Robot interaction
Lecture 2 + Literature
,Lecture 2 - summary
st
1 paper:
Roboti cs and AI: diff erences and similariti es
Difference between robotics and artificial intelligence:
1. Robots -> analogue world
Branch of electro-mechanical engineering. Robots are programmable, physical machines.
2. artificial intelligence -> digital world
The digital world branch of computer science. Computer programs that complete tasks which would
otherwise require human intelligence.
You can have one without the other
1. Robots interact with the physical world - with sensors and actuators
2. Robots are programmable (similar to AI)
3. Robots are usually (semi) autonomous, but they can also be remote controlled
4. AI algorithms can tackle learning, perception, problem solving, language processing and/or
logical reasoning - without sensors or actuators
If you use AI in a robot, you talk about artificially intelligent robots (usually the case – but it does not
have to be!)
2th paper:
Design lessons from AI’s two grand goals: Human emulati on and useful applicati ons
Common goals in AI research:
1. Theory development
o Mainly to understand human/animal/plant abilities
Doing so you gain knowledge to build a system that outperforms living
organisms
2. Application
o AI methods to build widely used products and services
The “imitation game” tells us that Turing thinks it won’t matter if computers have a consciousness or
not, all that matters is what people think of them.
- Thus - AI is a rhetorical device to influence human perception
Theory development in AI concerns the following topics:
1. Pattern recognition (images, speech, facial, signal)
2. Natural Language Processing
3. Translation of natural language
4. Bipedal robots (make a robot walk)
5. Emotionally responsive human faces
6. Game playing (chess, checkers)
Why AI if other engineering solutions work better?
, You do not need intelligence when the whole problem space is determined - you can
simply use logics (if your memory is great enough). If you process logic fast enough, you can beat
humans. - Therefore Brute Force is capable
Brute force is used when no faster algorithm is known to solve a particular problem - it just explores
all possibilities to solve a problem (brute force), however it requires a lot of memory and computing
power.
Human emulation and simulation will sometimes offer worse solutions than “non-natural”
inventions.
Mumford (1934)
- First attempts at new technologies are misled by humans and animal models
o Because we tend to copy the knowledge gain from human and animal studies
and according to Mumford, humans and animals do not always have the most
desired characteristics.
o For example:
4 wheels have large advantages over 2 legs.
Airplanes fly much faster and do not flap like bird wings.
Summary
A blank sheet robot is very useful, since it let us start at the very beginning. This makes us realize that
what we think we know is not enough to create another living being.
Robots are electro-mechanical, they can have AI insight, but it is not necessary.
A robot does not have to be autonomous.
AI do not have to make use of sensors and actuators.
Thus, any AI that, apart from a computer, screen, keyboard, is not concerned with physical 1/O
devices is not a robot. Thus a web crawler (a computer program insight a computer) is not called a
robot.
An avatar or chatbot is also not called a robot, although these applications are highly convenient for
simulating and testing human behavior with robots.
The different possibilities with AI:
1. AI
2. AI + on-screen avatar
3. AI + electro-mechanical robot (pepper, Sophia)
4. AI + electro-mechanical devices (Apple’s siri, Amazons’s Alexa, Google’s Home)
o This is something in-between an avatar and a robot
Theory and application are two different worlds
AI and robotics for theory development is to simulate humans to understand humans (or other
organisms).
Application can employ all tricks and engineering shortcuts to optimize the user experience. It is wise
to be informed by theory but not always follow it.
