Could the CAPTCHA of the future distinguish a human from a machine?

The tests that are used usually are problems related to human abilities that we perform unconsciously, as identify images, sounds, understand the language, etc. Are things that are a difficult tasks to resolve by a computer while for us it is not nothing because our brain have been programmed by the evolution to perform it quickly, using for this a lot of resources of our complicated brain without that we realize it.

There are three types of tests that are usually used in the CAPTCHA:

1.- Images: It is the more habitual test. Consist in identify a serie of letters distorted o strikethrough, that we must try to read from an image generated by computer. This suppose an impediment to the persons with vision problems and even also to the people that don't have because a lot of times the letters can't be identified although we see perfectly.

2.- Audio: Trying to solve the problems generated to the prior CAPTCHA, sometimes with the image of the CAPTCHA we have as alternative a button to use an audio CAPTCHA, to identify the letters or the word that says a synthesized and distorted voice generated by computer. These CAPTCHA have the same problem than the previous, is an impediment for the person with hearing problems and a lot of times are equally frustrating to the persons that don't have it, because the synthesized voices have such poor quality that results impossible to understand what it saying.

3.- Questions: Finally we have the most accessible alternative, the one used to add comments in this blog and that consist in make to the user a simple question that have to answer.

But none of this methods is 100% effective to avoid the SPAM, because all of this methods are problems that can be solved making use of algorithms of artificial intelligence.

The spammers don't need invest in develop perfect systems of artificial intelligence to solve all the CAPTCHA because it is too much expensive and unnecessary. If the artificial intelligence program can surpass a little percentage of the tests, is enough for them. For instance, if the SPAM program make 100 attempts and the algorithm to surpass CAPTCHA pass a 10% of the tests, they will can sneak 10 SPAM messages.

This fact causes that the CAPTCHA every time have to be more complicated to avoid the spammers and the spammers have to exert every time more in perform more sophisticated artificial intelligence algorithms to pass it. The consequence of this fight is that will be every time more difficult to the users pass the CAPTCHA until arrive the moment that will be impossible solve it.

A possible solution would be find a kind of problem that a machine can't resolve and human can. But exists a kind of problem like that? I don't know any. Do you know some?