Is Human-Level Artificial Intelligence a False Promise?

Wednesday, April 22, 2015

 Artificial Intelligence
Film analyst Rob Ager has looked at the subject of artificial intelligence and points out some reasons why the promise of human-level AI is an over-optimistic pursuit.

Rob Ager is a film maker and creator of the Collative Learning website.  Ager's analysis of film is always thoughtful and entertaining.  Now he has condensed a longer look at the subject of androids and artificial intelligence, and broken down what he considers are the ten biggest modern myths about the subject.

1. Sequential vs. Parallel Processing

Ager considers the human brain's ability to compute hundreds of thousands of parallel signals simultaneously gives an advantage to our organic processing abilities versus today's linear computer processors, despite their speed. The underlying lack of knowledge of the neuroscience behind our brain's abilities are at the core of this argument.

Could neuromorphic computing be an answer to this issue?  GPU's are massively parallel and have opened the door to neural networks and recent deep learning advances that are remarkable.

2. Visual System Differences.

According to Ager, computer vision is limited to processing the scan lines of an input from top left to bottom right.  He also points out that movement is analyzed by a computer in frames, which does not have an equivalent in the human brain. This is an oversimplification of the tremendous amount of research and progress that has been done in computer vision, especially over the last few years. Ager also again points out that we do not yet have a complete working knowledge of how our brain processes the massive amount of visual information it receives.

Is the human brain too complex to understand?

3. Memory Storage Differences

Despite Moore's Law, Ager points out that the human brain is much more flexible at memory storage than our present computer systems.  "Even a 90 year old person with a still healthy brain is still able to store new memories," but if your hard drive is full, that's it. Neuroscience hasn't yet described how the human memory system works.

4. Precision vs. Flexibility

Reproduction predefined actions is the core ability of computers, whreas the brain is incredibly flexible.  "It seems that human evolution has prioritized versatility over precision," says Ager. Science and technology follow the opposite strategy he suggests.

5. Misleading PR Projects

Ager suggests that the so-called triumphs of AI, such as Deep Blue beating Gary Kasparov, the Eugene Goostman chatbot beating the Turing Test, or Watson's win on Jeopardy! were actually limited by extremely limited sets of conditions on the test.  The chatbot, for instance, was programmed to be a Ukrainian teenage boy, which covered up a lot of it's failings in the English test.

Related articles

6. The Processing Power Misconception

This reason ties in with the first one in Ager's list. Increasing computing power, as is represented by Moore's Law may lead to faster signal processing, but Ager points that the trend may be unsustainable, and even if another technology like quantum computing comes along as a new paradigm, the human brain's parallel processing is still superior.

7. Similarity vs. Difference in How We Perceive Robots and Computers

"The desire to be pampered and served contributes to the general over-optimism of AI enthusiasm."

"Its great being mesmerized by what the new robots can do, because of our unfamiliarity with the robot, but don't forget to keep a balanced perception, taking into account what humans can do, and the robots still can't," suggest Ager. While some task specific robots are incredibly life-like, like the T8 Spiderbot,  they are not at all flexible in their operation yet.

8. Scientism

The historical perspective skews our perception. To claim that our AI efforts will produce fruitful results is part of what Ager points to as scientism—the belief that science can achieve anything we apply it to.  The belief is so strong that it approaches a religious tone. An analogy is the comparison to the Singularity as the "Rapture of the Nerds." Landing on the Moon, for instance led to expectations of the space program that were not reached.  Human level AI is another case.

9. The Slave Motive

Our underlying desire for slave domestic robots (and even robotic slave sexual partners) is a key force driving the pursuit of artificial intelligence according to Ager.  "Here were are in 2015, and we still take the garbage out ourselves," he sardonically quips.

robotics servant

"The desire to be pampered and served contributes to the general over-optimism of AI enthusiasm," Ager says.

10. The Immortality Delusion

Ager calls mind uploading a silly idea.  The promise of digital immortality is for some, the same thing that drives many into the arms of religious institutions he says. "The promise of an afterlife is a very tempting thing."

The criticisms Ager makes do not seem to take into account the most recent developments in machine intelligence or factor in the exponential nature of much of this progress. In fact, he hasn't provided sources for much of this video essay. The video is interesting, but perhaps Ager should stick to film commentary rather than subjects where he is clearly out of his depth.

What do you think?  Is Ager right, or are we closer to human-level artificial intelligence than we think?

SOURCE  Rob Ager

By 33rd SquareEmbed