208

14.5

Exercises for Chap. 14

Task 14.1

What does Gödel’s theorem say?

Task 14.2

What does “Turing-computable” or “non-Turing-computable” mean?

Task 14.3

What’s the Babbage test?

Task 14.4

(a) Find out about neural networks. Find TMHMM on the net and use it.

(b) Look at the ELM server, what predictions of the ELM server do neural net­

works use?

(c) The protein secondary structure prediction “PredictProtein” uses neural net­

works. Get an overview by using the software and the given references.

Task 14.5

Find a neural network software.

Conclusion

• A computer (as conceived by Turing as a Turing machine) cannot reason about

itself. Formal systems have basic bounds (exactly proved by Gödel and Turing),

what they can prove or decide and what not. Humans (and living beings in gen­

eral) may not think formally exactly, but therefore can think about themselves

and, in general, all fundamental questions more successfully. Therefore, goals

and values must and should always be set by humans, especially when computers

become more and more powerful.

• Artificial intelligence, in particular deep learning algorithms and neural net­

works, is giving a further boost to the capabilities of computers. The more fea­

tures of a living being are emulated (e.g. acting in an artificial environment,

emulating language and emotions), the stronger its capabilities become.

• In bioinformatics, the properties of artificial intelligence can be used directly for

modern image processing, for example, but also in general for the recognition of

complex properties (feature extraction), for pattern recognition in large data sets

(training data set) and then also for individual molecules or sequences (predic­

tions, for example, for the secondary structure in the protein, for the localization

in the cell, etc.).

14  We Can Think About Ourselves – The Computer Cannot