In Frank Herbert's Dune, the affairs of the entire universe revolve around a psychedelic drug known as the Spice Melange that enables beings the ability to fold spacetime and see into the future. dune.fandom.com/wiki/Spice_Mel…
The spice is essential because without it travel between planets in the universe would be practically impossible. The universe is interconnected in commerce through psychedelics.
The spice however also allows its consumers to see into the future. Hence to make predictions of what might come. Does not one find it odd that success in our modern financial industry relates to our ability to see into the future?
An intelligence (natural or artificial) that can disentangle the events of the past and the present to uncover the possibilities of the future is equivalent to a perpetual money-making machine. It is also curious that in the Dune universe that artificial intelligences are banned.
In Dune, there exist concentrations of events in the future that its clairvoyant beings cannot see past. That the density of entanglement of events can't be separated to see future consequences.
It's such a fascinating idea that you can write up an entire sci-fi novel on the idea that quantum computers can become clairvoyant oracles.
The singularities of cognition, where non-intuitionistic logic originates from, are analogous to a kind of inference that C.S. Peirce identified 100 years ago. He conjured up a terrible neologism, which he called abduction.
Fermat's Last Theorem which he scribbled in the margins of a paper is a powerful example of abduction. It took nearly 4 centuries to discover the proof of Fermat's abduction. quantamagazine.org/why-the-proof-…
Human minds can bend space and time to arrive at certain conjectures that not only takes centuries to prove but requires at least 129 pages of proof. scientificamerican.com/article/are-ma…
To be fair, Fermat didn't know if his conjecture was true, he just guessed it by extrapolating from previously known equations. The universe is sometimes self-similar in that patterns in the small also recur consistently across all scales.
The business of abduction involves the identification of the patterns that we can observe and extrapolating these to much wider domains. It is not the same as bending spacetime, but it's very useful for predicting the future.
We are better at predicting the future when we recognize recurring patterns of this world. We know that the sun will set today and rise tomorrow. We know that the seasons will change. Knowledge of these recurring patterns makes useful the business of planning.
We employ the discoveries in science to engineer how we reshape our world. Modern human life is predictable because we have reshaped our world to make it predictable. Before the 1950s, most people could just drop dead without warning because of smallpox.
Humans however have engineered all kinds of protective technologies that allow us to survive unpredictable events. We build robust dwellings to survive hurricanes and earthquakes. We invent more agricultural methods to avoid famine. We discover vaccines to avoid plaques.
We invent financial instruments to mitigate against the financial uncertainty of agriculture. We build all kinds of technological infrastructure so that our realities are more predictable.
Yet we have for centuries people who push back against modernity. Why do they push back against a more predictable reality? Is it perhaps because predictability can be correlated to determinism. Too much predictability implies a loss of determinism and hence free will?
There has always been the philosophical conundrum that if the universe was deterministic, will we then have free will? The universe of course isn't deterministic, so let's not waste any more precious brain cycles on this alternative metaphysics.
The evolution of the universe is open-ended. What this implies is that we cannot predict the future with absolute precision because we do not have the vocabulary to express what has yet to be discovered in the future.
We have no trouble speaking about computer science and internet ideas today. This is because we've developed the vocabulary to speak about these topics. 100 years ago they did not have that vocabulary. That of course didn't hinder some imaginative sci-fi authors.
We create civilization so that our lives become predictable. But that predictability is a luxury that we should exploit so that we can explore new worlds that have not yet established any semblance of predictability.
When we shut our minds out to the possibility of the enormous diversity and richness of reality, we see only the illusion of a world with absolute predictability. Hence a world without free will. A world that is absolutely monotonous. A world without variety and surprise.
What's the logic behind DeepMind's universal Perceiver-IO block?
Perhaps we want to compare this with the original perceiver architecture:
Note how the input array (in green) is fed back into multiple layers as a 'cross attention' in the previous diagram. That cross attention is similar to how your would tie an encoder with a decoder in the standard transformer model:
This is also why money printing thingamajigs have so much persuasive value. Hence why cryptocurrencies have their appeal.
People are more easily persuaded to pay for something if they perceive that it is an investment. An investment is anything that makes more money than what you originally put in.
Why are neural networks unable to nail arithmetic or multiplication? That is, you might be able to ask GPT-3 what 5 plus 7 equals, but you can't calculate 59 + 77 (trust me, it can't). Why is that?
This is because neural networks are unable to formulate compositional models of reality. Would a caveman be able to invent arithmetic or multiplication? I seriously doubt it, it requires a gifted human individual to invent these from scratch.
What is wrong with knowledge representations that it has barely moved the needle in machine understanding? @danbri
Intuitively, KR should be useful in that it diagrammatically records how concepts relate to other concepts. Yet for a reason that is not apparent, it isn't very useful in parsing out new understandings of the concepts in its graph. Where did we go wrong here?
Perhaps it's because knowledge graphs are noun-centric and not verb-centric. Reality is verb-centric. To get an intuition about this, watch this explanation of the open-world game Nethack:
Noun-centric thinking is making it difficult to explain and understand the dynamics of covid19. People seem to not understand that everything is a process.
Knowledge discovery is a process. Covid19 is novel because it is a new virus and its characteristics are unknown without further investigations. Of course, one has to balance the time required for information and the need to act swiftly.
Science is a knowledge discovery process. It is not a label you slap on to something that remains unchanging for all time. Yes, we do these for food items up to an expiration date. But after the expiration date, all labels attached to the food item aren't expected to hold.