I see Google Translate as a prime example. It’s so easy to confuse it. Type more than 2 words or phrases and chances are high that there will be an error. The bots that code the words and phrases structure them as elements with context, but those who programmed them chose the wrong elements! Basically, they treat letters as elements (which are made of smaller elements xD), then compute relationships between letters to mark differences between words. Next step, issue requests to retrieve data about the word to create a context that may be meaningful to a user. That data is again structured in the same way. Basically all they code is ‘images’ of a word, its form, but nothing else – structure does not correspond to natural language and something more like semantics is far far away. You say 2050…that’s wishful thinking with the current approach. The way coding bots is going, the goal seems to be: memorize every linguistic combination as form, and then use algorithms to ‘guess’ with a high chance of success what the human wants, had intended or thought of. I mean that’s what search engines do right? They guess, and the best search engine, Google, is better at guessing since it contains remarkably more information in its databases than others.

But, it doesn’t have to be like that. You don’t have to program bots to guess! It is possible to structurise programmed words in such a way as to correspond to natural ones. Most of natural language has a fixed finite structure that can be properly quantified. That approach would be finalized in a MUCH shorter amount of time and more importantly it will mirror human language, contain its full etymology and later just be updated with new info when new meanings get generated in the real world.

In natural language ‘apple’ is a heavyweight, and ‘this’ is a grain of sand. ‘This’ is neither polysemic nor complex. It is just 1 element! Sure it has 4 letters in written and 3 sounds in spoken language. But conceptually it is 1. When broken down to simpler elements the word ‘apple’ has several HUNDRED conceptual 1′s. That is how each representation of a word needs to be structured if bots are to advance and get closer to human understanding. So basically when you see ‘this’ it should contain little information in itself, while the word ‘apple’ should be like a darn ocean of information. Those words need to be ALWAYS and to EVERYONE like that (some rich in info, others poor in info). And that info needs to be carried and copied EVERYWHERE. But for that to work you need: a new way of building information hardware and correspondingly software. Current hardware and software are nowhere CLOSE to producing anything as closely sophisticated as an AI. Change of paradigm – required, otherwise ‘this’ will still confuse many like it did some few sentences above :).