1) What is the biggest threat of advancing robots into the future?
First, I would tend to agree with Isaac Asimov that future robots will probably come with built-in safeguards to prevent any sort of rebellion. The danger will come when, as seems inevitable, their intelligence begins to exceed human intelligence. Then, as writers like Vernor Vinge have noted, we will be completely unable to predict what they will do, and their superior abilities will make the achievements of humanity less and less important. Further, though there is no reason to think that they will have any desire to enslave or oppress humans, such super-intellligent robots will inevitably come to dominate all human affairs, for the simple reason that, in any group or society, power naturally tends to flow to the most intelligent being in the room.

2) What kind of positive side effects do you think would come of advancing robots anyway?
Adopting a broader perspective, and not worrying about the petty concerns of humanity, one can say that more intelligent robots would be a boon to the future of the universe in general, since they would be capable of developing theories and technology that humans may not be able to develop. Even if humans were no longer in control, and felt like second-class citizens, their lives would probably be better because of their robots' accomplishments.

3) Do you think any more rules should be added to Asimov's three rules, now that we have the internet and more advancements?
This is actually a matter now being debated by specialists in robotics. One person I know about who has written about extending Asimov's laws is Roger Clarke, and I just found that his paper on the subject is now online. From a personal perspective, I would first agree with the later Asimov that something like his Zeroth Law is needed: if a robot faces a choice between saving one person and saving 100 people, the robot requires a law that would lead it to save the 100 people, even though it would violate the First Law by leading to the death of one person. Also, it always seemed logical to me that Asimov's Third Law should be followed by a Fourth Law like this: "A robot must protect the existence of other robots as long as such protection does not conflict with the First, Second, and Third Law." That is, we will want robots to protect human beings, but we would also want them to protect other robots, to feel a sort of altruism toward their own kind as well as our own species. Finally, Asimov always assumed that humans would always be humans, and robots would always be robots, and there would never be any problem in distinguishing between the two. But we now know that various sorts of cyborgs, combining organic and mechanical parts, are possible and perhaps inevitable. So, if we have a human who retains an organic brain but has that brain directly connected to a computer that assists in her thinking, should that person be considered a human, or a robot, in the context of the Three Laws? Would a robot have to rescue or obey such a being as a human? Given the choice between saving that being and saving a "real," 100-per cent human person, would the robot save the latter person, on the grounds that she is more human than the cyborg? I'm not sure how these issues could be resolved within the framework of Asimov's Laws, but perhaps they would need to be proceeded by what Clarke might term a Meta-Law like the following: "For the purposes of these Laws, any being primarily controlled by a human intelligence shall be considered human, and any being primarily controlled by a mechanical intelligence shall be considered a robot."

4) Should we build robots that are exactly like humans, or only ones with a specific task?
Actually, I think the original Star Wars movie provides a reasonable answer to that question. If robots are designed to perform tasks that do not involve a lot of interaction with human beings, like R2-D2, there is no reason to make them look like humans, and any sort of functional design would be fine. However, if robots are designed to regularly interact with people as servants or advisers, it would be better to give them a human form, like C-3PO. Whether we should extend this principle to make such robots look exactly like humans, so that they cannot be distinguished from humans, is another question, since people might be discomfited if they could not immediately be sure that their friend's companion is a human or a robot. In one story with a name I do not recall, all robots were required to have writing on their forehead that identified them as robots, to prevent this problem. Perhaps small children, who might be frightened even by something humanoid like C-3PO, might be best served by robots that looked exactly like humans, but in other cases, this would not seem to be necessary—a general human form should be sufficient to facilitate interaction with humans.

5) Should we create protection robots for the elderly or important people? Or would the risk be too high?
In fact, given that society faces the increasing problem of elderly people living alone who have no one there to help them in emergencies, robots might be the ideal solution: instead of something like Life Alert, an endangered person might signal her robot to come to her aid. Beyond such simple tasks, there is the broader question of whether a robot would ever have the far-ranging intelligence and imagination to function as an effective bodyguard, since members of the Secret Service, for example, must be constantly vigilant against every sort of possible assault on the President. Still, robots could also be provided with superior sense organs that would allow them to see or hear threats that humans could not detect. Perhaps truly important people could be best protected by a team of humans and robots. As for the "risk" involved in such robots, any piece of machinery might malfunction at any time, but as indicated I have little concern that such a robot might, for example, suddenly attack a human.

6) If so, should there be an emergency shut-down button in case of a malfunction?
From one perspective, if a protective robot did happen to suddenly get out of control and start a destructive rampage, an emergency switch to turn it off would be helpful. On the other hand, if a robot protector could be easily disabled, that would diminish its value as a protector, since an evildoer might find and employ that switch to leave his intended victim vulnerable to an attack. This is a question that might be addressed by asking another question: since human beings also "malfunction" sometimes and do things that we don't want them to do, should all humans have an emergency mechanism installed in their bodies that would, say, cause them to immediately become unconscious if the button is pressed? Again, there is a certain logic behind the idea, but some obvious problems as well. Overall, if we are building reliable robots to perform such functions, we should be able to build them so that there is no real danger of a ruinous problem that would require emergency measures.