chatgpt

T

tOsYZYny

Has anyone played around with chatgpt, thoughts?

I used it a bit and it is pretty impressive in terms of being able to provide relevant answers to a broad range of topics and perform various tasks. As a software engineer by day, I am both scared and excited of the future. Even my role could be outsourced to AI it appears.
 
I am not so much blown away by chatgpt as my understanding is it obtains the knowledge from the internet to formalise the chats/answers.
What will happen when at some point we as in humans stop posting things back to internet.
Now there is a urge to report back to the community when the proposed answer works, but will it happen when the answer is only given by a machine.

But, when I looked at the images created by the AI, I am truly blown away. And that is not something you can compile from scraping the internet.
 
My thoughts:
When someone asks me a question, I don't try to say "what would typically be said". I try to say what I think is the truth. I get the impression ChatGPT is just taking the verbal "path of least resistance", like water flowing downhill. That's a major downside, and a difference in the way humans and computers (so far) think, and the human way is better.
Another problem (with the whole scraping technique) is that AI output has now been spewed all over the Internet. So future AIs, unless measures are taken, will presumably be trained on the output (deluded ravings?) of previous AIs. That could be a good thing, but it's not guaranteed. The AI could just get more and more divorced from reality.
I'd be impressed if an AI came up with an original idea that connected two previously unconnected fields. Like if an AI spotted some previously unnoticed pattern that underpins say, psychology and religion.
 
Because of my thoughts on Artificial --so called-- 'intelligence', I didn't further look at it. IMHO AI is overrated, but because it is over- or high rated its use will be spread by those who admire it.

Computers can process vast amounts of data and combinations, probably faster than humans can. In the same time computers can only do 'as instructed', and even when computers are able to make new instructions themselves these are based on instructions as given. Big question is if AI can come up with something _really_ new and unseen.

The ethical dimension is something we humans have to see develop, but we may be too late when some unethical AI-solution comes up and is already spread out all over the AI nerve system (remember the chatbots that became nazi within a matter of cycles). Besides the question who provided the ethical rules to AI for the first time? Todays view of the world will be different from tomorrows -- big question is whether humans and AI will shift narratives in a harmonious way.

AI-art IMHO is a major demonstration of human stupidity on what AI can do. Art, painting, sculpturing, writing, etc. is a profound human expression of being, and a far from rational process and experience. The joy of Art is both in creating as in seeing, reading and listening. In analogy, this is why there never will be self driving motorcycles -- it's just too much fun to ride it yourself! Even writing code to run a computer is a creative and fun process.
I can't think of any reason to automate that, deliberately stripping the human joy of life.

And scolars using AI, ChatGPT for writing their thesis don't understand the joy of learning and the joy of writing to develop yourself as a human being. Just passing the exam isn't enough, by far. For me.
 
meine said:
. In the same time computers can only do 'as instructed', and even when computers are able to make new instructions themselves these are based on instructions as given
Click to expand...
I was interested to read about "non-deterministic Turing machines" because the name sort of suggests that such a machine could make actual decisions. But it turns out that an NDTM just makes every possible decision at each juncture, branching into an exhaustive set of new machines. And an NDTM can be simulated on a TM. Making every decision seems to me no smarter than making just the one decision.
meine said:
Art, painting, sculpturing, writing, etc. is a profound human expression of being, and a far from rational process and experience. The joy of Art is both in creating as in seeing, reading and listening
Click to expand...
Agreed. It would seem however that beauty is to some extent quantifiable. Tigger (pictured :D) is a work of art. I recently read an article explaining that Manx cats are objectively more beautiful than Persian cats due to the distances between their facial features. And if you look at a picture of a Manx next to a Persian it's hard to disagree.
On the flipside, Google recently tried to get me to buy a £150 Chinese robot cat that is "a great companion for children and seniors". It made me terribly sad to think of a lonely, confused elderly person stroking a fake cat. An almost indescribable feeling that something has gone really wrong with the world.
 
I played around with it, even used it for rubber-duck debugging. Problem is it gives you false answers many times. You can't trust the answers it gives you.
One example:
Code: 
Q: Optimize subtracting 8 from a number with a shift operation

A: One way to optimize subtracting 8 from a number is by using a bit shift operation. In most modern CPUs,
bit shift operations are faster than subtraction operations.

You can use the ">>" operator to shift the bits of a number to the right. Each shift to the right by one position is equivalent to dividing the number by 2.
So, shifting the bits of a number to the right by 3 positions is equivalent to subtracting 8 from the number.
You'll notice very quickly the beginning of every answer is constructed from your question. Not to mention answer is complete BS.
Now this is something I do understand. What about stuff I have no clue?

But I get it, it's learning. That's why they opened it to public for free.
If you ask clever enough it can be used as interactive google search.
 
ProphetOfDoom said:
I recently read an article explaining that Manx cats are objectively more beautiful than Persian cats due to the distances between their facial features.
Click to expand...
Is that objective or 'statistically most liked by people who filled in the quenstionnaire'?

'The best restaurant in town' is not the one that got most five star ratings per se. If you don't like the taste of that <fill_in> cuisine, the rating is useless. Besides, 'the best' might be unknowingly distorted e.g. by the persons you went with. With a loved one a simple meal can be great, a good meal in rotten company can be awful. 'Objective' questionnaires and ratings aren't that fine grated.

Just take a look at people combining all 'best' and 'most popular' dress, beard trim, spectacles, shoes, T&A. They look like clowns.

Objective as in Natural Science (temperature, density, etc) is very different from 'objectivated' in Human Science. Human Sciences fell in the pit playing 'natural science', neglecting unicity and diversity. Messing up correlation and causation did the rest. Besides that statistics is a way to review observations done, and not a way to predict the future. Models only give insight in possible future outcomes, but only within the boundaries set by the maker. What when the maker neglected something, or something that is very hard to measure?

When AI presents 'objectivated outcomes' based on crude rating and correlation only, we're up shit creek. And we already are, because this huge failure is used all around us allready.
 
en.m.wikipedia.org

Golden ratio - Wikipedia

en.m.wikipedia.org en.m.wikipedia.org
It was based on the assumption that the golden ratio is strongly correlated with beauty.
So, objective within the context of that. Everything has to be within the context of something.
IIRC it was done by measuring cats' faces rather than questioning people about cats.
 
ProphetOfDoom said:
en.m.wikipedia.org

Golden ratio - Wikipedia

en.m.wikipedia.org en.m.wikipedia.org
It was based on the assumption that the golden ratio is strongly correlated with beauty.
So, objective within the context of that. Everything has to be within the context of something.
IIRC it was done by measuring cats' faces rather than questioning people about cats.
Click to expand...
Interesting article about the Golden Ratio, TNX!

Of course some ratio are more appealing --'more beautiful'-- to the eye than others -- for the majority (!) of people. It is human to search for the math behind things, even the unexplainable. But IMHO the beauty of things is something deeper than just correlate with a mathematical ratio and/or lots of people who have that opinion.

The techno- and rationalization of our world and thinking also might fall in the pit that correlation of personally perceived beauty and the math that describes so, isn't the same per se.

Could 'the smell of the colour nine' be an 'objectively' nice or bad odour?

[BTW, nice philosophical discussion!]
 
I've read that ChatGpt is political af.

I don't know if it's true but here in Germany someone asked it to write a "positive poem" about a certain right-wing politican.

ChatGpt refused to do so because of the "controversial nature" of said politican.

Then it has been asked the same question but this time with someone of the Green Party.

This time it was ok.
 
eternal_noob said:
I've read that ChatGpt is political af.
Click to expand...
Could be the 'ethical' part of it, making it questionable the ethics of whom it is.

Apart from that in a creative an satirical way, it can be quite OK to write such a poem. Wonder if AI can understand satirical proze. Mr. Data from StarTrek had some trouble understanding human humour.

[US-based AI would probably skip my English `ou`, as redmondisized spelling checkers already do]

[`redmondisized` -- unknown adjective, not in dictionary. Do you mean Slytherin?]
 
Let me know when a version comes out that can write a Device Tree Overlay for Hummingboard Pro to enable i2c3 on FreeBSD.

(Hint:It can't as overlays are broke on Hummingboard and crash it on startup.)
Obviously AI would know that and would show how to edit the boards dtb directly to suit.

Yup I will be 10ft under before it gets there....
 
I think google search algorithms are more impressive.

For vaporware I refer to:
-Quantum-computing
-Transmission of data using entanglement.
 
_martin said:
You'll notice very quickly the beginning of every answer is constructed from your question. Not to mention answer is complete BS.
Now this is something I do understand. What about stuff I have no clue?
Click to expand...
This is intentional, to mimic human conversation... and it looks like ChatGPT can run for POTUS already... Current POTUS will parrot back some of your question, his answer would be complete bullshit, and betray the fact that he has no clue what you just asked him. Just watch media sessions on TV... 😩
 
In the little bit I have interacted with it, I feel I get better results with it than Google. It was interactive. But yes, it isn't perfect and the amount of computing power it requires is insane. But yes, to that end, maybe it won't replace 'us', but instead it would be another tool in which we can leverage. Perhaps we can use it to do the mundane stuff like write unit tests or cleaning up code.
 
My opinion is that maybe Elon should focus on providing what his customers paid for.
Full Self Driving.
The name is not difficult to comprehend.
Mercedes came later to the game and has met Nevada DOT standard for level three autonomous.
Elon?
Buying twitter and doing dumb things with AI.

Starship could not launch because FAA approvals. Well, its been 3 months.
You going to fly that thing or tweet with ChatGPT all day?
 
ChatGPT is great tool no matter what you guys say. A hammer is great tool, you can use it to knock down nails or whatever you need it to do, you can also use it to break in somewhere illegally, you can also use it to kill someone. Tools are tools, they have their purpose, you can use it for good or for bad.
 
You must log in or register to reply here.
Back
Top