Artificial Intelligence is trained to map the inputs and outputs in a specific manner. So if you train it to take left as right and right as left it will do so. If you want that in certain conditions it should take the input as it is and in other conditions in an opposite manner. Then you will have to add another parameter or parameters to the input data to characterize the tone, accent etc. Surely it will add to complexities.

Yes definitely. Artificial intelligence is nothing but guiding a machine by training, so whatever commands you impose on the machine, it will be prompt to your guidelines only without any extra possibilities unless you provide them more options.

Just change the Target of the training data to the opposite meaning...You are the one who put the meaning to whatever data you train.

Yes, especially if you used the right weight of a token within a corpus.

HTH.

Dr. Samer Sarsam

Yes definitely

Artificial Intelligence is trained to map the inputs and outputs in a specific manner. So if you train it to take left as right and right as left it will do so. If you want that in certain conditions it should take the input as it is and in other conditions in an opposite manner. Then you will have to add another parameter or parameters to the input data to characterize the tone, accent etc. Surely it will add to complexities.

It is essential to use more inclusive data and train the system to map on the reverse as well. It will definitely become very complicated and one has to make sure of the usefulness of training a system on such mappings.

Sarcasm is part of the context in which a sentence or words are understood by AI. Machines will always find context dependencies difficult, however, I don't think that it is impossible. With adequate training, AI will not only understand sarcasm but may even reply with one.

Do you think Artificial Intelligence can pick up on human term resonance?

If so then, yes, of course.

resonance[ˈrɛz(ə)nəns]NOUN

Thank you my colleagues about your excellent contributions and answers

It merely is about your definition of Artificial Intelligence. By my definition, we have two kinds of AI or Intelligent Systems so to speak. Systems that have an intelligence close and like Humans and learn and work almost like humans. And then we have systems that are more static and mostly work based on the Mathematical and Statistical rules and are not mimicked by Human Physiology. There are so many factors in training an AI that can influence the performance. The two main factors are Data and the method of training. If you have a good, solid and reliable data it is more likely for you to be able to train an AI to do what you are asking. Also it depends on your method of training which can be either Supervised or Unsupervised. In a Supervised Environment, the chances to train an AI to be able to find Antonyms of words are way higher than Unsupervised Environment. Generally the method that involves around words and analyzing them is called NLP(Natural Language Processing). So in my opinion, Yes, it can.

Human intelligence can be shallow, deep and medium. Similarly artificial intelligence. At present we should develop, in the first step, shallow AI. If it works well we should try to develop more advanced intelligence. The subject you ask about belongs rather to tasks of deep AI. Was my comment deep or shallow? Greetings-A

AI processes input numerically. In this case, we might have two different meaning of 'beautiful'. Let say we denote it by b_1 and b_2. We need to explicitly explain or tell the AI how to differentiate between these two beautifuls.

One of the aspect that might differentiate between b_1 and b_2 is intonation. Suppose we don't have any other variable to differentiate between b_1 and b_2. If there are no difference between the numerical values of intonation, than it will be the same. We can also have several values for each b_1 and b_2. From there, you can teach AI on sarcasm, and it will adapt by analyzing the similarities between the numerical values for intonation.

And I am certain the values will be full of uncertainties. You may resort to using fuzzy set theory.

Establishing the semantically ordering relation and incorporating fuzzy hedge algebra among the near (related) synonyms could be possible to overcome such complexities. Semantically ordering relation among the adjectives such as young and old, bad and good, etc., and moreover the pair of opposite adjectives. Also we should train the machine in such a way that it could recognize the contextual differences and contextual importance.

With AI is like with Chinese room, then we could not say that AI something understand. If you programming AI and want to honesty tthey set of notions, or differential notions you could add new notion in this case synonymous I hope in your AI will be to chose this synonymous or second synonymous like in dictionary T9.