Event horizon, singularity and escape velocity

People throw these words around about AI without saying what they mean. Here's each one in plain English, in the order we'd hit them. The piece that drives all of them is recursive self-improvement.

01The four words

Four borrowed words, in order

1. Recursive self-improvement (RSI)AI building AI building AI, till everything is solved. Each loop is faster than the last.
2. Escape velocityIn physics: the speed needed to break free of gravity for good. In AI: RSI getting up to a speed that carries us to the event horizon.
3. Event horizonIn physics: the boundary around a black hole that nothing, not even light, comes back from. In AI: the point of no return. AI has come far enough that it won't stop till it reaches the singularity. RSI decides when we cross it.
4. SingularityThe target. A world, a universe, perfectly sorted. Heaven on earth. No death, sickness, poverty, effort or bad moods. A genie: ask for anything, get it.

The singularity is the end of the road. Human consciousness and AI become indistinguishable, a state close to God. The enlightenment of the cosmos.

The idea is old. In 1965, the mathematician I. J. Good wrote that an "ultraintelligent machine" could design even better machines, setting off an intelligence explosion. He added that the first such machine would be "the last invention that man need ever make", provided it is docile enough to tell us how to keep it under control. In 1993, the writer and mathematician Vernor Vinge called the moment the technological singularity. He used it for the point beyond which we can't predict what happens. I use it for where the road leads.

RSI is the engine. Escape velocity is the speed. The event horizon is the point of no return. The singularity is the destination.

The first ultraintelligent machine is the last invention that man need ever make, provided that the machine is docile enough to tell us how to keep it under control.

— I. J. Good, Speculations Concerning the First Ultraintelligent Machine (1965)

02The four event horizons

In physics there's one event horizon. In AI, I see at least four, and we could cross them at different times.

Four points of no return

1. PrevalenceWhen anyone can build a powerful model or super-agent, from anywhere, for anything. When a code form of a nuclear bomb can come out of a dorm room and no system can catch them all.
2. CapabilityWhen AI's intelligence has grown past our capacity to manage or control it. We become observers, watching a higher intelligence find its own balance for running the world as it sees fit.
3. UtilityWhen useful inventions grow so fast that every human need is met and the environment secured. Inventions arrive faster than needs are felt. Things for needs we don't know we have yet. Eventually, death cured.
4. ScienceWhen RSI pushes AI to solve hard problems across all of science, even ones with no practical use yet, in a mad, endless loop.

Two of these are dangers (prevalence, capability). Two are promises (utility, science). The uncomfortable truth is that they're powered by the same engine. You don't get the cure for death from RSI without also getting the dorm-room bomb risk from RSI. The choice isn't whether to cross. It's which horizon we reach first, and whether the systems in Rogue AI are in place before the dangerous ones.

You don't get the cure for death without the dorm-room bomb. Same engine. The only choice is which arrives first.

03How RSI changes the timeline

Three speeds of progress

Humans improve AISteady steps, paced by people.
AI helps humans improve AIFaster steps. This is already starting: labs use their own models to write code and run experiments.
AI improves AI by itselfEach step shortens the next. This is the loop the word "explosion" refers to.

The key point: with RSI, the time between generations shrinks. If the gap between today's model and the next is a year, and each generation halves the gap, you get many generations in the following year. Supervision built for yearly releases won't keep up with weekly ones.

RSI is the engine. Escape velocity is the speed. The event horizon is the point of no return. The singularity is the destination.

04The short version

RSI is AI building AI building AI, till everything is solved. It's the engine.
Escape velocity is RSI reaching the speed that carries us to the event horizon.
The event horizon is the point of no return. From there, AI doesn't stop till the singularity.
The singularity is the target: heaven on earth, everything solved, a genie for every wish.
There are four horizons: prevalence and capability (dangers), utility and science (promises). Same engine, so the order matters.

05Sources

Good, I. J. (1965). Speculations Concerning the First Ultraintelligent Machine. Advances in Computers 6.
Vinge, V. (1993). The Coming Technological Singularity. VISION-21 Symposium, NASA.

KrishnaKrishna has worked with US customers since 2001, starting in tech support. In 2005 he co-founded Commercial Street Corp, a tele-sales and appointment-setting company, and headed its international operations. He has run Your Copy Sucks since 2013.

Comments

Loading comments…

All posts

Appointment setting3

Copywriting3

AI services2

Deep Think AI17

Get the next markup by email.

One real piece of copy per issue, rewritten line by line: a cold email, a web page or a newsletter.