202

2016

14.3

2016

(https://www.nature.com/

nature/journal/v529/n7587/full/nature16961.html

More generally, the strength of artificial intelligence programs is based on partially

emulating biology so that autonomous learning is possible, e.g. neural networks or through

hidden Markov models (Maccorduck 2004). Such strategies are used in bioinformatics, for

example, for genome annotation (exon-intron domain; e.g., GenScan program) or for the

prediction of domains (e.g., Pfam, SMART databases), signal proteins (e.g., the SignalP

program), and membrane regions in proteins (e.g., the TMHMM program; Käll et  al.

2004) (Chap. 3). For complex optimization problems, such as in protein folding, artificial

evolution by genetic algorithms is also used. Evolutionary strategies are another important

method of artificial intelligence. The more efficiently learning is replicated, the closer one

gets to artificial intelligence. Deep learning seems to bring a new quality to it. We are cur­

rently using this for image recognition, for example, of microscopic images.

In general, we can say that artificial intelligence is very good at simulating the recogni­

tion of complex features (technical term: feature extraction; pattern recognition). To do

14  We Can Think About Ourselves – The Computer Cannot