Be Scared, and Fearless

Let me explain how a computer neural network works in a way that I hope will terrify you, if you believe only humans can exhibit intelligence. I am not going to explain how it works, I am going to explain how easy it is to use machine learning without knowing how it works.

In my day job, I have built several systems that depend on neural networks as a core function, to do things like recognize objects on video. It’s actually absurdly simple to do, and any reasonable software engineer on the planet could have a functional machine learning system in a week if they wanted to.

I’m not joking. Google the word Tensorflow and you can build one yourself, no more complicated than building a shed. You literally just follow instructions, as long as you have the data and know what you are looking for.

So you build these things, and train them to recognize patterns, and then you test and refine till when you show it a picture of a jar of pickles, it can tell you with a 98% certainty if it’s pickles, as opposed to some other class of things.

It can ID you as Republican. Or someone likely to buy a jar of pickles. From any of your data we can build models that classify you.

I can do this. You could to. So can the trolls.

There are several million of us on this planet who could build an ML to be used for good or evil at the current state of the tech, 2018. There are even graphical user interfaces to build these things and run them on your Google home or Alexa. Making training AI into a common feature of everything will turn the several millions into billions who will each act as a force multiplier in the disruption.

Given a few more years it is likely these large companies will simply spin up helpful tools as they interact with us, crafted to us. Software will be an extension of us, building more and more refined models of us to give us what it thinks we want.

The gap between people who use Machine Learning to gain an advantage over those who do is vast and getting vaster. But that will change soon, and as the internet brought constantly and rapidly evolutionary change, the rise of machine learning has barely begun – once non programmers build the models, ad hoc, based on their own constantly recorded experience, it will change reality for the species.

It will work this way: your phone will record all your actions and words, as it does to some degree today. It will also monitor your internal states in numerous other ways.

When it detects something is amiss, like talking to a certain person causes you problems, it will alert you to this. Will keep you informed of invisible anomalies in the world.

Not your friend. We have people for that. AI will be far closer to a new sense, and my guess is that is how we will integrate it. The new sense will be available to all our current senses.

The model of you is just a data pipeline problem.

The models based on you already exist. But if you defensively build your own model of you, then you will own you. You can get paid for companies to access your model and query it. We could actually get paid to pay attention to an ad.

You can be assured that your robot will always protect you and your circle. But only if you can trust the model.

When we all have these models, when the crafting and curating of them becomes the primary day to day experience of hundreds of millions, I predict that new forms of government will arise, based on the immediate communication between our models on behalf of us.

Because people know what we want. We just don’t have the right tools. It is time for the people of the planet to build the models of ourselves, and for good companies to build open source tools that let humans do this.

For robots to not be our enemies, they must become our protectors. And all other forms of robots without human operators should be seen as dangerous, poorly implemented tools.

Chainsaws running untended on the schoolyard.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.