The AM Forum
March 29, 2024, 11:59:44 AM *
Welcome, Guest. Please login or register.

Login with username, password and session length
 
   Home   Help Calendar Links Staff List Gallery Login Register  
Pages: [1]   Go Down
  Print  
Author Topic: CHATGPT is Scary!  (Read 4604 times)
0 Members and 1 Guest are viewing this topic.
WA2SQQ
Contributing
Member
*
Offline Offline

Posts: 1090


« on: March 17, 2023, 10:02:33 AM »

Have any of you tried CHATGPT? I decided to throw a few questions at at. The responses really shocked me. This technology will replace jobs, and it makes Google look like a mentally challenged person. I’m not a math wiz, so I’m curious of anyone sees any obvious errors.

QUESTION “Design a matching network to resonate a 21 ft antenna to operate at 7.285 mhz using an impedance of 50 ohms”

ChatGPT Responded

To design a matching network to resonate a 21 ft antenna at 7.285 MHz with an impedance of 50 ohms, we can use the following steps:
1.   Calculate the length of the wire required to resonate at 7.285 MHz using the formula:
L = 468 / f
where L is the length of the wire in feet, and f is the frequency in MHz.
Substituting the values, we get:
L = 468 / 7.285 = 64.25 ft
Since we have a 21 ft wire, we need to use some additional elements to create a resonant circuit.
2.   Choose a matching topology that can match the 50 ohm impedance to the antenna impedance. One common matching network is the L-network, which consists of a series inductor and a shunt capacitor. We can use an online calculator or a Smith chart to determine the required values of the inductor and capacitor.
3.   Determine the reactance of the 21 ft wire at 7.285 MHz using the formula:
X = 2πfL
where X is the reactance in ohms, f is the frequency in Hz, and L is the length of the wire in meters.
Substituting the values, we get:
X = 2π * 7285000 * (21 / 3.281) = 278.47 ohms
The reactance is capacitive, so we need to add an inductor to cancel it out.
4.   Calculate the required inductance using the formula:
•  L = X / (2πf)
Substituting the values, we get:
L = 278.47 / (2π * 7285000) = 0.00215 H
We can use a 2.2 μH inductor with a tolerance of at least 10%.
•  Calculate the required capacitance using the formula:
C = 1 / (2πf * Xc)
where Xc is the desired capacitive reactance, which is equal to the impedance minus the reactance of the inductor.
Xc = 50 - (-278.47) = 328.47 ohms
Substituting the values, we get:
•  C = 1 / (2π * 7285000 * 328.47) = 6.17 pF
We can use a 6.8 pF capacitor with a tolerance of at least 10%.
•  Connect the inductor and capacitor in an L-network as shown below:
----(L)----(antenna)---- | (C) | GND
The L-network should be placed as close to the antenna as possible.
•  Verify the match using an antenna analyzer or a SWR meter. Adjust the values of the inductor and capacitor as necessary to achieve the desired resonance and impedance match.

Logged
WA2SQQ
Contributing
Member
*
Offline Offline

Posts: 1090


« Reply #1 on: March 17, 2023, 10:06:27 AM »

Question #2

"What are the specifications for the modulation transformer used in a Viking Valiant transmitter"

CHATGPT RESPONDED

The Viking Valiant transmitter was a popular amateur radio transmitter produced by the Johnson Company in the 1950s and 1960s. It uses a plate modulation scheme with a high-voltage transformer to modulate the RF output. The transformer used in the Valiant transmitter is known as the "Valiant modulation transformer" or the "J-237".

Here are the specifications for the transformer used in the Viking Valiant transmitter:

    Primary impedance: 50,000 ohms
    Secondary impedance: 5,000 ohms
    Power rating: 175 watts
    Frequency response: 20 Hz to 20 kHz (±1 dB)
    Maximum input voltage: 2,500 volts RMS
    Maximum output voltage: 250 volts RMS

The transformer has a laminated core and is designed to operate at high voltages and currents. It is used to couple the audio signal from the modulation source to the RF output, and to provide the necessary impedance matching and isolation between the circuits. The Valiant modulation transformer is a critical component in the Valiant transmitter, and its specifications must be carefully matched to ensure proper performance and reliability.
Logged
WA2SQQ
Contributing
Member
*
Offline Offline

Posts: 1090


« Reply #2 on: March 17, 2023, 10:17:41 AM »

Question #3

"Design the input matching network for a 3-500Z tube, operating at 3.8 mhz, using an input impedance of 50 ohms"

CHATGPT RESPONDED

To design an input matching network for a 3-500Z tube operating at 3.8 MHz with an input impedance of 50 ohms, we can use the following steps:

    Determine the plate impedance of the 3-500Z tube at 3.8 MHz. This can be found in the tube's datasheet or by using a plate impedance calculator. For a typical 3-500Z tube operating at 3.8 MHz, the plate impedance may be around 1500-2000 ohms.

    Calculate the required input impedance transformation ratio using the formula:

    Transformation ratio = √(Plate impedance / Input impedance)

    Substituting the values, we get:

    Transformation ratio = √(1500 / 50) = 7.746

    This means that the input matching network needs to provide a step-up transformation ratio of approximately 7.746.

Choose a matching topology that can provide the required impedance transformation. One common matching network is the L-network, which consists of a series inductor and a shunt capacitor. We can use an online calculator or a Smith chart to determine the required values of the inductor and capacitor.

Calculate the reactance of the input impedance at 3.8 MHz using the formula:

X = 2πfL

where X is the react (error in body stream)

So this question resulted in a system error, but the web site clearly says that version 1 still has some bugs
Still, so far it looks ligit

Logged
KC2ZFA
Member

Offline Offline

Posts: 441



« Reply #3 on: March 17, 2023, 10:35:37 AM »

in the first example the bot assigns to L units of length at one point and units of inductance at another point.
Logged
WD8BIL
Contributing
Member
*
Offline Offline

Posts: 4409


« Reply #4 on: March 17, 2023, 10:52:24 AM »

"L = 278.47 / (2π * 7285000) = 0.00215 H
We can use a 2.2 μH inductor with a tolerance of at least 10%."

0.00215 H = 2150uH
Logged
W1ITT
Member

Offline Offline

Posts: 573


« Reply #5 on: March 17, 2023, 04:14:29 PM »

Question 3 appears to be asking for the Input network for the amplifier.  The algorithm then takes off and works on the output network. 
This is where human common sense and experience is still needed, at least for a while.  Most experienced humanoids will allow that the Output impedance would be in the vicinity of 1500 to 2000 ohms, but that the Input impedance is generally taken to be in the vicinity of 100 ohms.  And any humanoid solution that wasn't in that neighborhood would cause some head scratching.
And who uses L-networks in a 3-500 output anyway?  Maybe at 6 meters using the plate output C with a variable L......  But that's still a Pi-net.
Things in the algorithm world will get better.  I have some darn good antenna programs now.  I recall back when antenna analysis programs came out, fellows would stay up all night "designing" antennas that claimed to have super gain until it was pointed out that the unworkably low impedances in some of the sections would have incurred losses that ate up all the gain.  A few decades back, at a company where I was employed, we had a couple of these whizbang new EE grads who were certain they could eliminate the test and field engineering departments and just ship parts in a box straight to the site based on software magic.  They didn't.  But I'd be tempted to use GPT if I had to write a paper on Jane Austen for Englsh class, or some other such useless exercise.
Now a retired guy, a burden on society, I'm not worried.  But it's fun to watch.
73 de Norm W1ITT
Logged
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« Reply #6 on: March 17, 2023, 07:37:45 PM »

I believe the errors here are mainly caused:
1.) because it is an early AI and still learning to try to figure out what people's complicated questions mean, and is not used to lengthy or complex questions that require many variables to be stated.

2.) or not 'knowing' that it should ask for very detailed questions, perhaps because it is not programmed to understand all of the ways something like the 3-500Z question could be answered. (the circuit could be grounded grid or grounded cathode, a variety of matching networks could be used, certain Q values or bandwidths could be wanted, and the operating voltages and currents are not stated, so the grid or cathode impedance, voltage, current, drive power is not calculated properly ior the question isn't 'understood').

3.) because it does not 'know' that the imprecisely stated question is imprecise.

4.)  because the possible solutions to the question are different depending on how the question itself is interpreted.

The program may have been confused by:

in question 3, not having access to the specifications in the tube manual and thereby having the opportunity to either state an answer based on one of them or to ask the supplicant for those quantities and using a known model or the data from the characteristic curves.

in question 2, it was not supplied with the operating voltages, powers, currents of the tubes in use, nor with the transmitter's frequency response (which it could simply parrot from the manual instead of assuming hi-fi standards and spewing hi-fi assumptions)

In question 1, it apparently missed the velocity factor if the wire, though that is minor. It may have made a math error? I don't know. Maybe someone can run the calculations it suggested and find out why it apparently made an error.


I did ask it (via someone else's account because I won't intentionally give my cellphone numbers out to random AIs) to explain the relationship between the Higgs boson and property of mass, and it did a decent job against which I could not argue.

I've been told that there is a limit to the amount of thought-time it's allowed to devote to an inquiry in the free interface and will time out or quit session if that's exceeded, and that it is said to hand off queries to another processing unit (whatever that means) in mid-calculation. It definitely takes a moment to formulate a response to a significant query unless the rote answer is well-published.
Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
KF9CM
Member

Offline Offline

Posts: 78



WWW
« Reply #7 on: March 18, 2023, 09:53:26 AM »

I asked CHATGPT one question, "Do you have feelings ?", it replied , "No I don't but, I hope to have some in the future." This is what we have to worry about, DON'T GIVE THEM FEELINGS! Remember Data and Lore?

73 de, Gary KF9CM
Logged

73 de Gary, KF9CM




www.kf9cm.com
Jim/WA2MER
Contributing
Member
*
Offline Offline

Posts: 298



« Reply #8 on: March 18, 2023, 03:03:13 PM »

Whatever can be done, will be done. We will be assimilated. Resistance is futile.
Logged

Anything worth doing is worth doing to excess.
Since you have to die anyway, you might as well die from something you like.
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« Reply #9 on: March 18, 2023, 06:02:17 PM »

Dr Frankenstein part II he creates the expert-devil instead of a brain damaged monster.
Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
WA1QHQ
Member

Offline Offline

Posts: 111



« Reply #10 on: March 19, 2023, 04:46:49 AM »

The values for the matching network will only work if it is placed at the antenna feed point, otherwise the length of coax between the antenna and the transmatch will have to be factored in.

qhq
Logged
WA2SQQ
Contributing
Member
*
Offline Offline

Posts: 1090


« Reply #11 on: March 20, 2023, 08:48:32 PM »

This movie is a preview of coming attractions.
Colossus The Forbin Project
https://youtu.be/u2cISMlN-YE
Logged
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« Reply #12 on: March 21, 2023, 10:53:22 PM »

Revelation 13:15

just in case some people don't want to sleep tonight.

The TV set wasn't/isn't the enemy (debatable)

but an uppity AI is an unpleasant AI.
Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
W3SLK
Member

Offline Offline

Posts: 2651

Just another member member.


« Reply #13 on: March 27, 2023, 01:03:56 PM »

Now the 'main-stream-media' has caught a whiff of it: https://www.foxbusiness.com/media/chatgpt-who-what-behind-artificial-intelligence-tool-changing-tech
Logged

Mike(y)/W3SLK
Invisible airwaves crackle with life, bright antenna bristle with the energy. Emotional feedback, on timeless wavelength, bearing a gift beyond lights, almost free.... Spirit of Radio/Rush
Dave K6XYZ
Member

Offline Offline

Posts: 103



« Reply #14 on: March 29, 2023, 11:43:24 AM »

Reminds me of a book I saw one time called 'To Serve Man'.
Turns out it was a cookbook!
Logged
W3SLK
Member

Offline Offline

Posts: 2651

Just another member member.


« Reply #15 on: March 29, 2023, 11:58:14 AM »

The latest from the AI craze: https://www.foxnews.com/politics/elon-musk-apple-co-founder-tech-experts-call-pause-giant-ai-experiments

K6XYZ said:
Quote
Reminds me of a book I saw one time called 'To Serve Man'.
Turns out it was a cookbook!
That was from the 'Twilight Zone' if I remember correctly!
I'm more fearful of 'Colossus: The Forbin Project'
They say art imitates life but it seems that life is imitating art!
Logged

Mike(y)/W3SLK
Invisible airwaves crackle with life, bright antenna bristle with the energy. Emotional feedback, on timeless wavelength, bearing a gift beyond lights, almost free.... Spirit of Radio/Rush
WD8BIL
Contributing
Member
*
Offline Offline

Posts: 4409


« Reply #16 on: March 29, 2023, 02:37:14 PM »

In the Star Trek episode " I Mudd" the androids were designed and programmed to serve. That didn't work out too well either!  Grin


* I mudd.png (717.38 KB, 745x466 - viewed 127 times.)
Logged
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« Reply #17 on: April 09, 2023, 01:33:33 AM »

THIS UNIT IS THE ULTIMATE ACHIEVEMENT OF COMPUTER EVOLUTION.
Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
N2CQR
Member

Offline Offline

Posts: 29


« Reply #18 on: April 09, 2023, 11:52:07 AM »

And just look at how GPT AI was used (abused?) in a recent SolderSmoke podcast:

https://soldersmoke.blogspot.com/2023/04/soldersmoke-podcast-245-cuban-dsb-dc.html


Logged
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« Reply #19 on: April 14, 2023, 09:19:10 PM »

Is there a time(s)  when the chatgpt was played with? The audio is > 1 hour.
Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
N2CQR
Member

Offline Offline

Posts: 29


« Reply #20 on: April 15, 2023, 10:02:49 AM »

I'd tell you, but we have a contest on this on-going.  See if you can ID the AI-produced segment.  I can give you a hint:  it is less than 3 minutes.  73  Bill
Logged
Pages: [1]   Go Up
  Print  
 
Jump to:  

AMfone - Dedicated to Amplitude Modulation on the Amateur Radio Bands
 AMfone © 2001-2015
Powered by SMF 1.1.21 | SMF © 2015, Simple Machines
Page created in 0.065 seconds with 19 queries.