AI help example

Here is something off-beat, as I closed socials and blog posting, as per my main blog.

Update: Nature article on the continuing relevance of Young's experiment.

The question

One day I remembered the famous double-slit experiment - handy intro to quantum physics in the news - I actually remembered it partially wrong, as this   v e r y   l o n g   AI conversation will reveal: it starts correcting me with French resources as I'm in France, and I asked to switch to English ones incl. Feynman's original, but check this: it ended up proposing online experiment analogs to try out myself! And as I'm new to Python (this century's standard coding, I go back last century to Java & HTML), I asked for Python tools to load so-called Jupyter Notebooks (Wikipedia). 

The answer

As mentioned, the first conversation is a very long read, AI summarizes it neatly:

The pattern does not vanish because a human looks at it; it vanishes because the setup records which-path information, which destroys the interference... The key idea is that the interference pattern disappears when the experiment records which slit the particle went through, not because a person consciously “looks” at it.

That las part was my misunderstanding, which in turn made me wonder at first:

What if AI agents were to run a mathematical simulation of interference vs. observed-path variant of the double-slit experiment. Would that change anything, coz agents aren't observing, they're just running the experiment w no knowledge?

AI answer was quite simple, really:

 No — a mathematical simulation run by AI agents would not change the physics of the double-slit experiment. It would only calculate the outcomes; the interference pattern or its loss still depends on the actual experimental setup, especially whether which-path information is physically available... If the AI simulates the unobserved case, it will compute interference. If it simulates the observed-path case, it will compute no interference. But in both cases, the AI is only predicting the result; it is not causing the experimental outcome.

The simulation 

AI then obligingly asks:

Would you like a compact Python simulation of both cases, with a plot of the two patterns?

Then it answers: 

I can show you the simulation logic, but I can’t generate the Python plot in this environment because chart/code execution is unavailable here.

First it proceeds to offer a series of step-wise  study of available simulations you can follow (no need to login, just say "Close" and it likely goes to the top, so press "CTRL F" & search for "Python"). 

Second I asked for DIY Python tools, and again AI was helpful: I found that in Google Colab, the easiest interface was Jupyter Notebook, and check this out: toward the end of the conversation, AI offered to cast Python scripts into Notebook-ready scripts! Stored in my Colab Notebooks folder, start with readme.txt... Python-ers go! Alternatively get the Colab Project & page down to the animation for these begin & end results: 

Begin (click to enlarge, picture here)

End (click to enlarge, picture here)

It means that when you start with single trace, eventually the double-slit shows a Gaussian distribution of numbers of hits along intervals explained in the first AI conversation. The which-path recording however destroys the interference, and the distribution is more random. That first conversation expands on more options. The second AI conversation helps non-Python-ers launch their own. Enjoy!

Footnote: as with my main blog, this is under Creative Commons License.