zijlstravideo

Well, object detection and text generation work more or less the same.

The "AI" weights which word would likely follow up the previous word the most (based on all the text in its data set), and picks the next word(s) accordingly. Also including some rules to randomize choices and weighting in the many synonyms, calculating the distance between keywords to determine how much these terms are related (and weight their importance).

If you take away some other on and off page SEO ranking factors that might impact the search results...
In the end, isn't the "AI" doing more or less the same thing as what search engines have been doing for many years now?

- Take user input
- Calculate the relation distance between keywords
- Factor in synonyms
- Use weights to calculate the relevance between words, then predict which are likely most important
- Output best matches

In that case, wouldn't Google be considered the biggest AI? It's even trained on the largest data set possible. But I barely hear anyone mention Google as even being an AI, even though they literally try to get the same results.


Anyway... Besides all the AI buzzwords being overhyped, this stuff is really impressive!

A year ago, I would have never thought that one year later we would be able to turn a few pixels back into a photo realistic image... But we can. While not 100% perfect yet, it's insanely, almost scary good:



Same for other things, like...
Having to worry about small sensor camera's generating too much noise during low light shoots, used to be the most common thing in the world...

Suddenly, a ton of great machine learning models get released and noise on raw footage is no longer any issue at all (if you know what you're doing).
Feels kinda surreal as it changes so fast.

A lot of these machine learning models are already that freakin' good.
This will effect a lot of jobs one way or another, and probably sooner than later.

__________________
Contact: email