Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > sci.physics.relativity > #617436

Re: ChatGPT

Cross-posted to 2 groups.

Show all headers | View raw

On 5/28/2023 6:17 PM, Physfitfreak wrote:
> On 5/28/2023 3:31 AM, Physfitfreak wrote:
>> On 5/28/2023 12:22 AM, The Starmaker wrote:
>>> Physfitfreak wrote:
>>>>
>>>> On 5/27/2023 9:40 PM, The Starmaker wrote:
>>>>> Physfitfreak wrote:
>>>>>>
>>>>>> On 5/27/2023 8:45 PM, The Starmaker wrote:
>>>>>>> ChatGPT, an online chatbot created by OpenAI, has shown success
>>>>>>> passing several national benchmarking exams, including the SAT, the
>>>>>>> GRE and a bar exam. It has even shown the ability to pass the U.S.
>>>>>>> Medical Licensing Exam...
>>>>>>
>>>>>> Bullshit.
>>>>>>
>>>>>> I gave it a simple but tricky (to some idiots) math question and it
>>>>>> failed!..
>>>>>>
>>>>>
>>>>> What was the "simple but tricky (to some idiots) math question"?... so
>>>>> that others here can confirm if
>>>>> what you say is true or not.
>>>>>
>>>>>
>>>>> If you don't post it here...then that confirms it isn't true.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>> Hehe :) I can make you interested anytime I wish no matter how you try
>>>> to weasel your way around it.
>>>>
>>>> Telegram, for a limited period, allowed I believe less than 10 
>>>> questions
>>>> be asked from ChatGPT4 (latest version back then) free of charge. It
>>>> isn't free anymore.
>>>>
>>>> I asked it: "If you have 100 pounds of potatoes, knowing that they are
>>>> 99% water (weight percent) and let them dry in air long enough to get
>>>> them down to 98% water, how much do they weigh now?"
>>>>
>>>> The AI's answer reflected the same mistake that fools make :) Even 
>>>> after
>>>> telling it the answer was wrong, it kept making a slightly different
>>>> mistake. The poor thing was just like Alsing :-))) Couldn't really
>>>> think. It just searched and searched in other areas each time for the
>>>> answer. So about four of my questions got blown for nothing. But I got
>>>> to understand its abilities, what it could really do and what it was
>>>> helplessly not made for.
>>>
>>> Maybe your problem is...you need to Learn to Prompt.
>>>
>>>
>>> Here is the answer ChatGPT ver 3 free version gave:
>>>
>>> .................chatgpt........................................................
>>>
>>> To solve this problem, we need to calculate the weight of water in 
>>> the original 100 pounds of potatoes and then determine the weight of 
>>> potatoes with 98% water content.
>>>
>>> Let's start by finding the weight of water in 100 pounds of potatoes 
>>> with 99% water content:
>>>
>>> Weight of water = 99% of 100 pounds
>>> = (99/100) * 100 pounds
>>> = 99 pounds
>>>
>>> Now, let's calculate the weight of potatoes (solid matter) in the 
>>> original 100 pounds:
>>>
>>> Weight of potatoes = Total weight - Weight of water
>>> = 100 pounds - 99 pounds
>>> = 1 pound
>>>
>>> Since we want to find the weight of potatoes when the water content 
>>> is reduced to 98%, we can set up the following equation:
>>>
>>> Weight of potatoes = (Weight of potatoes + Weight of water) / (1 - 0.98)
>>>
>>> Let's calculate the weight of potatoes with 98% water content:
>>>
>>> Weight of potatoes = (1 pound + 99 pounds) / (1 - 0.98)
>>> = 100 pounds / 0.02
>>> = 5000 pounds
>>>
>>> Therefore, when the potatoes are dried to 98% water content, they 
>>> would weigh approximately 5000 pounds.
>>>
>>>
>>>
>>>
>>>
>>>
>>
>>
>> Hehe :) As you see!
>>
>> AI's problem is, it cannot think for itself. It can only refer to near 
>> endless sources of information based on its take of the task you give 
>> it. But that does _not_ equal thinking. Ability to analyze is 
>> something beyond having all the information in the world accessible to 
>> you and knowing the English language. A library cannot think.
>>
>> The correct answer to the question is of course 50 pounds. When one 
>> pound of dry matter was 1% of total weight before, then having the 
>> same 1 pound of dry matter as 2% of total weight requires that the 
>> total weight be half of what it was before, i.e. 50 pounds. But AI 
>> cannot do this reasoning cause it is an Alsing type of equipment. You 
>> can have an Alsing endlessly toil for you in various other ways, but 
>> thinking is not one of them.
>>
>> By the way, if the AI you tried is free to use, then it is not even as 
>> good as the one I tried some months back.
>>
>>
>>
> 
> I just used the free version of the AI online. This is what it gave me:
> 
> ---------------------------------------------------------------
> To solve this problem, we need to calculate the weight of the potatoes 
> after the water content decreases from 99% to 98%.
> 
> Let's assume the initial weight of the potatoes is 100 pounds, and their 
> water content is 99%. This means that 99% of the weight is water and 1% 
> is the dry matter.
> 
> Weight of water = 99% of 100 pounds = 99 pounds
> Weight of dry matter = 1% of 100 pounds = 1 pound
> 
> Now, let's calculate the weight of the potatoes when their water content 
> decreases to 98%.
> 
> New weight of dry matter = Weight of dry matter / (1 - final water content)
> = 1 pound / (1 - 0.98)
> = 1 pound / 0.02
> = 50 pounds
> 
> The weight of the water will remain the same since it's not evaporated. 
> Therefore, the weight of the potatoes after drying to 98% water content is:
> 
> Weight of dry matter + Weight of water
> = 50 pounds + 99 pounds
> = 149 pounds
> 
> ---------------------------------------------------------------------
> 
> As you can see, it cannot keep the logic for long. As soon as its 
> attention shifts, the line of logic suffers when it comes back to the 
> previous point. So it can sure start on the right path, but commits 
> simple mistakes in logic. It's nothing beyond an Alsing. At best, some 
> hit-and-miss equipment.
> 
> 
> Now why is that? Let's be a physicist and take a look. The developers 
> have jumped logical steps early on, banking on the assumption that 
> intelligence is some emerging process. Well, tough luck. It is not.
> 
> This project hasn't worked so far because bullshit doesn't get you 
> anywhere. Bullshit only works in the business of fooling people to make 
> money. It does not work in science.
> 
> The very first step of development of it has thus had a flaw :) What 
> _was_ the very first step, you might ask. It is the same step that 
> Backus (creator of FORTRAN language) demanded around 1960 to get done 
> and never saw it fully done. It was only half done, which is what now we 
> know as FORTRAN language.
> 
> Backus was very irritated by the extent of spoon-feeding work required 
> in the assembly language. So he aimed at a language where you'd give the 
> necessary information to it and what you want done (and knowing in 
> advance that with that information that purpose can be fulfilled by a 
> programmer), then would ask something like, "Now create a code to 
> implement that purpose." He wanted the compiler to take care of all the 
> little steps and get the code correctly written in the machine language. 
> Well, this never got done! Instead of going towards achieving this goal, 
> the fucking "engineers" kept repeating the same shit over and over in 
> the form of newly made programming languages that did nothing but 
> repeating the already existing abilities. The task got shifted from the 
> hands of scientists and fell into the hands of "engineers" :-( So now we 
> have 750 different programming languages, and not _one_ of them does 
> what Backus wanted.
> 
> In the interview he gave 6 months before his death, he expressed his 
> bitter disappointment in the way programming has "developed." In fact, 
> programming never developed beyond the FORTRAN that he had created in 
> early 60s. Nobody took one step beyond that.
> 
> Now these "AI" guys want to jump ahead of the horses. They want to have 
> their programs do that without solving this fundamental issue first. 
> Their idea is, "Just give it all the information in the world and train 
> it by telling it what is right in various different situations". As if 
> by some magic, the program itself - THAT SAME FORTRAN! - can somehow 
> learn to do that. But this is bullshit of course.
> 
> So when the day comes that what Backus wanted is achieved (which is 
> achievable indeed), only then, the process can successfully continue. 
> But "engineers" don't and won't and cannot do that. Mathematicians and 
> scientists can. The departments of "natural philosophy" can do it, not 
> your fucking "engineering" schools.
> 
> 
> 
> 
> 
> 
> 



By the way, it is fair to add that Backus later did the math required 
for taking that vital next step forward in creating and increasing 
actual intelligence in a programming language (i.e. making it more and 
more so-called "high-level"). But by then (late 1960s) millions of 
"engineers" had hijacked programming languages for making money off of 
them. A "cook" doesn't care about the science of nutrition. He just 
looks up the recipes and follows the steps and boom - his food is ready, 
it's time for him to get paid and go home.

So nobody in 1969 (I think that's when he got it done) even understood 
Backus's vital work aimed at that goal. In fact, only one mathematician 
in the world ever studied his work fully and to some extent participated 
as well. And Backus tried and tried from then on for decades to put this 
into the minds of those who should've taken such concepts and 
implemented them in the programming languages. He never succeeded. 
Businesses and governments and scientific institutions packed with 
"engineers" instead of scientists had gotten their "FORTRAN" and were 
happily cooking meals with it. They didn't want anything more. And to 
this day, 63 years later, any programming language you pick, you see the 
same old FORTRAN disguised, thanks to individuals with nothing in mind 
other than paying their mortgages with it.

So Backus did his part, both in the first vital steps (creating 
FORTRAN), and more so in about 10 years of additional work thereafter, 
focusing on adding "intelligence" to the language enough to at least 
successfully handle his original 1960 question, or demand. But nobody 
else did their parts. Instead of doing it, they would only give him 
"rewards" :-( And very soon it was Carter era, the beginning of the end 
of the USA as a first world nation, culminating in the full-speed 
descent with Reagan and whoever came after to lead and deal with such 
matters and opportunities for advancement of humans.

Hehe :)

Dark Ages happen, don't they :) ... Just leave your science to your 
"engineers." To "cooks". Leave your Churches in the hands of prostitutes.







-- 
This email has been checked for viruses by Avast antivirus software.
www.avast.com

Back to sci.physics.relativity | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

Re: ChatGPT The Starmaker <starmaker@ix.netcom.com> - 2023-05-27 18:45 -0700
  Re: ChatGPT Physfitfreak <Physfitfreak@gmail.com> - 2023-05-27 21:05 -0500
    Re: ChatGPT The Starmaker <starmaker@ix.netcom.com> - 2023-05-27 19:40 -0700
      Re: ChatGPT Physfitfreak <Physfitfreak@gmail.com> - 2023-05-27 22:03 -0500
        Re: ChatGPT The Starmaker <starmaker@ix.netcom.com> - 2023-05-27 22:22 -0700
          Re: ChatGPT Physfitfreak <Physfitfreak@gmail.com> - 2023-05-28 03:31 -0500
            Re: ChatGPT Maciej Wozniak <maluwozniak@gmail.com> - 2023-05-28 02:09 -0700
            Re: ChatGPT Physfitfreak <Physfitfreak@gmail.com> - 2023-05-28 18:17 -0500
              Re: ChatGPT Physfitfreak <Physfitfreak@gmail.com> - 2023-08-12 01:13 -0500
                Re: ChatGPT Physfitfreak <Physfitfreak@gmail.com> - 2023-08-19 08:01 -0500
                Re: ChatGPT Physfitfreak <Physfitfreak@gmail.com> - 2023-08-19 08:32 -0500
                Re: ChatGPT Prokaryotic Capase Homolog <prokaryotic.caspase.homolog@gmail.com> - 2023-08-19 06:49 -0700
                Re: ChatGPT Prokaryotic Capase Homolog <prokaryotic.caspase.homolog@gmail.com> - 2023-08-19 06:42 -0700
                Re: ChatGPT Python <python@invalid.org> - 2023-08-19 15:51 +0200
                Re: ChatGPT Prokaryotic Capase Homolog <prokaryotic.caspase.homolog@gmail.com> - 2023-08-19 12:05 -0700
                Re: ChatGPT Prokaryotic Capase Homolog <prokaryotic.caspase.homolog@gmail.com> - 2023-08-19 12:12 -0700
  Re: ChatGPT "Chris M. Thomasson" <chris.m.thomasson.1@gmail.com> - 2023-05-27 21:16 -0700

csiph-web