Some General Thoughts About the Future

This is pretty much an extract of a conversation I had with some friends about our future, AI, and everything that could be in it. It was mainly sparked after finishing Superintelligence and Life 3.0 (both great books, by the way).

The main point was this: we are 20 now, and AI seems to be booming. Physical AI or superintelligence might be in our near future, the former more likely than the latter. What could society look like? What should we be doing?

First let's start with the two main revolutions that I see in our future, physical AI and superintelligence, related but not correlated.

Physical AI

Have you seen the latest robot from Boston Dynamics? From Figure? From the 100th Chinese startup? Probably at least one of them. Have you ever stopped to think that these might actually be able to perform real tasks?

I remember watching Atlas, the original robot from Boston Dynamics, in middle school. It seemed unreal what it could do, but it was clear that it was a research project. It was bulky, clumsy and constantly under supervision.

Now we're at a point where combining the intelligence of the current models with an advanced enough body could lead to a superior worker in many fields. Actual deployments in the real world, with hourly costs that over time could become negligible, running 24/7 on cheap energy. This sounds exciting to me, and maybe scary to you. But what I find truly exciting is their digital brain.

Humans take years to develop and most people stop actually learning after university. However, the few of us who continue doing so can develop extraordinary physical and mental skills. Up until now a lifetime of personal work was needed to achieve them, and every generation started pretty much from a clean slate. Evolution is painfully slow.

Now think about humanoids. This applies to any robot, but for simplicity let's stick to them. In a lab somewhere they can teach one to land a perfect backflip. Upload the code, and now every single one can do it. Teach one how to change a tire, and now every single one can do it. Teach them how to shoot a rifle, and they can all shoot.

This is the key point: the ability of an individual to simply install the learnings of the whole species. Imagine if you could download the Einstein DLC, or maybe in my case the aerospace engineering DLC. Three years of hard work done in maybe a couple of hours.

We will have perfect workers for every task, provided their dexterity and mobility improve enough. Physical labor will be dead.

The best part: this is not really an if, but a when. Two years? Five? Ten?

Superintelligence

Now, this is the gamble. Anthropic says Mythos is too powerful, that it's a threat to national security. Superintelligence might be close, as Mythos can rewrite and improve its own code.

In reality, are we close? According to the arguments in the book Superintelligence, honestly nobody knows. The best AI researchers on the planet can't agree on when, or if, we will ever get there.

We might hit an intelligence plateau, never arriving at true superintelligence. If we do get there, though, what would our future look like? According to Superintelligence there are several scenarios. To sum some of them up:

• Paradise: we don't work if we don't want to and we can live our lives however we want. Abundance.
• Dystopia: AI turns rogue, its goals don't align with ours and we are annihilated. For example, computronium trying to produce paper clips.
• Observant: like a god it observes and only acts in extreme situations; we would not feel its effect directly.
• Oracle: we ask, it replies. It can be used for destruction or for goodness, but we are in control.

In some we keep our freedom of choice, in some we die, and in some we simply gain instant access to the hardest questions we can think of.

These are extremely different futures. All of the positive ones would make our lives easier and better than those of any human who has ever lived. We could live forever, explore the stars, follow what truly guides us with no limits. Is this worth risking our lives? Should we keep going with superintelligence development? To be honest we are nowhere near experienced enough to answer this, or to affect it in any meaningful way, so we should just consider it as a possibility, move on, and build our lives and relationships normally.

Some of my friends responded to these scenarios with "it will never happen", or "we're all doomed". I believe the only way to approach them in a healthy way is to be positively cautious. What does this mean? Accept that they might happen, see everything good that could come of them, reflect on what could go wrong, and vote and work on ways to mitigate the downsides.