We were promised Strong AI, but instead we got metadata analysis

Further, meaning is an ever-evolving, ever-mutating process much like the universe.

Training on symbols cannot arrive at meaning, since the meaning isn’t contained in those symbols. Using past symbols, also means no room for evolution.

Machine learning does work though in areas where the needs of the end goal are densely present in the symbols being used for training.

Like recognizing text. We learn to recognize those marks from just the marks, and nothing else. And so those marks contain all that is needed to recognize them. This can be encoded/learned.

But what they mean isn’t encoded in them, nor is it in words, in sounds, in facial expressions, in tones, in body gestures. It may even lie in between us, rather than in us.

I am not really engaged with AI-research, but I follow the area with interest and my impression is, that if strong AI will emerge in the next time, then only by accident. I mean there are lot's of awesome advancements and for example I did not expect Go to be solved since years already, but still - I see no way from current tech, to a general AI, that can really understand things.

Or is someone aware of more groundbreaking research?

Further, I'd wager more than half of AI researchers are at least surface-level familiar with brain science, not (as you claim) less than 1 in a million (are there even a million AI researchers?). There's significant work between computational neuroscience, mathematics, philosophy, computer science, etc, etc in the field.

Many smart people are giving it their best effort to understand different pieces of the puzzle from many different viewpoints and angles; FAANG corporations might be among the most visible, but their AI is necessarily profit driven and close to the ground, relevant to currently tractable problems (amenable to 'mere statistics').

And in fact, slavishly copying nature is something which has long been on the AI back-burner, but we're on the order of at least a decade from being able to create a computer system with enough transistors to do so.

Not sure what else to say, really.

1) metadata is helpful - not all of it, but some; 2) ML is obviously needed when metadata is missing, and metadata is missing very often; 2) Even when metadata is present, pure ML-based extraction often beats it in quality, with right ML models. A combination of ML+metadata fallbacks is even better.

Website creators often make mistakes providing metadata, they may misunderstand the schema and purpose of various fields, have metadata auto-generated incorrectly, etc. It is rarely about deceiving for the tasks we're working on (though it also may happen).

So, I don't see Zyte falling back to metadata analysis, ML models are already better than this human-provided metadata - but metadata is helpful, as one of the inputs.

We're going to publish product extraction benchmark soon, where, among other things, we compare automatic extraction with metadata-based extraction. In this evaluation we've got a result that ML + metadata is better than metadata not only overall (which is expected), but on precision as well.

I wonder if the reasons metadata is sometimes preferred are not related to quality, or to failure of ML approaches. If Google doesn't get data right, it is not Google's fault anymore, it is website's fault.