>

SUBHODEV DAS

SongSoptok | 1/15/2016 |





SUPERINTELLIGENCE: ARE WE DOOMED BY IT?

"The development of full artificial intelligence (AI) could spell the end of the human race," warns Stephen Hawking. Elon Musk fears that the development of AI may be the biggest existential threat humanity faces. Bill Gates urges people to beware of it.

AI or the field of study concerned with developing intelligent behavior of non-human kind, i.e., of machines and software, has seen explosive growth over the past several decades. The ever increasing power of computers and the rigorous scientific work by researchers have brought AI within the reach of common man. Strangely, in this dawn of AI, we hear such dire predictions about our future.

Dread that the abominations people create will become their masters, or their executioners, is hardly new. But voiced by a renowned cosmologist, a Silicon Valley entrepreneur and the founder of Microsoft—hardly Luddites—and set against the vast investment in AI by big firms like Google and Facebook, such fears have taken on new weight. With supercomputers in every pocket and robots looking down on every battlefield, just dismissing the premonitions as science fiction seems like self-deception. The question is how to worry wisely.

Today’s AI produces the semblance of intelligence through brute number-crunching force, without great interest in approximating how minds equip humans with autonomy, interests and desires. Computers do not yet have anything approaching the wide, fluid ability to infer, judge and decide that is associated with intelligence in the conventional human sense.

Yet AI is powerful enough to make a dramatic difference to human life. It can already enhance human endeavor by complementing what people can do. For example, a recent study has shown that an AI framework-based simulation modeling that understands and predicts the outcomes of treatment could reduce health care costs by over 50 percent while improving patient outcome by nearly 50 percent.

IBM’s Watson supercomputer is evaluating evidence-based cancer treatment options using analytics to help a physician consider all related texts, reference materials, prior cases, and latest knowledge in journals and medical literature when treating an illness. The analysis could help physicians determine the best options for diagnosis and treatment in a matter of seconds!

Because of the 'humanity gap' between artificial and a real human mind, a human working in conjunction with any AI machine will always be more powerful than AI working on its own. If, however, Stephen Hawking’s premonition comes to truth: “It (AI) would take off on its own, and re-design itself at an ever-increasing rate.”

Many philosophers, futurists, and AI researchers have conjectured that human-level AI will be developed in the next 20 to 200 years. If these predictions are correct, it raises new and sinister issues related to our future in the age of intelligent machines. The power of atom laid dormant throughout history until humans unleashed it in 1945. Can the current century be the age of “intelligence explosion?”

In “Super intelligence: Paths, Dangers, Strategies” (Oxford University Press, July 2014), Nick Bostrom, a futurist at the University of Oxford, argues that if machine brains surpass human brains in general intelligence, then this new super intelligence could replace humans as the dominant life form on Earth. If that happens, it would be stealing a page from the book of human history.

We are already seeing Google’s Deep Mind and Microsoft’s Project Adam claiming to better human performance at image recognition tasks. Unlike human brain, the AI machines have no physical limitation. These can be as fast – a 604 GHz microprocessor has been reported by the University of Illinois recently, while neurons fire at 200 Hz – and as vast – a computer can be warehouse-sized but a human brain is enclosed in the cranium.

The critical threshold beyond which “intelligence explosion” or singularity happens is still a theoretical conjecture and is yet to be confirmed experimentally. There are great uncertainties about technological progresses, particularly hardware and software bottlenecks to achieve the “full” AI capabilities. However, it is undeniable that the future after the creation of smarter-than-human intelligence is absolutely unpredictable.

While specific predictions regarding the consequences of super intelligent AI vary from potential economic hardship to the complete extinction of humankind, many futurists like Nick Bostrom agree that the issue is of utmost importance and needs to be seriously addressed.

The hysteria around super intelligent AI or artificial super intelligence may be overblown, according to most serious AI researchers. Those who have dabbled in the AI technologies can appreciate the enormity of the task in ushering a super intelligence revolution. As a practitioner of general (as opposed to friendly) AI, I’m yet to see the first glimmers of monstrous sentience gestating in today's code. Perhaps Gates, Hawking and Musk, by virtue of being incredibly smart guys, know something that I don't.

We certainly don’t know how the neural processing speed or the size of the brain is related to intelligence in a qualitative sense. Mammals like elephants and whales have much bigger brain than humans. However, their intelligence may be limited when compared to that of humans. We do not know the relation in brain size to intelligence across animals, because we have no useful measure or even definition of intelligence across animals. And these quantities certainly do not seem to be particularly related to differences in intelligence between people.

Bostrom claims that once we have a machine with the intelligence of a man, super intelligence will be achieved just by making the machine faster and bigger. However, all running faster does is to save time. If there are two machines A and B and B runs ten times as fast as A, then A can do anything that B can do if one is willing to wait ten times as long. Similarly, his claim that a large gain in intelligence would necessarily entail a correspondingly large increase in power seems far-fetched.

Some AI researchers argue that seemingly super intelligent systems may have limited autonomy. IBM’s Deep Blue could beat the world chess champion but it may not share the same level of intelligence as humans. Soon, airplane auto-pilots and self-driving systems for cars will be more reliable than human pilots and drivers. Does that mean they are more intelligent than people? In a very narrow way, these systems are “more intelligent” than people, but their expertise applies to a very narrow domain, and they have very little autonomy. They can’t really go beyond the task they were designed to perform.

The difference between panic and caution is a matter of degree. So, random, unsupported comments – yes, even from Bill Gates – can do more harm to the psyche of the common masses when the intention is to urge caution.  Certainly a general artificial intelligence is potentially dangerous; and once we get anywhere close to it, we should use common sense to make sure that it doesn’t get out of hand.

The programs that have potentially far-reaching capabilities, such as those controlling the power grids or the nuclear bombs, should be conventionally designed whose behavior is very well understood. They should be protected from subversion by AI’s; but they have to be protected from human sabotage anyway, and the issues of protection are not very different. A machine should have an accessible “off” switch; and in the case of a computer or robot that might have any tendency toward self-preservation, there should be an off switch too that it cannot block.

Even so, one might reasonably argue that the dangers involved are so great that we should not risk building a computer with anything close to human intelligence. Something can always go wrong, or some foolish or malicious person might create super intelligence with no moral sense and with control of its own off switch. I certainly have no objection to imposing restrictions that would halt AI research far short of human intelligence.

It is certainly worth discussing what should be done in that direction. However, Bostrom’s claim that we have to accept those quasi-omnipotent super intelligences are part of our future, and that our task is to find a way to make sure that they guide themselves to moral principles beyond the understanding of our puny intellects, does not seem to me a helpful contribution to that discussion.

[SUBHODEV DAS]

Comments
0 Comments

No comments:

Blogger Widgets
Powered by Blogger.