I am a bit pessimistic about the limit of machine power. Yes, we observe huge growth in its capacities. It now can play chess better than champions. The core issue of machine is fast speed, higher than for human. Another component is the program written by human. The question is, whether this program is able to cover all peculiarities of language translation. Contrary to chess (with fixed rules), language is a complex evolving system. There may be some slang understandable for subset of people. Meaning can often depend on context.

I think that while future machines ca perform well in classical language (like standard Oxford English), they may under-perform in evolving language and slang.

Yes, I think it's possible.

While I agree that it's possible, I think the larger question is whether artificial intelligence would actually understand the original language or the translation. There I think the answer is no on both counts.

Yes, It is possible. It will not only translate but also able to understand and able to answer correctly.

Dear Adriana,

The rapidly developing advanced machine learning techniques makes it quite possible. Let us make the topic a little bit specify, for example, nowadays deep learning as a strong machine learning tool can be useful to make it true.

For example see:

https://medium.com/@ageitgey/machine-learning-is-fun-part-5-language-translation-with-deep-learning-and-the-magic-of-sequences-2ace0acca0aa

https://www.bibsonomy.org/bibtex/09104f020101005f73cbd1d380edda8d

So, not so far it would be possible translation complex text using by Deep learning.

Best,

Reza

yes,i think so

I am a bit pessimistic about the limit of machine power. Yes, we observe huge growth in its capacities. It now can play chess better than champions. The core issue of machine is fast speed, higher than for human. Another component is the program written by human. The question is, whether this program is able to cover all peculiarities of language translation. Contrary to chess (with fixed rules), language is a complex evolving system. There may be some slang understandable for subset of people. Meaning can often depend on context.

I think that while future machines ca perform well in classical language (like standard Oxford English), they may under-perform in evolving language and slang.

It depends on how far the "artificial intelligence" evolves in general. Furthermore, it depends on what are "complex sentences".

All in all, if the question is "will AI be able to accurately and fluently translate all domains/genres between any languages", my answer is "no".

If the question is "will AI be able to accurately and fluently translate some domains/genres between certain language pairs", then "yes".

It is already done, more or less, see e.g. Google Neural Machine Translation Model. A brief explanation of the model (including information on how it scores in comparison to human-quality translation) is available at: https://www.youtube.com/watch?v=HcStlHGpjN8&t=1597s, starts at 27:52. The speaker, Jeff Dean, is a Google Senior Fellow in the Google Brain team. Reza Mahini above provides a link to the paper in which the model is explained in detail, the title of the paper and its web address are also given in the video at 28:30.

Yes.

Yes, definitely...

https://newsclick.in/artificial-intelligence-learns-language-translation-without-human-interference

Hi. No. The question is similar to question "Will AI start being alive".

Velina (liking science fiction).

@Velina Slavova

RE: No. The question is similar to the question "Will AI start being alive?".

There is a debate over whether viruses (not the computer-kind) are alive. They have some of the features considered to be paradigmatic of life forms but lack others. * They straddle the boundary between biology and chemistry; or better, there is no sharp boundary.

Developers are already incorporating some small snippets of biological material into hardware components. ** Future AI would likely have neural networks built from organic/biological components cultured in a lab, thus "straddling" the boundary between biology and electronics (or nanobiology and nanoelectronics). The point is that there are spectra from the nonliving to the living and from the nonbiological to the biological, and there's no reason to believe that full-fledged life as we know it is necessary for engineering advanced translation processes. Something short of that may be sufficient.

So I say probably YES, because AI will in certain respects be similar to things that are alive: not mere hardware but hardware-cum-wetware.

___________

* https://www.khanacademy.org/test-prep/mcat/cells/viruses/a/are-viruses-dead-or-alive

https://www.sciencealert.com/are-viruses-alive

** https://newatlas.com/cyborg-biologically-powered-chip/40815/

Koniku corporation has already developed a prototype 64-neuron silicon chip: http://koniku.com

Hi Karl. OK. agreed.

https://www.theatlantic.com/science/archive/2018/01/brain-cells-can-share-information-using-a-gene-that-came-from-viruses/550403/

Article The Neuronal Gene Arc Encodes a Repurposed Retrotransposon G...

That's all... clear for now. I should wont to be the author of a deep learning system which will do this... Very deep learning, pardon.

The translation, in general, is possible because the Meaning of the announsment is clearly Understood by the Person who translates it.

Constructive idea: we should incorporate viruses in the machines in order to make them create meaning inside. So let us start from the semi-alive stuff, it seems (from the links) this is one of the information tranfer mechanisms.

The very deep learning has to start whith the Big Bang, I guess.

I go on thinking that - No.

Kind regards,

Velina

Certainly YES !

I strongly recommend reading this:

https://twitter.com/tarfandy/status/958461556135727104