Speed super-intelligence
I wanted to write about a very specific type of super-intelligence known as the speed super intelligence. I have been reading about Artificial Intelligence for a while now and, wanted to share my two cents. This post is for anyone who basically is interested in knowing how would artificial intelligence evolve over the years. Let’s begin with what the current state of AI is and where it seems to be headed.
How do we categorise AI now?
Artificial Intelligence or AI is currently on a broad level been classified into 3 categories:
- Artificial Narrow Intelligence (ANI)
- Artificial General Intelligence (AGI)
- Artificial Super Intelligence (ASI)
Where does AI stand currently?
Narrow Intelligence — Current world AI
Narrow intelligence is domain specific intelligence. Artificially developed narrow intelligence, broadly known as Artificial Narrow Intelligence or ANI, are good at specific kind of task. Most basic and ubiquitous example of an ANI is, say a calculator. A calculator can multiply say 6892 x 456 at an accuracy and speed that is at least several orders of magnitude above the average human. However, that is the only task it is good at. To look at some more specific examples, Deep Blue (AI) which defeated two-time reigning World Chess Champion Garry Kasparov in the game of chess would also fall under Artificial Narrow Intelligence or ANI. In more recent scenario, AlphaGo (AI developed through Reinforcement Learning by DeepMind at Google) which defeated Lee Sedol in the game of Go, a game which is much more advanced and has much more degrees of freedom than that of the game of chess. However, AlphaGo still gets classified under the banner of ANI.
Now, what about upcoming technologies such autonomous driving cars. Are they above ANI? No! As Elon Musk, himself mentions in his talk that self-driving cars fall under the strict paradigm of narrow AI. They are just very, very good at one thing, driving better than humans. In fact, all of artificially created intelligence in history till this point falls under this very narrow category very aptly called Artificial Narrow Intelligence.
What is the human level intelligence benchmark?
General Intelligence
General intelligence in the current world exists only in biological form and has not been artificially developed yet. To define in simpler terms, define a spectrum with Artificial General Intelligence (AGI) when developed would be machines which can match human-level intelligence in all general activities. It would include intelligence in general tasks such as reading a book and understanding, talking, writing or anything that any human can do with ease. Any machine that can match humans in general activities would be classified as AGI.
Journey of intelligence beyond the humans
Super Intelligence
Super intelligence is the spectrum of intelligence which starts at intelligence level slightly above humans (say Einstein) and extends beyond to infinity. So basically anything more intelligent than humans no matter in how infinitesimally small manner would still be broadly classified as super intelligent.
Summary:
- Narrow Intelligence is domain specific intelligence.
2. General Intelligence is human-level intelligence
3. Super Intelligence is intelligence beyond human-level intelligence.
What happens after AI achieves human level intelligence?
Now, that’s a question many have been asking for a long time. And, a very small number of people do feel that it can never be achieved. However, most AI practitioners that the arrival of AGI may vary from as soon as 10 years from now to as far as maybe the end of this century. In fact, Elon Musk strongly believes that 10 years is the upper-limit of the arrival of Artificial General Intelligence (AGI).
“The pace of progress in artificial intelligence (I’m not referring to narrow AI) is incredibly fast. Unless you have direct exposure to groups like Deepmind, you have no idea how fast — it is growing at a pace close to exponential. The risk of something seriously dangerous happening is in the five-year timeframe. 10 years at most.” — Elon Musk
However, in spite of the varied opinions for the arrival of AGI, almost the entire community strongly believes that once AGI arrives, it would take-off into the ASI (super-intelligence) space in no time (literally, as low as microseconds). From a matter of few seconds to maybe a few days. In fact, some have beliefs that on the artificial intelligence station, AGI may not even be a stop. AGI when seen from a larger picture is nothing but just a benchmark on the intelligence spectrum that we as humans feel is special checkpoint since, it would mark the crossover of machines and humans.
It takes decades for the first AI system to reach low-level general intelligence, but it finally happens. A computer is able to understand the world around it as well as a human four-year-old. Suddenly, within an hour of hitting that milestone, the system pumps out the grand theory of physics that unifies general relativity and quantum mechanics, something no human has been able to definitively do. 90 minutes after that, the AI has become an ASI, 170,000 times more intelligent than a human.
-Tim Urban, The Road to Super-intelligence — Waitbutwhy.com
So, it may take as less as 90 minutes for AI to exponentially grow 170,000 times once it reaches AGI. Tim Urban talks in great detail about how we could transcend from artificial narrow intelligence into an era of artificial super intelligence. It’s a must read if you wanna explore this topic in further depth.
So, moving forward and getting into the more intriguing topic of what could be possible ways and means by which AI superior to humans could be developed (if it can ever be developed).
What are possible paths to super-intelligence?
Super Intelligence though very tough be articulated by beings which possess general intelligence can be broadly classified into 3 types again.
- Speed Super Intelligence
- Collective Super Intelligence
- Quality Super Intelligence
However, for the purpose of discussion in this post — I shall only be delving into the topic of Speed Intelligence and how would it feel for being which attains such a state.
How would it feel to be Speed Super Intelligent?
I shall be getting to the part which this post is about — Speed Super Intelligence. And, it’s the simplest section of this post. It’s just as simple as it sounds. Speed Super Intelligence is intelligence achieved by sheer speed of computation and calculation efficiency.
Let’s say you are at Starbucks for coffee. The guy at the counter just called out your names and your friend walks over to get his Latte and your frappe. While walking back, a girl in rush, brushes past him dropping your coffee to the floor.
How would this entire scenario play out for you if your brain worked 10,000x the normal brain?
But, now you get to slowly watch the cup fall to the floor like an excruciatingly long predictable feature movie where you know the entire plot. For you, the journey of the cup to the floor would be nothing less than say watching a snail cover a distance of 1 meter across the floor.
So, you might as well go out and order some Chinese, eat it hopelessly with chopsticks, have some desserts, go to your friend’s place, grab a fresh t-shirt for him (and also say a towel), come back to Starbucks, order yourself a new Java-chip Frappe and sit back at the same old spot while that cup now starts to touch the surface of the ground.
Speed Intelligence: Most probable path to super-intelligence
This apparent time dilation gives such an incredible intelligence advantage to an entity which would probably be the first and one of the easier way to achieve super-intelligence through increased computational efficiency. So now you know how an artificial speed super-intelligent machine would feel about us humans. The same as we do about snails.
