https://www.zdnet.com/article/meta-...l-ai-model-to-compete-with-openai-and-google/

Since it's an open-source AI technology, commercial access to LLaMA gives businesses of all sizes the opportunity to adapt and improve the AI, accelerating technological innovation across various sectors and potentially leading to more robust models.

Meta's LLaMA is available in 7, 13, 33, and 65 billion parameters, compared to ChatGPT's LLM, GPT-3.5, which has been confirmed to have 175 billion parameters. OpenAI hasn't said how many parameters GPT-4 has, but it's estimated to have over 1 trillion parameters -- the more parameters, the better the model can understand input and generate appropriate output.

OMg, there's the magic word "understand" being used again.

 

No, it's not a simulation of intelligence. It's a regurgitation of human intelligence by a system not designed or built to simulate intelligence. It index's, stores and retrieves processed human intelligence.

Thus, the "brute-force" to which i referred. Force, by definition, is the antithesis of intelligence.

 

Thus, the "brute-force" to which i referred. Force, by definition, is the antithesis of intelligence.

Brute-force usually means inefficient, not non-existent. Force, like aggression, can be directed intelligently or it can be wasted. A LLM doesn't' Brute-force, it uses sufficient power for the functionality it provides. That functionality is not intelligence in any form.

 

That functionality is not intelligence in any form.

Fundamentally, I'm not disagreeing with you.

The word "intelligence" is being diluted by those who insist AI is artificial intelligence.

 

Fundamentally, I'm not disagreeing with you.

The word "intelligence" is being diluted by those who insist AI is artificial intelligence.

These systems seem intelligent because they work on languages, pictures, graphics, which we link to intelligence. They do have a superpower, the prediction (using massive amounts of human pre-processed examples) of the next word to fool a human intelligence.

Does a baby need hundreds or even tens of exemplar faces to train it to recognize its mothers face? No, because a baby has intelligence.

 

They do have a superpower ... to fool a human intelligence.

And said superpower will eventually pass a Turing Test ... which IMHO is not truly a scientific test, but simply a metric of human perception.

 

https://www.zdnet.com/article/meta-...l-ai-model-to-compete-with-openai-and-google/


OMg, there's the magic word "understand" being used again.

which is back to the original question
how do we define and measure intelligence ?
people used to think only humans were intelligent,
then they found apes doing the same things, so they are now classed as intelligent
then dolphins and elephants,
even gold fish have showed intelligence

Is a machine that can answer questions on these forums intelligent ?
am I or anyone else on here intelligent ?
how do we measure and prove it ?

once we know that , we can use the measure to classify ,

 

which is back to the original question
how do we define and measure intelligence ?
people used to think only humans were intelligent,
then they found apes doing the same things, so they are now classed as intelligent
then dolphins and elephants,
even gold fish have showed intelligence

Is a machine that can answer questions on these forums intelligent ?
am I or anyone else on here intelligent ?
how do we measure and prove it ?

once we know that , we can use the measure to classify ,

It's pretty easy.
Any way you want to scientifically measure and define intelligence you will find that Large Language Models don't fit that definition unless a spreadsheet is also intelligent. In other words if you make the definition stupidly broad to cover them then anything can be intelligent.

 

It's pretty easy.
Any way you want to scientifically measure and define intelligence you will find that Large Language Models don't fit that definition unless a spreadsheet is also intelligent. In other words if you make the definition stupidly broad to cover them then anything can be intelligent.

What do you suggest @nsaspook as the test ?
Don't think a spread sheet could reply to forums , so is an ai that can reply intelligent

 

What do you suggest @nsaspook as the test ?
Don't think a spread sheet could reply to forums , so is an ai that can reply intelligent

Why couldn't a spreadsheet reply to forums? Spreadsheets go out and scrape screens and query databases and post results. Nothing really different.

 

Why couldn't a spreadsheet reply to forums? Spreadsheets go out and scrape screens and query databases and post results. Nothing really different.

Didn't think of that.
So how would one tell that a spreadsheet that replied to forums was intelligent or not ? What test would one use ?

 

Didn't think of that.
So how would one tell that a spreadsheet that replied to forums was intelligent or not ? What test would one use ?

I haven't been closely following the thread, but I think that this has probably been mentioned -- the first issue is defining what it means for something to be "intelligent" in this context. People often throw around the Turing Test as if passing it means that something is intelligent. But there are a couple of issues with this. First, the Turing Test was never intended to test if something was "intelligent", but merely whether it was exhibiting "intelligent behavior". The question of whether something has to be intelligent in order to exhibit intelligent behavior is not addressed. Second, the Turing Test wasn't really a test devised to evaluate if something exhibited intelligent behavior according to some already-accepted definition of what such behavior was, but rather an attempt to define what such behavior was -- essentially, any behavior that passes the Turing Test is considered to be intelligent behavior by definition. That leaves completely up for debate whether it's a good and/or useful definition.

I don't think it's possible to come up with a single, all-purpose definition of intelligence. I think that it really matters what the context of the specific discussion is. Whenever someone mentions the word "intelligence", the first response should probably be, "in what sense?"

 

I haven't been closely following the thread, but I think that this has probably been mentioned -- the first issue is defining what it means for something to be "intelligent" in this context. People often throw around the Turing Test as if passing it means that something is intelligent. But there are a couple of issues with this. First, the Turing Test was never intended to test if something was "intelligent", but merely whether it was exhibiting "intelligent behavior". The question of whether something has to be intelligent in order to exhibit intelligent behavior is not addressed. Second, the Turing Test wasn't really a test devised to evaluate if something exhibited intelligent behavior according to some already-accepted definition of what such behavior was, but rather an attempt to define what such behavior was -- essentially, any behavior that passes the Turing Test is considered to be intelligent behavior by definition. That leaves completely up for debate whether it's a good and/or useful definition.

I don't think it's possible to come up with a single, all-purpose definition of intelligence. I think that it really matters what the context of the specific discussion is. Whenever someone mentions the word "intelligence", the first response should probably be, "in what sense?"

" the first issue is defining what it means for something to be "intelligent" in this context "
I think thats the nub of the "problem"
the forum seems to break into those that say AI can never be "intelligent" , and those that say "it can be", and those that ask, how do we prove / measure it one way or another

I wonder if "the Greeks" or whatever used to ask the same as to were animals intelligent or just repeating what we had taught them ?
to me it sounds the same question for AI

Now some have started saying on the forum this or that type of AI "can never be intelligent"
which is may be a bit more easy to demonstrate,
but the AI I know is never of one type,

 

Well
https://www.theregister.com/2023/07/20/gpt4_chatgpt_performance/

 

Well
https://www.theregister.com/2023/07/20/gpt4_chatgpt_performance/

LoL
It's all about the Benjamin's. Free good dope to spread the word, then watered down crap for sale once they are hooked.

 

Google Tests A.I. Tool That Is Able to Write News Articles

So....what's the difference?

Perhaps it's a more honest approach.

 

https://www.zdnet.com/article/meta-...l-ai-model-to-compete-with-openai-and-google/


OMg, there's the magic word "understand" being used again.

We need to be a bit careful about how we react to the use of words like this. Humans have a long history of anthropomorphizing anything and every thing we can. We talk about opamps "wanting" to hold the voltage difference between the inputs to near zero or about a system "knowing" what the temperature is based on the voltage on a thermistor. So it's not unreasonable to talk about a system "learning" something or coming to "understand" something. The issue is that (usually) no one in a discussion about opamps or temperature controllers is going to misinterpret those descriptions and think that anyone is actually claiming that an opamp actually "wants" anything or that a temperature controller actually "knows" anything. But with AI, one person uses these kinds of terms in a perfectly reasonable way and there is no shortage of other people that seem willing to go out of their way to misinterpret them in a literal way.

 

I agree but in this case of the use of the word is more that just anthropomorphizing, it's IMO being used to 'sell' a product using property's the product doesn't have. That's as IMO bad as Tesla calling their level 2 driving technology , Full Self Driving.

Sometimes words matter, like when there's a likely threat of harm from the product.

https://www.drive.com.au/news/tesla-full-self-driving-level-2-autonomous/

Tesla has admitted its controversial ‘Full Self-Driving’ semi-autonomous driving technology is less capable than similar systems developed by its rival car-makers

https://www.iotworldtoday.com/trans...-steering-wheel-weights-used-to-trick-teslas-

Amazon Pulls Steering Wheel Weights Used to Trick Teslas

 

Now that I've been reading that ChatGPT's performance has been decreasing over time. I think that it proves that the darn thing has no understanding of what it is that it's supposed to do in the long term.
I mean, it was fed a bunch of organized information by its "handlers" in such a way that it developed patterns to follow. But now that it's been left on its own, it has absolutely no idea of what direction the "knowledge" that it's been absorbing is supposed to grow into.

Only humans know what it would actually want it to do. It follows that the software must be kept under human supervision at least for the time being.

 

Now that I've been reading that ChatGPT's performance has been decreasing over time. I think that it proves that the darn thing has no understanding of what it is that it's supposed to do in the long term.
I mean, it was fed a bunch of organized information by its "handlers" in such a way that it developed patterns to follow. But now that it's been left on its own, it has absolutely no idea of what direction the "knowledge" that it's been absorbing is supposed to grow into.

Only humans know what it would actually want it to do. It follows that the software must be kept under human supervision at least for the time being.

Hi
you say " no understanding "
is it not analogous to "a child" , they learn from everything they are plumed with,
and that can lead to better or worse "responses" over time
But never say a child is not intelligent,

Which comes around again to the fact we have no definition on what counts as "intelligent" and how we measure it