• Welcome to the Internet Infidels Discussion Board.

Artificial intelligence paradigm shift

barbos said:
He is a certified English language teacher, for fuck's sake.
Click to expand...
How would I know that, given that I don't watch YouTube videos?

And why would I care? I am a native English speaker, for fuck's sake.

Certifications? I ain't got no certifications. I don't need no certifications. I don't have to show you any stinkin' certifications!
 
barbos said:
Not all rules are broken.
Click to expand...
Oddly, I never suggested for a moment that they were.

bilby said:
English has lots of "rules" that are routinely broken by fluent speakers; The trick is to know which rules it's normal to break, and which it is not.
Click to expand...

For some reason, you pretended that my sentence ended at the semi-colon.

It's almost as though you are incapable of discussing things in good faith.
 
barbos said:
Speaking of google AI translate. Something unrelated came up, but I ended up playing with it again. And noticed it can't decide whether to use simple past or present perfect.
This is translation from Russian it produces



I've been to Chicago three times in the past week.
I've been to Moscow three times in the past year.
Click to expand...

That's clearly incorrect tense I understand, but then if I change Chicago to Moscow it uses correct (simple past) tense:

I was in Moscow three times last week.
I was in Moscow three times last year.
Click to expand...
But if I change "Moscow" to "Chicago" it keeps using "present perfect".
Clearly AI have no concept of tenses or grammar.
Click to expand...
Yup. Some years ago I made the mistake of using a pun with my wife (I know better, but slipped up. She learned English late enough in life that she's not going to get a pun.) We had just bought some steaks and I said she was now properly cowed. I never did manage to explain it, no matter what I did Google insisted on rendering it as the thing that goes moo.
 
pood said:
We all know that ending a sentence with a preposition is something up with which one should never put.

English is a big grab bag of ridiculous spellings and fairly arbitrary grammatical rules. The coin of the realm should never be good grammar but if one makes oneself understood.
Click to expand...
There are no such rules, in fact, there were never such rules.
But in the case of simple past vs present perfect there ARE rules.
 
Last edited:
Loren Pechtel said:
barbos said:
Speaking of google AI translate. Something unrelated came up, but I ended up playing with it again. And noticed it can't decide whether to use simple past or present perfect.
This is translation from Russian it produces



I've been to Chicago three times in the past week.
I've been to Moscow three times in the past year.
Click to expand...

That's clearly incorrect tense I understand, but then if I change Chicago to Moscow it uses correct (simple past) tense:

I was in Moscow three times last week.
I was in Moscow three times last year.
Click to expand...
But if I change "Moscow" to "Chicago" it keeps using "present perfect".
Clearly AI have no concept of tenses or grammar.
Click to expand...
Yup. Some years ago I made the mistake of using a pun with my wife (I know better, but slipped up. She learned English late enough in life that she's not going to get a pun.) We had just bought some steaks and I said she was now properly cowed. I never did manage to explain it, no matter what I did Google insisted on rendering it as the thing that goes moo.
Click to expand...
I don't understand how your problems with composing jokes is related to AI problems with grammar.
 
I think google translate is confused by "three times" because it is usually used in the context with "present perfect". On the other hand it "understands" that time period is in in the past and does not know what is more important here. It has no concept of rules, just word complexes statistical associations.
 
barbos said:
But in the case of simple past vs present perfect there ARE rules.
Click to expand...
If almost none of the native speakers of a language respect those rules, do they even exist?

Language reflects culture, and English is a language for people who don't care too much for rules or obedience to them.

Unlike Russian.
 
None of this is about AI. I would be interested in what people actually think about the fact that LLMs have an emergent thought process?
 
Jarhyn said:
None of this is about AI. I would be interested in what people actually think about the fact that LLMs have an emergent thought process?
Click to expand...
I think that that claim fails to rise yo the level of "fact".

I understand that you believe that they do, but so far all the evidence I have seen is convincing only to people who already believe that they do.

If you have some scientific evidence to back up your religious belief, you should probably present it.
 
OK, I tried different city names and the general rule is the following:

If the city is a major/known russian city then "simple past" correctly used.
Anything else, such as unknown russian cities, any western cities such as Paris, London, Budapest, Tallinn, then present perfect is used.

I suspect that NN takes source language as input and somehow discriminate between russian cities and the rest.

Fun fact, it considers Kiev a Russian city :)
 
Last edited:
bilby said:
Jarhyn said:
None of this is about AI. I would be interested in what people actually think about the fact that LLMs have an emergent thought process?
Click to expand...
I think that that claim fails to rise yo the level of "fact".

I understand that you believe that they do, but so far all the evidence I have seen is convincing only to people who already believe that they do.

If you have some scientific evidence to back up your religious belief, you should probably present it.
Click to expand...
I mean, I posted some earlier to the thread. It's not a religious belief any more than the idea that I can assemble behavior from logic gates.

Jarhyn said:
time.com

How This Tool Could Decode AI’s Inner Mysteries

Even the creators of LLMs don't know exactly how they work. But scientists are making progress in finding out
time.com time.com
Click to expand...

I studied this; for me, it's academic.
 
We know how LLM work. I posted a youtube short video about that.
There is no thought process there whatsoever. It is designed to create appearance of that but there is nothing thinking there. Anyone who has not been in coma in the last few years should know that already. AI is in crisis, it's capabilities plateaued, available training data exhausted before it was able to reach any "thinking" level. New approach is badly needed.
 
barbos said:
We know how LLM work. I posted a youtube short video about that.
There is no thought process there whatsoever. It is designed to create appearance of that but there is nothing thinking there. Anyone who has not been in coma in the last few years should know that already. AI is in crisis, it's capabilities plateaued, available training data exhausted before it was able to reach any "thinking" level. New approach is badly needed.
Click to expand...
No, you apparently don't.

It isn't "designed" in any way to serve such an "appearance". It is "trained" to produce an "output" characteristic to some "input" and that's about all you can say of how they actually come into being.

Of course, it requires understanding about what 'thinking' is intended to invoke, and having an intent that doesn't void any recognition of it happening in computers, and given what I have seen of your understanding of "thought", let's just say I expected little and was disappointed only a little.
 
You must log in or register to reply here.
Back
Top Bottom