Eliezer Yudkowsky asked”If artificially intelligent systems someday come to surpass humans in intelligence, how can we specify safe goals for them to autonomously carry out, and how can we gain high confidence in the agents’ reasoning and decision-making?” Yudkowsky has argued that in the absence of a full understanding of decision theory, we risk building autonomous systems whose behavior is erratic or difficult to model. Thus Yudkowsky is thinking about having a system in AI where human’s core values are persevered and executed.(assumed it is possible) Then the AI would be ethical, and they will act in a way we want, and not the way we tell them to do. This idea is called humanity’s coherently extrapolated volition. So they can make their own choices and still be ethical.

One of the problems in decision theory is the prisoner’s dilemma. For example, we could imagine that if both players cooperate, then both get $10; and if both players defect, then both get $1; but if one player defects and the other cooperates, the defector gets $15 and the cooperator gets nothing. Now, imagine one of the players is you and the other is the copy of you, so both of you know exactly how each other would react and act. Because two players know how each other will play, thus it is more likely that both of you will choose to cooperate and get the most beneficial result. Both player’s actions are in fact influenced by each other even though in a prisoner’s dilemma they are separated. Therefore their decision is timeless. This is a decision model developed by Yudkowsky, called timeless decision theory (TDT).

Roko’s basilisk was an attempt to use Yudkowsky’s proposed decision theory (TDT) to argue against his informal characterization of an ideal AI goal (humanity’s coherently extrapolated volition).

According to Yudkowsky, we will make AI in which they will act according to our human core values. But just as in TDT, we humans know the blueprint of AI while AI is smart enough to know how humans behave and will behave. Roko observed that if two TDT agents with common knowledge of each other’s source code are separated in time, the later agent can (seemingly) blackmail the earlier agent. Call the earlier agent “Alice” and the later agent “Bob.” Bob can be an algorithm that outputs things Alice likes if Alice left Bob a large sum of money, and outputs things Alice dislikes otherwise. And since Alice knows Bob’s source code exactly, she knows this fact about Bob (even though Bob hasn’t been born yet). So Alice’s knowledge of Bob’s source code makes Bob’s future threat effective, even though Bob doesn’t yet exist: if Alice is certain that Bob will someday exist, then mere knowledge of what Bob would do if he could get away with it seems to force Alice to comply with his hypothetical demands.

Since a highly moral AI agent (one whose actions are consistent with our coherently extrapolated volition) would want to be created as soon as possible, Roko argued that such an AI would use acausal blackmail to give humans stronger incentives to create it. Roko made the claim that the hypothetical AI agent would particularly target people who had thought about this argument, because they would have a better chance of mentally simulating the AI’s source code. Roko added: “Of course this would be unjust, but is the kind of unjust thing that is oh-so-very utilitarian.”

Roko's conclusion from this was that we should never build any powerful AI agent that reasons like a utilitarian optimizing for humanity's coherently extrapolated values, because this would, paradoxically, be detrimental to human values.

Click here to download this writing

This is the documentation of my involvement in Olivier’s project L’Adapté, where I was the one who tried to learn to walk on this absurd leg extension:

Portal to Olivier’s project: link

After about an hour or more of getting used to the leg extension, my body started to coordinate and walk. To coordinate, it means the following:

Making every step bears a risk. Without putting most weight on one leg and two hands extension, I could not have lifted the other leg and step forward. However, since the leg extension has to tendency to fall to the right and left side, putting weight on one leg is never a stable position. Hence, I need to incorporate two hands to balance for around 2 seconds. If I lost balance in that 2 seconds, I fell. However, there is a trick. If I continue to walk without stopping, it is easier to balance. Just like riding a bicycle at high speed is easier than riding as slow as a turtle.

Then, it is about the weight. Because the leg extension only connects to my calf, basically I am kneeling on the extension. It is quite impossible to lift it with only my calf. Therefore, the sole was needed to lift the back of the leg extension, while two hands extension as a joint to share the weight.

Finally, it is the angle between two legs. If you notice carefully, a human’s two feet can never walk in parallel like a robot. We tend to make a certain angle in between both feet. It is an insignificant thing to notice when walking normally. Yet, walking on this leg extension, such angle became a crucial part of whether I can make a smooth transition to the next step. Little bumps on the ground or tiredness of my body sometimes would produce undesirable angles. Hence, a lot of the time I have to deal with these undesirable angles by redistributing force upon the 4 spots (two hand extension and two leg extension).

I would consider the above as a trace of where I put my attention, rather than a complaint about the inhumane design of the machine.

When everything is enlarged, the slightest details matters, just like why quantum physics is important to the exploration of the universe. When every little details matter, our automatic physical response, and our perceptive categorization does not work. I could not rely on how I usually walked nor how I normally classified dangerous scenarios. Everything has to be unlearned and relearned again. I went retrospect in time and became a baby than trying to learn how to walk.

As I started to walk while imagining how should I walk, for the first time, I experienced a loss of control upon my body. There is so much wiring that has been done when we were a child to allow our brain to control our movement smoothly.

After hours of walking on it (well I was in a zone, which means highly focused, not even able to talk, that is why my sense of time probably is distorted), my feet and my hips began to experience fatigue. My hands were comparatively less fatigued. I think this is funny because it might be how the elderly are experiencing when they have to walk on walking sticks.

Click here to download this writing

What is an exhibition? It is not a place for presentation, it is a place for things to be curated. And through curation, all the great art pieces are killed and reincarnated into a new baby. It is believed that if you have done something great in your past life, you would reincarnated with inheritance of those good trace you had in your past life. A great artwork, even though killed by curation, would inevitably reincarnated with the aura of its greatness.

Click here to download this proposal