Artificial Intelligence is tomorrow’s electricity: Future of artificial intelligence P1
Artificial Intelligence is tomorrow’s electricity: Future of artificial intelligence P1
Human evolution takes a giant leap forward every time we gain control of a profoundly new source of elemental power. And believe it or not, we’re close to our next great leap.
Our ancestors looked much like today's modern ape—a relatively smaller skull, larger teeth and a far stronger jaw to chew through the pounds of raw plants that our oversized bellies would spend hours-to-days digesting. But then we discovered fire.
After exploring the remains of forest fires, our ancestors found charred animal carcasses that upon closer inspection … smelled good. Cutting them open was easy. The flesh was more flavorful and effortless to chew. And best of all, this cooked meat digested quickly and more of its nutrients absorbed into the body. Our ancestors got hooked.
After learning to tame fire and use it to cook meals, the generations that followed saw incremental changes in their bodies. Their jaws and teeth grew smaller since they didn't need to chew endlessly through tough, raw plants and flesh. Their intestines (bellies) grew smaller because cooked food was far easier to digest. And the increased absorption of nutrients from cooked meat, and arguably our newfound need to hunt our food, helped spur the development of our brains and minds.
Millennia later humanity gained control of electricity, sparking the Industrial Revolution in 1760 and leading to our modern day. And here too, our bodies are changing.
We’re living longer. We’re growing taller. Our ballooning population is interbreeding to create far more variations of humanity. And as we master the tech behind genetic engineering by the mid-2040s, humanity will gain the ability to directly influence its physical evolution at a much faster clip. (Feel free to read more in our Future of Human Evolution series.)
But by the early 2030s, humanity will realize a new power: true artificial intelligence (AI).
The rise of the personal computer and the internet have given us an early taste of how access to an augmented intelligence (basic computational power) can change our world. But in this six-part series, we're talking about truly limitless intelligence, the kind that learns on its own, takes action on its own, a magnitude of intelligence that can liberate or enslave the whole of humanity.
This is going to be fun.
Clearing up the confusion around artificial intelligence
Putting aside the overly dramatic opening, let's get real about AI. For most people, AI is just a really confusing topic. A big part of that confusion comes from its sloppy use across pop culture, the press, and even in academia. A few points:
1. R2-D2. The Terminator. Data from Star Trek: TNG. Ava from Ex Machina. Whether portrayed in the positive or negative, the range of fictional AI blurs the public's understanding of what AI really is and its potential. That said, they are useful as educational references. That's why for the sake of conversation, throughout this series, we'll namedrop these (and more) fictional AIs when explaining the different levels of AI that exist today and will be created tomorrow.
2. Whether it's your Apple smartwatch or your autonomous Tesla, your Amazon Echo or your Google Mini, these days, we're surrounded by AI. But because it's become so common, it has also become entirely invisible to us, much like the utilities we depend on, like electricity and water. As humans, we’re susceptible to a range of cognitive biases, meaning this increasingly commonplace AI is pushing us to redefine our concept of ‘real’ AI to become far more mythical than realistic.
3. On the academic side, neuroscientists, biologists, psychologists, et al., the professionals most concerned with the brain and mind still don't have a complete understanding of how the brain works. Without this understanding, science can't effectively identify whether an AI is or isn't sentient (alive).
4. Putting all this together, our pop culture, our science, and our human biases are skewing the way we think of AI from the get-go. As humans, we naturally tend to understand new concepts by comparing them to things we already know. We try to understand AI by anthropomorphizing them, attributing to them human personalities and forms, kind of like Amazon Alexa's female voice. Likewise, our instinct is to think of a true AI mind as one that will work and think just like our own. Well, that won't be how it plays out.
The thing to remember is that the human mind, alongside all the animals and insects we share this planet with, represents a form of evolved intelligence (EI). How we think is a direct result of two factors: millennia of evolution that shaped our base instincts and the sensory organs (vision, smell, touch, etc.) our brains use to collect information.
The AI we create won't have these hang-ups.
Current and future AI won’t run on vague instincts or emotions but defined goals. AI won’t have a handful of sensory organs; instead, depending on their scale, they’ll have access to dozens, hundreds, thousands, even billions of individual sensors all feeding them reams of real-time data.
To sum up, we have to start thinking of AI less as machines and more like aliens—entities that are entirely different than ourselves.
With this in mind, let's shift gears and focus on the different levels of AI currently in the pipeline. For this series, we'll highlight the three levels commonly discussed by most AI experts.
What is an artificial narrow intelligence?
Sometimes called "weak AI," artificial narrow intelligence (ANI) is AI that specializes in one field or task. It perceives and then acts on its environment/situation directly without a concept of the wider world.
Your calculator. All the individual single task apps on your smartphone. The checkers or Starcraft AI you play against online. These are all early examples of ANI.
But since 2010, we’ve also seen the rise of more sophisticated ANIs, these ones with the additional ability to consider past information and add them to their preprogrammed representations of the world. In other words, these newer ANIs can learn from past experiences and progressively make better decisions.
The Google search engine is an obvious example of an epicly advanced ANI, serving you the info you’re looking for seconds before you even finish typing your question into the search bar. Likewise, Google Translate is getting better at translation. And Google Maps is getting better at directing you where you need to go faster.
Other examples include Amazon's ability to suggest products you might be interested in, Netflix's ability to suggest shows you might want to watch, and even the humble spam filter that gets better at filtering out tempting 'get rich quick' offers from supposed Nigerian princes.
At the corporate level, advanced ANIs are used everywhere these days, from manufacturing to utilities to marketing (e.g. 2018 Facebook-Cambridge Analytica scandal), and especially in finance, where specialized ANIs manage over 80% of all stock trades in US markets.
And by the 2020s, these ANIs will even begin diagnosing patients and recommending medical care specific to the patient's medical history or DNA. They will drive our cars (depending on the local laws). They will begin giving legal counsel for routine legal cases. They will handle most people's tax preparation and begin processing increasingly complicated corporate tax accounts. And depending on the organization, they will also be given managerial functions over humans.
Keep in mind, all of this is AI at its simplest.
What is an artificial general intelligence?
The next level up from an ANI is an artificial general intelligence (AGI). Sometimes called “strong AI” or “human-level AI,” the future invention of an AGI (forecasted by the early 2030s) represents an AI that’s as capable as any human.
(This is also the level of AI that most fictional AI represent, again like Data from Star Trek or the T-800 from The Terminator.)
This sounds odd to say given that the ANIs described above, especially the ones powered by Google and Amazon, seem so all powerful. But in truth, ANIs are amazing at what they were designed for, but ask them to do anything else and they fall apart (metaphorically, of course).
Humans, on the other hand, while we're hard-pressed to process terabytes of data per second, our minds excel at being amazingly adaptable. We can learn new skills and learn from experience, change objectives based on our environment, think abstractly, solve all types of problems. An ANI can perform one or two of these traits, but rarely can they do all of them together—this cognitive weakness is what AGIs will theoretically overcome.
To learn more about AGIs, read on to the second chapter of this series that explores this level of AI in depth.
What is an artificial superintelligence?
The last level of AI is what leading AI thinker, Nick Bostrom, defines as an artificial superintelligence (ASI). An ASI would surpass current human performance in every factor, from logic to wisdom, from creativity to social skills. It would be like comparing the smartest human genius, with an IQ between 120-140, to an infant. No problem would be outside an ASI’s ability to solve.
(This level of AI is seen less often in pop culture, but here you can think of Samantha from the film, Her, and the ‘Architect’ from the Matrix trilogy.)
In other words, this is the kind of AI whose intellect would theoretically surpass any intellect that has ever existed on Earth. And this is why you hear Silicon Valley heavyweights sounding the alarm.
Remember: Intelligence is power. Intelligence is control. Humans can casually visit the world’s most dangerous animals at their local zoos not because we’re physically stronger than these animals, but because we’re significantly smarter.
To learn more about the opportunities and the threats ASIs present to humanity, be sure to read on through the rest of this series!
Future of Artificial Intelligence series
Next scheduled update for this forecast
The following popular and institutional links were referenced for this forecast:
The following Quantumrun links were referenced for this forecast: