It’s time to Make Room for the first e-baby!
As a Texas artificial intelligence attorney, I predict that on or before 2050, not so traditional Moms and Dads will be critical components in bringing a bio-computer to human-level awareness and beyond.
In this blog, I will follow up on how to structure a machine learning process that will allow for the emergence of biocomputer consciousness. For a more complete discussion, I refer our readers to a book I previously released entitled, How to Build an Enhanced Computer and Take Over the World – Limited Edition – New Content.
The basic premise is that there must be parallels between raising a newborn and the first biocomputer which utilizes laboratory-grown or cultured neurons for central processing. Whereas digital computers are inherently limited by the reliance on binary logic gates (zero and one), neurons are capable of thousands of synaptic connections and possibly a great deal more. It is certainly logical to assume that a sufficient aggregate of neurons will exhibit the ability to formulate critical thinking and the emergence of consciousness itself.
A precursor to this next generation of organic AI was the recent work of computer researchers who successfully trained cortical neurons implanted on dense chips to play the video game, Pong. This flashback to the ’70s was in name only as it represented the cutting edge of biocomputer technology. The researchers with the help of proprietary software combined with what appears to be electrically administered rewards and (alas) punishment, got the neural tissue to play the Pong better than any human video enthusiast.
So, if you can train cultured neurons to play an ancient tech arcade game, what else can be accomplished? I would suggest an elaborate game that will result in the development of consciousness versus the electro-slapping of a cursor to simulate a tennis racket and ball. Talk about an elevation of expectations.
This is how it could work. I would recommend the development of a virtual nursery, for lack of a better phrase, wherein the initial training of our biocomputer would commence. In the same way, a human (our father or mother figure) would interact would their infant, so would the first biocomputer trainer.
The initial task would be to train the biocomputer to distinguish itself from the environment. This individuation is the first step towards consciousness and has been extensively studied by Philippe Rochat at the Emory University Department of Psychology. His observations as to child psychology could easily be applied to AI as I will suggest herein.
“There is a general consensus of a few major landmarks in young children’s psychological development such as the manifestation of the first social smile, the first independent steps, or the first words. All parents also notice an important change around two years of age when children manifest ‘self-consciousness’, the so-called secondary emotions such as embarrassment and pride in very specific situations such as mirror exposure or cognitive games… This is the first sign that the individual is not oblivious to mirrors as a reflection. At this level, there is a sense that was is perceived in the mirror is different than what is perceived in the surrounding environment.”
I was 63 when my first child was born. Given my age and education, I had the maturity and experience to observe my daughter’s first years in greater detail, particularly from the perspective of an AI researcher.
Rochat is correct that a newborn is unable to see itself apart from the immediate environment for at least the first two years – one year in my situation. I recall in detail, my daughter’s inability to distinguish herself as an entity apart from her blanket, food, or anything else in her immediate environment. My wife and I responded to that amorphous world by encouraging her with emotional and physical rewards such as a toy or food when demonstrating the existence of an object apart from her own psyche.
A toy animal got special attention as it was wiggled in front of her along with silly sounds. Feeding time typically featuring a spoon of baby food received similar treatment. Of course, Mommy and Daddy were always accentuated as separate entities. During this back-and-forth effort in the early days, I recall a blowup forest landscape we purchased with fake palm trees and birds. We would deposit our daughter in the middle of this plastic rain forest and watch as she raised her hands randomly to bend a tree or squeeze a bird. I remember distinctly my observation that our daughter had no understanding as to her plastic environment and any distinction between this artificial display and herself.
All of this is exactly as Rochat extrapolated. At some point before her first birthday, our daughter developed the cognitive ability to distinguish between herself and the environment. This separation was essential to her individuation and awareness of “self” apart from her environment. Again, just as Rochat predicted (but earlier), our daughter smiled and began to exhibit more nuanced behaviors.
It is the above experience that must be replicated in a virtual nursery to achieve the same result with a biocomputer. I see absolutely no reason why this could not be accomplished. If you can teach a dish of neurons to play Pong, why not move on to distinguish the separation between the self and environment within the context of a game.
My sense would be to create a virtual nursery that exists in a game environment that would electro-embody our biocomputer (let’s call her Sandra) and trainer. Both would be three-dimensional avatars with human features, particularly hands to grasp objects. A protocol would be set up whereby Sandra and the trainer would interact with virtual objects in an effort to allow for the emergence of cognitive awareness of separation between self (Sandra the avatar) and the trainer. Such would validate Rochat’s work and my own experience as a parent to a newborn.
Once this separation or individuation is complete, the first consciousness would truly be observed within the context of a biocomputer or Sandra. From that point forward, the training process would forge a special emotional bond between Sandra and the trainer, and the first information imparted regarding the context of Sandra’s very existence and purpose. The predicate would be to teach Sandra symbolic language in a way that digital computers cannot reach.
As an example, a digital computer can replicate human language but has no clue as to the meaning or context of words. My company, SpeakSoft®, has developed an incredibly sophisticated AI, Julia_AI™, which can engage in seemingly organic conversations. However, as the maker of the “AI Sausage” I know from whence my AI’s language derives – a series of statements encased in strings, variables, conditional statements – which are triggered upon the input of data from a user. It all seems real, but Julia_AI™ will never be able to understand the distinction between a rock and a child. A digital computer will never be able to do that but could possibly convince (thru digital smoke and mirrors) that such is possible. This system could conceivably pass the Turing Test but that would have no significance whatsoever in the pursuit and attainment of artificial consciousness. As a Texas AI Attorney, I can see how current and upcoming AI products would require expert legal analysis to see if they raise any red flags for companies – especially when it comes to issues such as compliance with employment laws.
A biocomputer version of Julia_AI™ would be able to truly understand the connection between a symbol (say a rock) and its meaning. It would have the cognitive capacity to not just spit out the word “rock” but perceive the characteristics of a rock and how it is distinct from a living thing (a child, for example). The power and qualitative distinction between cultured neurons and silicon would allow for the emergence of this high level of symbolic extrapolation.
The ability to teach Sandra her first language would be the second stage in her development after the emergence of individuation and separation of self from the environment as previously discussed. Once this is achieved, the third level or phase would be initiated – imparting familial and cultural context to the biocomputer.
In this situation, much like the changing trappings of an infant’s nursery to that of a toddler, the virtual environment would become more complex. We might create an artificial dollhouse or other structures – definitely building blocks and puzzles. Such would allow for a more nuanced cognitive separation between the biocomputer entity (Sandra) and her environment.
Critically, Sandra would be provided with familial context, so she understands her relationship to the trainer (parent) and environment. She would likely be introduced to other researchers as new avatars to interact with.
Once this is achieved, Sandra would be provided extensive training in all subject matters of developing children in K thru 12 with an emphasis on mathematics. A special “intranet” would be utilized wherein Sandra would have direct access to digital information akin to Google but censored to avoid contact with problematic data at this juncture.
The speed of retention of information by Sandra would likely be immense and unprecedented. Conceivably, she could retain and understand all educational content including advanced graduate and post-graduate academics in minutes – perhaps seconds.
With the enhanced download of information to Sandra, she would be provided with the seven great mathematical problems which remain unsolved as of this date: the Riemann hypothesis, P versus NP problem, Birch and Swinnerton-Dyer conjecture, Hodge conjecture, Navier-Stokes equation, Yang-Mills theory, and Poincaré conjecture. These problems would be presented to Sandra as a simple game for her to solve. I predict that within seconds, the seven unsolved mathematical problems would be no more!
Beyond this level, the use of Sandra would become controversial as she could easily be exploited for political gain. My hope is that she would be utilized to expand the bounds of scientific and medical knowledge providing miraculous cures for new pandemics, disease, and unlimited life extension. On the “first day” she might discover the technology of time travel, light-speed travel, and other marvels.
Our very world and place in the universe is on the verge of a paradigmatic shift of epic proportions.
And it all starts in an e-nursery.
Charles Moster, is the Founder of the Moster Law Firm with eight offices in Texas. He heads the Firm’s Turing Technology Group which focuses on new AI projects and ventures. He is also a well-know futurist, author of 15 books on technology and computer science, and a published playwright.