Sunday, April 19, 2015

Deep learning AI - current dangers

This article about deep learning AI scientist Geoffrey Hinton mentions expert opinions that doomsday scenarios of intelligent AI being destroying humanity scenario are more than 40 years away at the least, but focus on such doomsday scenarios are distracting attention away from other current dangers like the following.

"The National Security Agency in the U.S. has a huge amount of data at its fingertips. It would be shocking if it wasn’t using neural networks to make sense of it. The U.S. Department of Defence continues to fund AI research: how much autonomy can we as a society comfortably transfer to intelligent drones or robots? Appropriate boundaries for lethal autonomous weapons systems are an ongoing international debate. And if you’re already uncomfortable with ads that pick up keywords from your Facebook posts and email correspondence, you might not look forward to those systems getting smarter.

Then there’s the job question. Traditional computing replaced many menial tasks; neural nets are adept at navigating deep reservoirs of knowledge. Startups such as San Francisco-based Enlitic believe that deep learning algorithms can do a better, faster job of reading medical scans than the best-trained human beings."

  1. In June 2013, in this article, I expressed concern about the how most citizens are not aware about the technology capabilities available to their governments and how the secrecy about governmental technology capability development efforts reduces the control of citizens over ways and means governments can use to repress them if the governments decide to act despotically. Despots have an incentive to increase this secrecy to build technology to help them retain power.
  2. Degree of autonomous systems will increase and interactions between these autonomous systems will increase and humans might lose control/knowledge-of-control. I reflect on such scenarios in this article. I face one such scenario every morning when I walk through my living room littered with toys by my twin daughters earlier night. I never know which action by me will trigger unpredictable series of interconnected actions/reactions (of sound, lights, motion) between toys and have to be careful to avoid waking them up. Rogue software can complicate these scenarios as illustrated in this article.
  3. In this article, I talk about how it is important to regulate correctly to give citizens control over their data using correct technology, (something that is being done through data broker regulations and consumer privacy regulations being considered by USA) so that the scenario of leveraging ever advanced technology to process this data using deep learning AI will become illegal. Of course if this is not done or not done correctly, freedoms of citizens will be at risk. Their thoughts, words and actions will become vulnerable to influence from those mining their data and planting advertisements and/or other content.
  4. In this article, I talk about how the quickly emerging scenario of large scale job destruction due to technology is dangerous because our social sciences (including economics) are not able to predict the kind of society and state we will soon have as this happens. I worry that this might be intentional.
As deep learning AI technology in its various forms and scale accelerates, we will become more and more of a knowledge based society. Generating, leveraging and protecting intellectual property (algorithms and data) is already quite important and over time will become the most important differentiator between success and failure. Intellectual property has already become the organizing principle of humanity. Due to its nature, humanity will need to organise its affairs differently than at present to survive and thrive. This article talks about issues and challenges in protecting intellectual property and the implications for vulnerability of humanity. This article talks about whether ALL intellectual property should be allowed to be private irrespective of its nature.

When I wrote all these articles, I expected the issues they talk about to become important soon. So I am not surprised to find that they are becoming important. My only concern is that we are not ready to handle these issues. We need to do much more and soon.


Pratap Tambay

19 April 2015
Post a Comment