of machine learning is
to formulate a good predictive model.

Machine Learning

A good model predicts well for new data.

Machine Learning

(Pichler and Hartig 2023)

Typical bias-variance tradeoffs in classical machine learning (left panels) and deep learning (right panels) models for interpolation (in-distribution) and extrapolation (out-of-distribution) tasks. In contrast to the classical bias-variance trade-off in panel (a), the bias-variance trade-off for DL in panel (b) shows that after the interpolation threshold (pink dotted line) the training loss is constant (i.e. bias is not improved by increasing model complexity), but the test loss (and thus variance) can be still reduced by increasing the model size. Note also that the total generalization error in extrapolation tasks (panels c and d) is usually higher and the optimal model complexity lower, as the bias will not go to zero with increasing model complexity (depending on the similarity between training and test data).

Kinds of ML

Kinds of machine learning

unsupervised learning
supervised learning
reinforcement learning

unsupervised learning

supervised learning

reinforcement learning

important classes

Important classes in unsupervised learning

agglomerative hierarchical methods
a priori methods

important classes

Important classes in supervised learning

distance based classes
tree based classes
neural nets

and deep learning?

What is deep learning after all?

and deep learning?

Deep learning models are neural networks.

In huge.

Criticism

So, AI gives us the perfect new world?

Criticism

Participants who used AI produced fewer ideas, with less variety and lower originality compared to a baseline.

(Wadinambiarachchi et al. 2024)

Criticism

at least some generative AI systems may produce plagiaristic outputs

(Marcus and Southen 2024)

Criticism

Expectations

What you can expect from this course:

a deepened understanding of algorithmic thinking
introduction to machine learning
hands-on experiences in machine learning
introduction to deep learning & LLMs
critical perspective on the topic of AI
ethical considerations

What do you expect from this course?

Learning by doing

We are using R

Why R

Why R?

high level language that abstracts many complexities
learning curve not too steep
useful for your area of expertise
visualizations look pretty good out of the box

Let’s get started

References

Block, Ned. 1981. “Psychologism and Behaviorism.” Philosophical Review 90 (1): 5–43.

Commons, Wikimedia. 2023. “File:turing Test Diagram.png — Wikimedia Commons, the Free Media Repository.” https://commons.wikimedia.org/w/index.php?title=File:Turing_test_diagram.png&oldid=772447387.

Jones, Cameron R., and Benjamin K. Bergen. 2024. “Does GPT-4 Pass the Turing Test?” https://arxiv.org/abs/2310.20216.

Marcus, Gary, and Reid Southen. 2024. “Generative AI Has a Visual Plagiarism Problem.” https://spectrum.ieee.org/midjourney-copyright.

Millière, Raphaël, and Cameron Buckner. 2024. “A Philosophical Introduction to Language Models – Part i: Continuity with Classic Debates.” https://arxiv.org/abs/2401.03910.

Pichler, Maximilian, and Florian Hartig. 2023. “Machine Learning and Deep Learning – a Review for Ecologists.” Methods in Ecology and Evolution 14 (4): 994–1016. https://doi.org/https://doi.org/10.1111/2041-210X.14061.

Turing, Alan M. 1950. “Computing Machinery and Intelligence.” Mind 59 (236): 433–60.

Wadinambiarachchi, Samangi, Ryan M. Kelly, Saumya Pareek, Qiushi Zhou, and Eduardo Velloso. 2024. “The Effects of Generative AI on Design Fixation and Divergent Thinking.” In Proceedings of the CHI Conference on Human Factors in Computing Systems. CHI ’24. ACM. https://doi.org/10.1145/3613904.3642919.

Wikipedia. 2024. “Blockhead (Thought Experiment) — Wikipedia, the Free Encyclopedia.” https://en.wikipedia.org/w/index.php?title=Blockhead_(thought_experiment)&oldid=1221080543.