Artificial intelligence could exceed human intelligence
What keeps Stephen Hawking and Elon Musk up at night?
Artificial intelligence (AI) gone awry.
We’ve actually been studying AI for over 60 years. Famed computer scientist John McCarthy is credited with introducing the term “artificial intelligence” while at MIT in 1955. We still haven’t worked out what “real” intelligence is yet, which makes you wonder what the artificial version will look like.
For many years, researchers working in AI struggled with the challenges of making a machine “think” like a human. Programs were written to play chess, which became powerful enough to beat many of the world’s champions. But the programs often relied upon brute-force computing power. True intelligence – artificial or otherwise – always seemed elusive.
Recent advances in neural networks have changed all that. A neural network is software that mimics how the human brain works. It “learns” how to solve a problem by comparing many possible solutions, and then improves the solution with more data. These programs learn very rapidly, and quite soon they can solve problems much larger and more complicated than a human could.
The scary bit is that, over time, the neural pathways the programs act like a mysterious black box. The programs solve problems, but we’re not entirely sure how. Imagine if the programs could generate a new version of themselves, an improved version, that had an even more sophisticated neural network from the get-go, and it too grew in power. The growth via artificial evolution is exponential, which means the computational capabilities of the program would soon eclipse that of it’s makers – us.
This is what futurist Ray Kurzweil refers to as the “singularity”: the point when human-level intelligence is realized in a thinking machine. Several things have to happen before the singularity occurs, such as a complete understanding of how the brain works, but we’re well on our way to doing so. You can view an excellent video on the singularity by Mr. Kurzweil at https://goo.gl/WNqGFQ.
When the singularity occurs, people like Hawking and Musk fear that these machines will become uncontrollable, because they will evolve so fast that we’d be left wondering what happened. In neo-Darwinian terms, we’ll be victims of unnatural selection. And ironically, we’d be the engineers of our own demise.
The concerns over AI have become so prevalent that we now have academics with the title of “AI ethicist.” They discuss how to manage the coming revolution in thinking machines and their impact on society. When AI is coupled with the incredible advancements being made in robotics, we may soon enter a new post-human world.
Scott Tilley is a professor at the Florida Institute of Technology in Melbourne. Contact him at TechnologyToday@srtilley.com
Source: Artificial intelligence could exceed human intelligence
Via: Google Alerts for AI