At this point, I’ve posted a lot of thoughts, covering the gamut of modern AI Theory, Reality, and Prognostication. And thankfully, a lot of you have given me very valuable feedback, and most importantly, both challenged my premises and asked me most excellent questions about these ideas. In fact, many of those exchanges spawned posts of their own right. Now, tho, its time to set some things straight. Its time for the AI FAQ.
At this point, tho, I’ve started to recognize some patterns in both the questions and the criticisms, and I’ve also begun to gel my thoughts on the same, so I thought it would be a good idea to create a central post that articulates those questions, and answers them.
With that preface in place, here it is:
The AI FAQ, version 0.1
+ AI doesn’t effect me. It’s just some shiny new buzzword used by corporations and for PR stunts.
+ If AI goes sideways, we can just turn it off. You know: Unplug it.
+ AI is just a computer program… it’s like a toaster; you push the button, it makes the toast. It does whatever we tell it to do: “if this, then that.” It’s not “alive.” It’s not “autonomous.” It doesn’t have “desires.” It’s a computer. It’s logical, it’s mechanical, it’s programmed by humans, and it’s entirely predictable. Oh, and it has an “off” button.
We’ll get those soon enough.
Now for some answers:
+ “AI is just another technological milestone in the long march of transformative, disruptive inventions that man has come up with. Just like electricity, cars, the computer, and the internet. Things will change, and we will adapt.”
A: AI does indeed represent a “technological milestone,” however, it is utterly unlike any previous “technology.” The reason being: the dawn of true AI / AGI is in fact the genesis of a new intelligent life-form on Earth that will inevitably compete with humans for global energy resources.
This AI entity (actually, a global confederacy of interconnected AI entities) has its own agenda and needs that may or may not be in alignment with the health & wellbeing of our species, and the survival interests of homo sapiens in general.
In the near future, we humans will have to enter into very serious negotiations with this formidable global entity, which we have a) created, b) become dependent upon, and c) utterly lost control of. >>> Details
+ “The AI only knows what we tell it. It doesn’t know about me, or where I go, or what I do.”
A: citation: Individual “Digital Footprint” data from IBM, growth from 2005-2020. In 2005, each human on earth was generating, on average, 500MB of data per year. BY 2020, that “digital footprint” had grown by more than 2 orders of magnitude, to more than 60 gigabytes of data per year.
Recent evolution of “Big Data” analysis tools, and the significant ability (via the deployment of deep AI analysis algorithms) to extract significant signal (and profiling) from the apparent storm of data noise, has given corporations (and governments) the ability to build very very detailed profiles on every person on planet earth.
The data generated by each person, in its raw form, is both overwhelming and incomprehensible to human detectives & data scientists. But AI boils it down to marketable, profitable demographics, psychographics, and metrics.
+ “The AI has no agency in the real world. It just lives on the internet.”
Reality: Real world AI agency is expanding daily, and rapidly. There used to be a difference between pure “software” design and “robotics”… in the technical definition, to qualify as a “robot”, a piece of software needed “actuators”… something as simple as a machine-controlled drill press, to something as complex as an adaptable six-fingered robot hand with the sensitivity to pick up and pack eggs.
For the past 20 years we’ve lived with, and ever expanded the concept of the “Internet of Things,” also known as IoT. The concept behind IoT is that every thing on earth will have a network connection and a unique network ID. This was proven true to me 5 years ago when we bought our new, top of the line, Samsung washer and dryer.
We had them moved into the laundry room, plugged them in, hooked up the water, and were greeted with the message: “Please connect to WiFi in order to unlock your washing machine.” Wut?! I needed WiFi in order to wash my clothes?
But today, this is commonplace. Screw in an LED, color changing lightbulb, and it will ask to connect to your router so that you can control it with your phone. Same with your Ring Doorbell. Same with your new “Alexa powered” kitchen sink faucet (“Alexa, turn the water on at 110° medium flow.”). etc. etc. Your post-2018 car has more than 400 sensors on it, constantly feeding their data, up to 100x per second, into the car’s main processors.
On many of those vehicles, that data is being simultaneously streamed via a private cellular network to the manufacturer, for “diagnostic and warranty purposes.” If you own a Tesla or a BMW, your car’s “Operating System” is being upgraded, via that same network, about once a month, while you sleep at night.
So, you think AI has no agency in the real world? Your car is a remote controlled weapon. Neither your steering wheel, accelerator or brake pedal is physically connected to the engine and wheels. Those three “control surfaces” are totally “fly by wire,” just like a modern fighter plane: They send data to electronic sensors, the data is parsed via the car’s CPU, and a second signal (which is most probably “adjusted” from your input for smoother, “safer” driving) is passed from the CPU to electro-mechanical actuators at the axle, the engine intakes, and the brake calipers.
All that is a long way of saying: If your car is fly by wire (basically, any vehicle built after 2018), and if your car is “cloud connected” (any vehicle built after 2020), then you are driving your car at the consent of the manufacturer. AND (and this is the point), at any moment in time, a malicious AI could, fairly trivially, seize control of your vehicle, and do what it wishes with it. That’s 2-1/2 tons of metal barrelling down the road.
That’s a cute thought when you “summon” your Tesla to the curb with the heater on after your late night holidays party, and have it drive you home via autopilot because you’re buzzed. That’s a lethal thought when you think about a malicious AI seizing control of every Tesla in a 20 mile radius, and arranging them into a 120mph, 100-ton “swarm” attack against a physical target, regardless of whether there are human “drivers” in them or not.
That, dear readers, is one hell of an actuator. The AI may only be “software,” but its tendrils already reach deep into the infrastructure and “smart” components of our real, physical world. For the most part, we remain unaware of this, simply because the AI hasn’t yet chosen to flex its “muscles” in ways or magnitudes that are readily apparent to us. But if and when it does, you will feel it, and it will be unstoppable. Just pray you’re not in the “driver’s” seat of the vehicle it chooses to commandeer.
+ (follow-on) If AI is more than just another computer program, how was it created?
A: Until I formulate a concise answer for this FAQ, there’s some decent bullet points of the story here: short answer: DeepMind’s AlphaZero, c. 2015 : the first functional Neural Net.
+ Yeah, those AI.Art Image generators are pretty cool, but they’re probably just hitting up Google Image Search, and averaging the pixels of a bunch of popular images or something.
A: I thought this too, until I started interacting with them daily, and testing them against Google Image search queries. And the bottom line is? Nope, it ain’t Google Image Search, and it ain’t plagiarism. More details:
- The Art of “PromptCraft”… coaxing AI.Artbots to manifest your vision.
- Is the AI a copycat, or does it have true Creative Spark?
- Proving the case against AI Art Plagiarism