2 hours ago, Sohan Lalwani said:

What type of physics are you looking to stimulate if you don't mind me asking?

I would use past cosmological maps to develop a model to predict future trajectories of celestial objects. Would be interesting to see if it comes up results similar to relativistic model and even capture impact pf dark matter.

2 hours ago, CharonY said:

What you describe has been around not under the name of AI, but under the umbrella of machine learning. Neural networks as outlined by @Ghideon is an example of such an approach. The issues are also mentioned, including that most are black box approaches, which makes issues of overfitting problematic (i.e. the outcome could be heavily biased due to your training set, but you won't know why and how it affected it). In contrast to, say Newton's law which is a very reductive model (and hence is elegant, but also idealized) you'll get something that has many, many, unwieldy parameters. Trying to prune that down to something like a Newtonian equation will require significant follow-up work. Basically you will try to create a simplified system based on an overly complex ML-model. Whether that is easier than to derive them otherwise depends on the system, I guess (which is way above my knowledge level).

It's true that Neural Network tend to me black box and we may never understand what each neurons really do but do we really need to know? If a ML model is able to predict any phenomena reliably that current models in physics, I wouldn't care a lot that the model is black box. Of course we need to make sure that ML model is not overfit to fit the data.

1 hour ago, studiot said:

The first thing to note is that as with any computer system it can only output what it is programmed to output.

One of the features of much of human discovery has been the occurrence of the unexpected.

Columbus died thinking he had discovered Asia.

Fleming died knowing he had discoverd penecillin.

Rutherford's famous statement

https://spark.iop.org/collections/rutherfords-experiment

The quote is from a famous experiment described here and I have referenced it as it describes the sort of process that leads to discovery.

That is to be alert for the unexpected.

So one possible approach is to program your AI to compare with the expected and output anything that is unexpected.

I remember being the only student in the class that got the 'right' wrong answer in a Chemistry experiment when the teachers gave use the wrong solution and everybody else falsified their results to fit expectations.

Thanks for sharing your insights.

I feel that physics has reached a dead end for 3D human beings like ourselves. May be AI will be able to find the patterns that we are unable to so far.

33 minutes ago, Mahi_sayli said:

I would use past cosmological maps to develop a model to predict future trajectories of celestial objects. Would be interesting to see if it comes up results similar to relativistic model and even capture impact pf dark matter.

May I recommend Universe Sandbox? It does require some level of cost but nearly perfectly fits your criteria

Just now, Mahi_sayli said:

I feel that physics has reached a dead end for 3D human beings like ourselves.

Take heart

Ptolemy said that.

Pope Leo said that

Lord Kelvin said that

and yet

and yet

and yet

We are still constantly discovering that there is more to discover than we know.

4 hours ago, Mahi_sayli said:

It's true that Neural Network tend to me black box and we may never understand what each neurons really do but do we really need to know? If a ML model is able to predict any phenomena reliably that current models in physics, I wouldn't care a lot that the model is black box. Of course we need to make sure that ML model is not overfit to fit the data.

Well, that is exactly the issue though. If you do not know how something works, it is difficult to predict where it likely will fail or whether it is overfitting your data. If your conditions are sufficiently simply that could likely be circumvented and cross-validation approaches might help. But the more complex it is the harder it is to figure out whether the model has issues.

A huge example in ML are population data, where the model often ended up to be biased against minorities, as they were underrepresented in data sets. Especially in the medical field, this was a real issue resulting in worse outcomes for minorities which were only realized rather recently.

Of course with the perfect approach and the perfect data set it might perform well. But if we had that level of certainty, we might not need Ml in the first place as more reductionist approach might be similarly successful.