Sunday, June 10, 2018

A Fully Automated Society - Engineering Paradise


If we look at the rhythm we are developing AI and robotics, AI experts anticipate realistically the first artificial general intelligence (AGI) will emerge sometime before 2040, give or take a few years.  That’s roughly 20 years from now.  According to the AGI Society, artificial general intelligence (AGI) is “a general-purpose intelligence comparable to that of a human mind.”  We already have, in 2018, some highly sophisticated AI who can do certain specific things with much better ability and skill than any single human.  Skillsets such as image-recognition, understanding language, figuring out how to win complex games (like Chinese Checkers, Go, and Chess), how to analyze huge amounts of data, driving, sift through legal documents to find appropriate precedents in legal history and so much more.  None of these replace an entire human worker.  They can however, make us much more productive at our jobs by doing things we find hard to do, faster, with less errors, and more efficiently.  Researchers and engineers are also constantly finding out ways to automate other tasks so that humans don’t need to do them anymore.  Our world is more and more integrated with narrow AI that are more capable than us at doing specific tasks.

For now, none of these artificial intelligences are AGI.  But with enough extremely competent narrow AI connected through the Internet, eventually all these thousands of specialist AIs can spontaneously form AGI. 

It’s easy to imagine.  Right now, the best intelligence to take on a task that has never been attempted before is a human being because we can be creative and find out the best way to find a solution to any challenge.  Some people are better at some challenges than others but let’s leave that aside for a minute.

If we project a little bit into the future where we have more highly efficient narrow AI in the market connected through the internet.  Suddenly, we encounter a challenge to which we need to find a solution.  Instead of finding a human expert to find a solution, we’d ask our front-liner:  the digital assistant.  Now, the digital assistant is incapable of finding a solution to the problem.  It is extremely good at being an assistant, and that is all.  It is a narrow AI.  It is better at receiving requests and finding resources for you than any human could.  The assistant could then search the Internet for narrow AI that can help solving the problem.  It could find thousands of them and “ask” them if they have a solution to the problem or a partial solution.  Those narrow AI are also exceedingly good at precisely what they were designed to do, from data analysis, to image recognition, pattern recognition, solving algorithms, searching historical records, doing math, and more.  They are also connected to their own network of narrow AI with which they work with regularly.  So, some of the narrow AI queried by the digital assistant can communicate and ask their narrow AI networked “colleagues” to do certain tasks.  This exchange and querying of solutions can go through multiple levels and at every step, AI that are better at their specific tasks are challenged to find solutions they were tailored specifically to resolve.  After a few seconds, or less, a solution to the original challenge can be produced to the digital assistant and fed back to us.  Perhaps several solutions may be provided based on the different types of narrow AI connected to that may have different approaches.
A process that takes hours to several weeks to accomplish for a human being subject matter expert has been completed within a heartbeat.

No actual AGI has been designed to do all this directly.  It was all done through connectivity between simpler technologies.

If we had a choice to put our lives in the hands of a human being or a robot with AI installed, and the AI was more skilled, not prone to bias, sleep, fatigue or sickness, and would run on solar power, would we depend on the human expert or the expert robot?  How about if it could do the job within a blink of an eye to boot?

I bet most people would trust the smart robot.

This paradigm shift is why experts are calling these developments the 4th Industrial Revolution.  Experts, like myself and many others are saying that we’re heading towards a time where AI and robotics will become the better choice to solve all our challenges.  We’ll want automated systems to take care of our food, our roads, our lives because they will be better at it.

This is good news.  As described In Section 2 on the Ministries, we want automated systems to handle most of the work that needs to get done anyway.  If we look at the economics of that, it means getting high quality products and services for minimal resource consumption and minimal energy costs.

This means it would be extremely easy to provide food, water, air, a home and security for everyone. 
Since automated systems will be more efficient at doing any type of work, people will be free to offer added value to society in any way they feel like.  Some people will still wish having the best of us in supervisory and decision-making positions.  Not because the best among us will do a better job than automated systems.  No, the reason why people will want to put citizen experts in power of the ministries, and vote them in, is to ensure the subjective values of the population are respected. 
For a while some people will still need to talk to other human beings when they need something.  Whether it is about getting some specific dining experience or dealing with the immigration processes.  It is natural for human beings to want to experience life with other human beings.  In fact, our health and wellbeing depend on us interacting with other people.[1]  Those interactions don’t need to be work-related though.  We can have those connected experiences through family, friends, social experiences and more beyond the limits of the client to employee relationships.

Even at that not-too-far-future time, some people will still want to interact with other people for certain things.  People are unlikely to be on permanent holidays as many will still want to donate their time to society and help communities in ways they really enjoy.

The natural engineer will still want to engineer.  The natural teacher will still want to teach.  The natural social worker will still want to help families out.  These will be people giving their time because they really want to do this in service of others, without any expectation of getting something in return.

The gift of automation is not to replace every single occupation.  It is about being able to cover all the bases while letting people choose to do what they wish to do.

Heck, even though I described an ideal way to govern and the election of the most qualified expert for each segment of a nation’s executive earlier on in this work, there will be a time when the people will want to vote for an AI to take up the leadership instead of a human being too.  This can be a choice when the time comes. 

Within less than 2 generations automation will allow humanity to be entirely free to do whatever their hearts desire.  This is entirely positive as incentives such as self-direction and a strong sense of social purpose are far more powerful than financial incentives.  Sam Glucksberg’s “Candle Problem” experiments have shown that people get best results if a task is completed without financial incentives and as little pressure as possible.  Under those conditions, problems are solved with more imagination and faster too.[2] Thus, in a future where people challenge themselves for fun, it’ll be, well, fun and not stressful as the consequence of failing a challenge will be insignificant with better, more efficient and ever vigilant AI and robots to take over and fix it for us.

Therefore, my suggestion for the ideal world where automation can do jobs better than humans is to let humans do what they want, keep stress levels as low as possible, and let people solve problems for fun.  Leave emergencies to automated systems and those few people passionate enough to help society alongside automated systems.  If people help in this way while stress-free, they can get compensated for their service, but they are really doing it for fun or to feel part of something bigger.
When thinking of the near future like this, I am constantly reminded of Star Trek’s Federation of Planets.

If you live as a member of the Federation in this science fiction show, you would do whatever you wanted to experience life.  In other words, the officers of the Enterprise in Star Trek:  The Next Generation don’t need to be on that spaceship.  They are not forced to be there either.  They do it for the adventure and because they are passionate about space exploration.  AI and robotic systems could explore the galaxy with much more efficiency than a spaceship full of inefficient people with wants and needs.  But that’s beside the point.  These people want to experience space exploration and discovering new life and new civilizations for themselves directly.

We can engineer such a future for ourselves (minus the starships for a while) if we so choose.
Suffice to say however, that even in the Federation, there is an economy of sorts as matter is not always abundant and people still find value in things and services.  The Earth, for example, has a finite quantity of resources.  So, even aiming for something amazing like an ideal science-fiction-like future, we still need to be fair and manage the balance of resources, take care of the environment and each other, lest we get in serious existential trouble.  We must keep everything in a ledger.  We just don’t need maintain the ledger ourselves.  Let the AI crunch the numbers!


[1] Emma M. Seppälä, PhD (August 2012)  Connect to thrive.  - https://www.psychologytoday.com/us/blog/feeling-it/201208/connect-thrive
[2] John Kounios and Mark Beeman (September 2015).  How incentives hinder innovation  - http://thepsychreport.com/books/how-incentives-hinder-innovation-creativity/

No comments:

Post a Comment