AI Designed Quantum Physics Experiments.

[FrumiousBandersnatch]

Two things struck me. The first is that Theseus is a much cooler name for a super computer than Melvin (not sure if there's an anagram in there or maybe there's some computing correlation with the Ship of Theseus?).

And secondly, this quote from the linked article:

'This process makes calculating the final quantum state much easier, although it is still hard for humans to understand'.

Stupid humans...

Theseus doesn't have to understand it - it uses the "Shut up and calculate" method

 

[FrumiousBandersnatch]

... Also that there is a deep connection between QM experiments and mathematics' Graph Theory (which was also mentioned in a more detailed way in the OP SciAM article). The 2017 article says:

This is along the lines of what Stephen Wolfram is working on with his project to generate all of physics from a (relatively) simple computational rule (derived from his work on cellular automata). The results of his computational models are many different types of mathematical graphs and mappings from which many fundamental rules of physics emerge - including, apparently, the Einstein equations...

He talks about it in this recent blog: Stephen Wolfram on Computation, Hypergraphs, and Fundamental Physics.

 

[FrumiousBandersnatch]

...So it appears that the real 'takeaway' here might be that the AI Reinforced learning technique, gives access to solutions which then mysteriously 'test out' (or are 'consistent with') real life measurements/observations .. So why should that be so?

Perhaps because this AI was based on 'toolbox' models of empirical methods?

 

[sjastro]

Another earlier report here (2018), says that these AIs (MELVIN and Theseus?) use reinforcement learning .. Ie: similar to AlphaZero and Leela Chess Zero, (discussed in a previous thread here).

More snippets here (going backwards in time - Dec 2017). This one clarifies more on what this exercise is showing - with that being, that human logic itself, is thought to be the main constraining issue when it comes to understanding QM experiments:

Also that there is a deep connection between QM experiments and mathematics' Graph Theory (which was also mentioned in a more detailed way in the OP SciAM article). The 2017 article says:

An answer to a quantum-physical question provided by the algorithm Melvin has uncovered a hidden link between quantum experiments and the mathematical field of Graph Theory. Researchers from the Austrian Academy of Sciences and the University of Vienna found the deep connection between experimental quantum physics and this mathematical theory in the study of Melvin's unusual solutions, which lies beyond human intuition.

An application of graph theory is in the mathematical modelling of the spread of COVID-19 as shown in this thread.
In the last century in my days as an undergrad graph theory was taught as a 3rd year subject in pure mathematics.
I find reading these links in combination with watching too much Star Trek in all its reincarnations it is easy to get carried away and think Melvin used graph theory in developing quantum physics experiments while the slothful dumb humans gradually came to this realization.
Whatever solution Marvin came up with, it is the "generalization" used by physicists, to deal with this query earlier in the thread, in establishing the connection between quantum physics and graph theory.
Marvin is not smart enough to make the connection otherwise mathematicians in particular should step aside and let AI do the research in math and its applications.

 

[sjastro]

Relevant to this thread, in a landmark decision the Australian federal court has decided AI systems can be legally recognised as inventors in patent applications.

Australia's Federal Court made the historic finding that "the inventor can be non-human".

Can artificial intelligence be an inventor? A landmark Australian court decision says it can

 

[sjastro]

Relevant to this thread, in a landmark decision the Australian federal court has decided AI systems can be legally recognised as inventors in patent applications.

Can artificial intelligence be an inventor? A landmark Australian court decision says it can

If this progression continues and AI is universally recognized as inventors what is the next step; AI being recognized of making independent contributions in all areas and therefore potential candidates for winning Nobel prizes or its equivalent in mathematics the Fields medal?

The Fields medal would be interesting, mathematicians are washed up by the age of forty and incapable of winning the medal; not sure how age would fit in the scheme of things.

 

[Halbhh]

If this progression continues and AI is universally recognized as inventors what is the next step; AI being recognized of making independent contributions in all areas and therefore potential candidates for winning Nobel prizes or its equivalent in mathematics the Fields medal?

The Fields medal would be interesting, mathematicians are washed up by the age of forty and incapable of winning the medal; not sure how age would fit in the scheme of things.

Would have to recognize those coders after all.

 

[sjastro]

Would have to recognize those coders after all.

Using the chess analogy Leela Chess Zero was coded and trained by machine learning by individuals who only needed to know the rules of chess.
When Leela Chess Zero plays moves that are not only well beyond the comprehension of these individuals but also the very best human players, one can't credit any human for the moves made.

We have now gone past the Turing Test.

To physicists asking the question whether AI will eventually win the Nobel Prize in Physics with further developments in artificial neural networks to facilitate deep learning as opposed to machine learning.

AI Podcast: “Could an AI Win a Nobel Prize?” Quantum Researchers Ask | The Official NVIDIA Blog

 

[partinobodycular]

They play chess at an evolutionary level beyond humans now this...
AI Designs Quantum Physics Experiments beyond What Any Human Has Conceived

Step 1: The AI designs it.

Step 2: The AI runs it.

Step 3: You're in it.

 

[SelfSim]

Step 1: The AI designs it.

Step 2: The AI runs it.

Step 3: You're in it.

AI still has to overcome some rather major hurdles between steps 2 and 3, such as:

- obtaining construction materials
- manufacturing new parts, including its smallest parts, and thinking apparatus
- providing a consistent power source
- programming the new members
- error correction of any mistakes in the offspring.

These aren't insignificant, when compared with the entrenched and widespread influence humans have achieved to date.