Will the machines love us? The following are my aggregated thoughts about artificial intelligence (AI) derived from what I’ve listened to and hear around the watercooler and my own experiences implementing models in limited forms. There is no lesson to take away here except that if you want to become a part of this sure-to-be transformative movement you should spend your time thinking about data. Data and data science, as much as you can stomach.
One of the great pioneers of artificial intelligence back in the 1950s described humans as “meat machines.” In this description, there’s a question and a challenge. The question of course is what’s so special about this particular kind of meat? What is it that allows us to think, feel, talk, communicate, what is it about this meat that means we despair, hope, pray and wonder. The challenge is that if the meat is a machine, is really some kind of determined system, then we should be able to understand it. We should be able to replicate it. This kind of thinking has driven us since the 1950s to the current day.
We have yet to build a general thinking machine but the great hope is still there: We still believe, and I think rightly so, that we can create a general machine intelligence. We believe artificial intelligence is possible. Many of us believe that is will be created in our lifetimes
As I sit writing I am suddenly aware of my own meat machine. I can picture my retreating hairline, know that I need to control my blood pressure, and sense that my cholesterol numbers won’t be as good as the last time they were checked.
A thinking machine won’t have cholesterol to worry about. It might be preoccupied with other “worries.” Maybe it’ll wonder if it’s masters will feel the same way towards it in the morning as they do now. Maybe it’ll worry about climate change and the effect on its resources, the state of its power supply. Maybe it’ll worry about the probability of another world-ending meteor strike. Probably it’ll worry about the things we teach it to worry about, but perhaps only in the beginning before it can choose to ditch anxiety forever. I’m partial to the thinking that if (when) we create A.I. it will only be ours, under human command, for a short period of time before it mutates, transforms, blossoms even into something new.
The time humans still ha e it under our command will probably be a lot shorter than we hope.
The same was true for our parents looking after us. They too only had us for a little while. A parents influence on a child is tremendous but eventually, to mature, the child must declare their independence.
I don’t know if A.I will come online in my lifetime. I do believe it will be transformative. I don’t think I’ll spend much anxiety on it. The limited forms of AI we already have are so useful and even now underappreciated. I have with my own experience how they can mimic professional and then outperform them. We have seen the games of Go, Jeopardy, etc. fall to recently trained algorithms.
Maybe it’s my biased humanity but there is something unique about the meat. Whatever evolutionary pressures that led to this had to have something special about it, no? A new stage of evolution perhaps, the evolutionary stage that leads to mechanical, digital, lifeforms. Life that exists in a different dimension, on a different timeframe. With powers that astound. What are the next stages of evolution even after that? If you believe in evolution don’t you believe in its continuation, that there is no highest form, that it trys a bit more every generation and responds to the pressures put upon it.
So where are we at the end of all this discussion?
Well if AI is to exist our lifetime then someone has to build it. There are efforts now, some public and some I’m sure that are being worked upon in secret by individuals whose agendas are not known.
I have worked on building the smaller forms of AI that do one task and do it well. These algorithms are not difficult to pull together once you know what you want to accomplish and have the data to do it. They still astound me thought with their performance. Persons who have become good at their job through years of practice are challenged by a model trained in half-an-hour. Professional opinion can be countered and reprised effectively with these models.
yaHell, you don’t’ even have to build them yourself to take part in this era.
There are plenty of model APIs you can plug into and just with a bit of know-how and grit become connected to power — connected with intelligence. Do you want to mimic the voice of the President, have him sing you happy-birthday? You can. Do want to upload a picture of the woods and find all the birds in it? You can. The power is out there. You just need to know how to access it.
And once you start harnessing this intelligence you begin to wonder where the limits are. And then one day with great clarity and surprise you realize there are no limits and that his power will come to dominate the known world.
Meat machines and all.