Hacker News Reader

Comment by gjitcbcz | Hacker News Reader

No context was given but here was his actual statement—-

> I can't foretell the future, but it is important to realize that ChatGPT is not artificial intelligence. It has no intelligence; it doesn't know anything and doesn't understand anything. It plays games with words to make plausible-sounding English text, but any statements made in it are liable to be false. It can't avoid that because it doesn't know what the words _mean_.

View on HN

His comments might have some weight if he actually provided evidence of his assertions. For instance he says "it doesn't know what the words _mean_." How do we indicate/prove that we know what a word means? We can give a definition. We can offer an example of how to use it in a sentence. We can explain and/or demonstrate how its meaning can vary depending upon context. Etc... ChatGPT can do all those things.

So, is there something more fundamental going on in our own brains? If so, what is it?

As far as false vs. true, humans also make many mistakes, confuse references, draw false conclusions drawn from unrelated evidence, and on and on. We also often get things right. ChatGPT is similar.

He just makes conclusory statements so it is just an opinion. I'm not saying any true intelligence has been demonstrated yet, but the progress that has been made in just the last year or two has been enormous. Does anyone doubt the progress won't continue? Something very important is happening and I personally think it is more than just "playing games with words." Just my opinion...

Meaning isn't defined in the brain, though it may effect and be effected by the contents of the brain.

Tracking the meanings of words over history shows some words stand still (rock, tree, bird), some words gain new meanings and lose older meanings, and some words change the underlying meaning entirely (silly, nice).

These words appear somewhere in history for a first time and are then repeated and modified over the centuries. Meaning-making of words is a cognitive, social and structural process that exists across society as a whole.

The key part is that they are never "proven", merely shown to exist and in use by people. Proving the meaning of a word, is to treat meaning like a science. Meaning cannot be reduced into the scientific process. It grows and changes organically, just as much as everything else does.

Got to say, I miss superhuman computing progress. Chess engines completely destroying chess players, even the world champions, was exciting. A human-equal chatGPT robot is marvelous and a huge step forward. But not "superhuman exciting"....

English isn't my first language. I learned a lot of words simply by inference when reading English texts. I used a lot of them before really I knew what they meant, because I had built my own understanding of the words through pure inference. Often that understanding was close to a words true meaning, other times not.

Clearly some meaning of words has been trained into the coefficients that make up ChatGPT, after all it's producing more eligible texts than several people I know...

A lot of people talk about what ChatGPT does. Few seem to talk about the data it was trained on. That data came from us, from people. So besides the algorithm, there is the training which is...what precisely? Everything on the web? Scientific papers? Everything in archive.org? I don't know. But what ChatGPT responds with is a reflection of people, which is a very interesting idea (its mistakes are our mistakes as reflected in our online speech?) People say ChatGPT "hallucinates", where did that come from?

It turns out that accurately predicting text using a model that's a couple orders of magnitude smaller than the inputs requires understanding what the words mean.

> It can't avoid that because it doesn't know what the words _mean_.

That’s true of a fair number of humans, too.

This is an uncharitable description of human understanding: your statement is simply not true in this context.

> your statement is simply not true in this context.

Thanks for clearing up the misconception with your cogent and well-reasoned reply.

> true of a fair number of humans, too

And your conclusion would be? If they show to have what it takes we will call them intelligent, otherwise unintelligent.

Some seem to want to cling to a weak definition of intelligence, covering all sentient beings in compassion - but we are engineering /tools/, things that have to be effective, "have what it takes"... This puts the "intelligence" of the narcoleptic carp in a special corner with little relevance to the context.

Reply on news.ycombinator.com