I wonder if the folks living through the beginning of the industrial revolution understood the dramatic impact mechanization and automation would have on their daily lives. It probably felt gradual. First it was cheaper to make good iron. Next, the amazing new steam engine caught their fancy. It would have felt slow at first. Marvelous inventions that promised increased prosperity and reduced human effort. In the long term this was certainly true, but only after a major upheaval of civilization.
I don’t think it’s far fetched or even uncommon to think that the intelligence revolution will be just as promising and problematic for humanity.1 I write this as someone that both loves and loathes technology. I’m also writing this as a father of a young child that worries about the world she is inheriting. I take great pains to keep from being a kook. I think the line between paranoid and prepared is very blurry and the deep experience required to actually understand artificial intelligence makes it hard to know which side I’m on.
I’ve included a large number of links in this article. It’s not a diatribe or dystopian fantasy. I consider it my research journal on a topic that I think is the greatest change in civilization since the industrial revolution. You and I have the opportunity to see it happen first hand. I’m wondering out loud if we recognize what we are seeing. You can find a list of all the linked references at the end of this article. Some are web articles, some are videos, and others are entire books.
Privacy and intimacy, as our parents knew it, is disappearing. Machine learning is transforming the ability of outsiders to pierce the veil of our personal relationships. This is a topic I care a great deal about because I believe privacy is vital to our self identity. There’s plenty written about the value of privacy so I don’t want to go on about that topic now. While Mark Zuckerberg may still think we no longer need privacy I think he’s probably changed his mind since getting married, having kids, and being investigated by congress.
I do not live under the illusion that we maintain our privacy simply by living mostly offline. Algorithms for mind reading are getting better every month with Goliaths like Facebook, Google, and even Apple driven to convert facial expressions into data. These technologies are hitting the part of the growth curve where new unimagined powers are developed faster than we can care.
We now have algorithms that can determine gender or sexual orientation just from looking at images of us. Imagine an uncle sitting down in front of Google Photos and saying “show me everyone in my family that might be gay.” Or how about an insurance company doing the same? Or a bar owner.
Running in parallel to all of this artificial intelligence innovation is dramatic improvements with image capture. Resolution is going up and physical size (of both the camera and the image file) is going down. This means we are developing an ever present array of cameras that capture every moment of our lives coupled with an artificial intelligence to draw conclusions and make decisions. I don’t believe that we will see regulations or controls over this technology in my lifetime. China has already begun deploying these technologies within their own public spaces and Russia is likely already do so as well. This is beyond the surveillance model we’ve previously seen, where humans were both the bottle neck and the decision maker. Now computers aggregate our social media, purchasing histories, and family relationships to create a profile. With the rapid development of image recognition algorithms, computers also determine meta attributes based on photo and video analysis to map further sociological conclusions about us.
They never even asked me any questions.
“It comes with colossal opportunities, but also threats that are difficult to predict. Whoever becomes the leader in this sphere will become the ruler of the world.”
After the recent U.S. election, I think he’s a man to listen to when it comes to the consequences of algorithms on the geopolitical landscape. The calculated application of disinformation is sinister but it is also an age old technique for destabilizing a culture. With so much attention focused on social media such as Facebook and Twitter we have large swaths of the public accessible from outside our own borders. Even worse, we have Facebook, Google and Twitter helping to algorithmically target the weakest links in society. The very networks we use are the tools of disinformation. There’s no distinction.
I don’t believe that A.I. Is inherently bad. I do believe that people bring their own biases with them to any project and those can be bad. The fear I have is that the first government to achieve the singularity will instantly become a new kind of superpower.
The stock market is an obvious target for any reasonable powerful A.I. Simply crashing the market would not be very useful in a global economy, but to subtly and undetectably manipulate the market in ways that weaken economies would be a powerful tool. We’ve already seen the perils of software based high speed trading. Now imagine a superintelligence whose mission is not to just make money.
While a percentage of people worry about their privacy because “they have something to hide”, I think it’s more concerning that we can be influenced to act against our own interests while believing that we aren’t. The infamous Target example, where they collected data to determine who was pregnant and what they could be offered to get them into their stores shows two things: It’s simple to do and very little effort is required to convince people they aren’t being gamed. It’s easy to pass this off as just par for the course with advertising. It’s not. A mountain of data is exchanged between a many different of eager partners to determine the exact right way to influence an individual.
From a Target spokesperson:
we found out that as long as a pregnant woman thinks she hasn’t been spied on, she’ll use the coupons. She just assumes that everyone else on her block got the same mailer for diapers and cribs. As long as we don’t spook her, it works.
Algorithms can certainly help us make better decisions, but if they are trained to take advantage of our insecurities and push us to choices that conflict with our interests, we’re in trouble. Now imagine an algorithm of this sort with virtually infinite compute power. Quantum computing may be the leap into artificial intelligence that people like Elon Musk worry about. It’s easy to write these “kooks” off as simple fans of the Terminator franchise when they alarm over world war, but it’s worth considering their achievements and what they could gain from such statements. For me, I’ll trust Musk long before Zuckerberg, who has already shown that he doesn’t understand how even Facebook works.
Elon Musk just a month ago:
China, Russia, soon all countries w strong computer science. Competition for AI superiority at national level most likely cause of WW3 imo,
The machine learning revolution will not be all sinister. It may shockingly beneficial to society. The effect of deep learning on medicine could bring about a new age in human health. We’re starting to glimpse a future where we can accurately model the atomic world. This may not (and I believe it will not) lead directly to new medicines, the power it provides to scientists could accelerate the invention of new, safer, therapies.
There will be social costs for even the most beneficial applications of artificial intelligence: job loss. As our society passed from the enlightenment to the industrial revolution and finally through the information age, our jobs became more cerebral and less manual. Here we are at the start of the twenty-first century and much of our prosperity comes from gathering and applying information. What do we do when algorithms are vastly superior at these jobs?
The field of Radiology is currently seeing the impact of machine learning. It started as enhancements for human specialists. In a short time professional radiologists trained algorithms to do their jobs more efficiently and more accurately. The debate is still on about how much of an impact A.I. will have on the future for Radiologists. Some think it will simply make humans better at their jobs:
When, and if, the AUC of an ML algorithm exceeds that of a radiologist, the machine will have learned more than what we now know and will be able to make decisions we cannot now make. At that point, we may be compelled to incorporate that algorithm into our practice. Incorporating these analytical tools into our practice, however, does not mean replacing radiologists. Rather, they will complement our quite remarkable human skills
But I don’t think anyone doubts that there will be a dramatic change for what was once a lauded and highly lucrative human career:
Computers can also quickly digest complex data sets, the authors added. While even the most trained radiologists will have cognitive limitations, no such issues exist with machine learning.
Chockley and Emanuel also detailed how machine learning is already being used in computer-aided detection (CAD), showing that machines can work “as well as or better than” veteran radiologists.
Quantum computing will change everything and it’s coming faster than I would have guessed. It’s difficult to understand what exactly quantum computing is yet I find it fascinating. This is one of those areas where I am obliged to take the word of an expert because there’s no chance I will develop an expertise in the field. The short version I’ve come up with is that a quantum computer co-ops parallel universes to perform its calculations. I love that idea even if it’s not a complete explanation.
A 500 qubit computer will probably be achieved before I die. At that point no encryption we use today will mean anything.2 Quantum computing is real today. It’s not a fantasy. We only know what is being done in the commercial realm. We have no idea how far countries like China will take the technology until they’ve already broken through all modes of encryption. It’s compelling that China is already working on un-hackable quantum encryption for their communications.
These two parallel lines of work will likely coalesce into the singularity. Even if we fall short of achieving the singularity, super intelligence will still be a tipping point for our culture that few people are preparing for.
We’re monkeys in a tree that just figured out how to make fire. Eventually we’re going to set ourselves a blaze and then blame the flammable tree.
This front is being moved from several directions. Some are in the public sector but many (maybe most) are not. I think it’s expected that national intelligence agencies are deeply concerned and energized about artificial intelligence but I also think the biggest advancements will come from private corporations working in private, such as Amazon’s efforts with their Barcelona group, Facebook’s Montreal group, or even Apple’s own deep learning team. Google already declared their intention of “AI First” and I’m sure we will only see the tail end of their achivements through their consumer products.
There are major hurdles to overcame yet. Many of the biggest advancements in this field will only be apparent through head-scratchingly impressive new features or scary-accurate new advertisements. When we see these advances, they will already be old technology. That’s what worries we. By the time the public sees the truly great breakthroughs the power will already have shifted. By the time a company can purposefully manipulate the market it will be too late. I don’t think we are talking about a scenario where large corporations do battle with each other using similar weapons. On it’s surface that would be fine because they would neutralize each other. I think what we will see is one company suddenly becoming so massively dominate that within weeks their control of the free market is insurmountable. Google reached this point in a gradual and organic method. I don’t think that’s what the future will look like as we climb the exponential curve of computing power and machine learning.
The links below are the same links from above. It’s just easier to read as a list.