Home
JAQForum Ver 24.01
Log In or Join  
Active Topics
Local Time 14:41 31 Dec 2025 Privacy Policy
Jump to

Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me.

Forum Index : Microcontroller and PC projects : AI Research Project Ongoing

     Page 3 of 3    
Author Message
PhenixRising
Guru

Joined: 07/11/2023
Location: United Kingdom
Posts: 1686
Posted: 01:29am 29 Nov 2025
Copy link to clipboard 
Print this post

@Revlac

The other thread was locked out after you suggested I "have a coffee and calm down"

In that message, you referenced the "ping pong ball through a bat" Myth buster video as if it was proof of something.

Do you even have a clue about this?
Did you see a spherical object pass through the bat as a sphere?

Did you pay attention to the proximity of the muzzle of the air cannon to the bat?

This is the downright ignorance that prevails today.

All that needs to happen is a misleading video to "prove" something.

Hint: there is a German product "Alarmo" that is totally devastating and all it shoots is 5000PSI of air...no projectile.

You DID NOT see a spherical object pass through that bat...watch again and let's start talking real physics here(?)
 
lizby
Guru

Joined: 17/05/2016
Location: United States
Posts: 3522
Posted: 01:32am 29 Nov 2025
Copy link to clipboard 
Print this post

  PhenixRising said  My latest AI debate is about Concorde: where did it store 7,000 gallons and provide a range of 4,000 miles at mach-2


How can one explain something that isn't true?

According to my sources, "Concorde’s total fuel capacity was about 26,300 Imperial gallons".

Perhaps you were thinking of the WW II B-29 (approx. 7,859 Imperial gallons).

So I kinda doubt you actually did ask AI about it.

~
Edited 2025-11-29 11:48 by lizby
PicoMite, Armmite F4, SensorKits, MMBasic Hardware, Games, etc. on fruitoftheshed
 
PhenixRising
Guru

Joined: 07/11/2023
Location: United Kingdom
Posts: 1686
Posted: 01:44am 29 Nov 2025
Copy link to clipboard 
Print this post

You must be thinking liters

Try applying some logic instead of accepting horse sh*te.

Look at the Concorde....26,000 gallons???

But you know what, they might as well state 26M gallons and you'd still accept it....because they told you.
 
lizby
Guru

Joined: 17/05/2016
Location: United States
Posts: 3522
Posted: 02:01am 29 Nov 2025
Copy link to clipboard 
Print this post

Are you denying that "typical London–New York Concorde flights took about 3 to 3.5 hours gate‑to‑gate" and that "likely on the order of 1.5–2 million people—flew on some combination of London–New York or New York–London sectors"?
PicoMite, Armmite F4, SensorKits, MMBasic Hardware, Games, etc. on fruitoftheshed
 
PhenixRising
Guru

Joined: 07/11/2023
Location: United Kingdom
Posts: 1686
Posted: 02:54am 29 Nov 2025
Copy link to clipboard 
Print this post

And herein lies the problem:

The critical thinker: Well I have analyzed the structure and calculated the required volume of the tanks and to be honest, I really can't see how this could even work. I mean, 0.57 miles/gallon with those Olympus turbo fans just doesn't add up.

Non critical thinker who is happy to be told what to believe: Are you saying that the plane didn't fly from London to NY.


Of course the plane flew. My friend used it several times, only for bragging rights.

But how was it really propelled at mach-2?


 
Mixtel90

Guru

Joined: 05/10/2019
Location: United Kingdom
Posts: 8401
Posted: 08:30am 29 Nov 2025
Copy link to clipboard 
Print this post

WHOA!....
I know I drift off track, but I do try not to for too long. :)
Lizby does have a valid point. This thread was started to continue the AI discussion after the previous one drifted too far out. It would be rather nice to keep this one running now. :)

Would you like to start some sort of "critical thinking" thread, PhenixRising? It could be interesting.
Mick

Zilog Inside! nascom.info for Nascom & Gemini
Preliminary MMBasic docs & my PCB designs
 
lizby
Guru

Joined: 17/05/2016
Location: United States
Posts: 3522
Posted: 11:28pm 02 Dec 2025
Copy link to clipboard 
Print this post

Failures as well as successes should be reported.

I asked Gemini if it could design circuits. It said yes. I prompted with something quite simple--a simulation PCB for testing control of my 3 boiler zones with a Tasmotized Sonoff 4CH Pro R3, with the PCB having only connections to that, a 2-pin 5mm terminal block for 9VAC input, and 4 10MM LEDs with appropriate resistors--all in EasyEDA format.

It produced a file, but was unable to tell me how to import it successfully. Everything I tried got error messages. It said it would do better with a Kicad file, but that, while imported with no errors, showed no circuit.

There was quite a bit of it telling me how I could lay it out, but I repeatedly said that was not the point: if I just wanted the circuit, I would do it with perfboard--what I wanted was for Gemini to do it.

After faffing about for over an hour, I gave up. Maybe Gemini 4 will be able to do it.

On the other hand, it has provided some very valuable and time-saving code about hardening my Seagate Dockstars running openWrt, and providing more logging and reporting. That interaction has been very satisfactory, and I've learned a lot.

But I also had a HAL moment. It lost context, and got stuck on a previous prompt, repeating itself over and over (regarding the circuit design). I had to close the browser page and start a new one.

~
Edited 2025-12-03 12:19 by lizby
PicoMite, Armmite F4, SensorKits, MMBasic Hardware, Games, etc. on fruitoftheshed
 
Mixtel90

Guru

Joined: 05/10/2019
Location: United Kingdom
Posts: 8401
Posted: 07:53am 03 Dec 2025
Copy link to clipboard 
Print this post

"I'm sorry, Lizby, I can't do that."

It probably can design circuits in the form of schematics, possibly as a netlist, but it may not be able to convert the netlist into a pcb. Netlists aren't easy for humans to understand either, and I don't think there is a defined standard for them. You'd have a job finding a netlist-pcb converter to suit a weird netlist.
Mick

Zilog Inside! nascom.info for Nascom & Gemini
Preliminary MMBasic docs & my PCB designs
 
lizby
Guru

Joined: 17/05/2016
Location: United States
Posts: 3522
Posted: 02:52pm 03 Dec 2025
Copy link to clipboard 
Print this post

  Mixtel90 said  "I'm sorry, Lizby, I can't do that."


It produced schematics in a number of formats, but EasyEDA couldn't import any of them. This is just another of those things which it can't do now, but probably soon will be able to do.
PicoMite, Armmite F4, SensorKits, MMBasic Hardware, Games, etc. on fruitoftheshed
 
lizby
Guru

Joined: 17/05/2016
Location: United States
Posts: 3522
Posted: 05:56pm 07 Dec 2025
Copy link to clipboard 
Print this post

On the advice of Gemini, I purchased a Sonoff 4CHPRO R3 4-relay module to provide additional control of the 3 zones on my boiler. I won't be present to do the installation until May, but I flashed it with Tasmota and wired up 4 LEDs on a perfboard to emulate it. I gave this prompt to Gemini:
  Quote  Write a bash script. For the Sonoff 4CHPOW R3, using for Power1, Power2, Power3, and Power4 with, e.g., curl "http://192.168.1.185/cm?cmnd=Power1%20On". Turn on each relay in sequence at 1 second intervals. Pause 3 seconds. Turn them off at 1 second intervals. Turn them all on, pause 5 seconds, then turn them all off. Then continuously at random turn one on or off every 2 seconds.


It took one adjustment, but this is the code it produced:
#!/bin/sh

# --- Configuration ---
IP="192.168.1.185"
# ---------------------

echo "Starting Sonoff 4CH Sequence (Ash Compatible)..."

# Helper function to send command
set_relay() {
   # $1 = Relay Number, $2 = State (On/Off)
   curl -s "http://${IP}/cm?cmnd=Power${1}%20${2}" > /dev/null
   echo " -> Relay $1 turned $2"
}

# 1. Turn on each relay in sequence (1 second interval)
echo "--- Sequence: Turning ON one by one ---"
for i in 1 2 3 4; do
   set_relay $i "On"
   sleep 1
done

# 2. Pause 3 seconds
echo "Waiting 3 seconds..."
sleep 3

# 3. Turn off each relay in sequence (1 second interval)
echo "--- Sequence: Turning OFF one by one ---"
for i in 1 2 3 4; do
   set_relay $i "Off"
   sleep 1
done

# 4. Turn them ALL ON
echo "--- ALL ON ---"
# We use 'Power0' which is Tasmota's shortcut for "All Relays"
curl -s "http://${IP}/cm?cmnd=Power0%20On" > /dev/null
echo " -> All Relays set to ON"

# 5. Pause 5 seconds
echo "Waiting 5 seconds..."
sleep 5

# 6. Turn them ALL OFF
echo "--- ALL OFF ---"
curl -s "http://${IP}/cm?cmnd=Power0%20Off" > /dev/null
echo " -> All Relays set to OFF"

# 7. Random Loop (Continuous)
echo "--- Entering Random Mode (Ctrl+C to stop) ---"

while true; do
   # ASH Random Logic (No arrays allowed)
   
   # Get a random number from system entropy if $RANDOM is missing,
   # but most OpenWrt builds include $RANDOM.
   # If this fails, we can read from /dev/urandom.
   
   # Pick Relay 1-4
   # Take a random number, modulo 4, add 1.
   RELAY=$(( (RANDOM % 4) + 1 ))
   
   # Pick State (0 or 1)
   RAND_STATE=$(( RANDOM % 2 ))
   
   if [ "$RAND_STATE" -eq 1 ]; then
       STATE="On"
   else
       STATE="Off"
   fi
   
   set_relay $RELAY $STATE
   
   sleep 2
done


Youtube Video: Sonoff 4CHPRO R3 exercised

~
Edited 2025-12-08 04:09 by lizby
PicoMite, Armmite F4, SensorKits, MMBasic Hardware, Games, etc. on fruitoftheshed
 
lizby
Guru

Joined: 17/05/2016
Location: United States
Posts: 3522
Posted: 11:31pm 07 Dec 2025
Copy link to clipboard 
Print this post

I took another step; to Gemini:
  Quote  In simple and straightforward html, produce code for a page entitled, "Boiler Zone Thermostat Bypass", with 3 lines containing the line titles "Kitchen", "Livingroom", "Upstairs", an on/off button, a circle to be red or green depending on off or on, an editable text box for the number of minutes to be on, and a non-editable text box for the number of minutes remaining. Provide an automatic update once a minute. Runtime will be limited to 60 minutes. In a subsequent prompt I'll ask you for the backend code and setup procedure for running this on a Seagate Dockstar flashed with openWrt.


It did this and I brought it up in the browser from the local drive.

I asked it how to set it up on my Seagate Dockstar (16-year-old gear) running openWrt, and it told me and provided the code: html in /www, script in /www/cgi-bin, monitoring script in home directory, crontab entry. All worked very straightforwardly.



In years past I have been ready to tear my hair out over this stuff. Now I'll be able next May to upgrade my control of the 3 zones of my boiler, in parallel with the thermostats so they still work.

The Sonoff 4CHPRO R3 cost me $25 on Amazon ("Black Friday" price). The boiler manufacturer wants $800 for gear of similar capacity ($800 Canadian--about $600 USD).
PicoMite, Armmite F4, SensorKits, MMBasic Hardware, Games, etc. on fruitoftheshed
 
MarkF
Regular Member

Joined: 01/08/2023
Location: Australia
Posts: 48
Posted: 03:15am 08 Dec 2025
Copy link to clipboard 
Print this post

I've been using various AI tools/LLMs since November 2022. Regarding AI use, a few things to remember:

1) The AI you use today is the poorest-performing it will ever be! Tools/LLMs like ChatGPT, Grok, Claude, and Gemini continue to improve daily. I've seen significant improvement in these models this year alone (2025). ChatGPT couldn't even access the internet back in March 2025.

2) Some models are better than others at certain tasks. If you don't get the results you expected, copy-and-paste the same prompt into a different AI model. For example, I've found Gemini to be the best for image generation.

3) Feed in helpful data. If you have a Word/PDF of your notes, thoughts, electrical data, diagrams, etc., include that as an attachment to your prompt. The AI's reply will be much better.

I personally like to take the AI's response and add selected points into my research notes document. Then I'll run the prompt again, sometimes in a different AI model, just to gain other insights.

4) Once recursive self-improvement begins (expected in 2026), AI will be better than anyone on the planet in any field. It's going to be scary times!
 
lizby
Guru

Joined: 17/05/2016
Location: United States
Posts: 3522
Posted: 03:50pm 10 Dec 2025
Copy link to clipboard 
Print this post

China Talk

  Quote  Dmitri Alperovitch: The interesting thing about AI is not that it’s going to make tasks easier and faster, but that it’s going to make other things that you would never, ever do before accessible.

PicoMite, Armmite F4, SensorKits, MMBasic Hardware, Games, etc. on fruitoftheshed
 
lizby
Guru

Joined: 17/05/2016
Location: United States
Posts: 3522
Posted: 03:03pm 29 Dec 2025
Copy link to clipboard 
Print this post

Prompt for ChatGPT-5:
  Quote  Consider Saul Griffith's assertion: "In every possible future of this planet, the cheapest electricity will always be your rooftop solar." ( Rooftop Solar  )

What are the chances that some other technology (like fusion or deep geothermal) will make solar a transition technology?


tl;dr:
  Quote  Saul Griffith’s assertion is likely correct in all plausible futures short of post-scarcity civilization: while other technologies may dominate bulk, firm, or industrial power, rooftop solar’s combination of free input, local delivery, modularity, and manufacturing learning curves makes it extraordinarily unlikely to be displaced as the cheapest marginal electricity for end users.


Long form:
  Quote  This is a profound question, because Saul Griffith’s statement is not a casual slogan—it’s a claim about **physics, economics, and political economy** taken together. The right way to answer it is not “is fusion possible?” but **what would have to be true for rooftop solar to stop being the cheapest electricity at the margin**.

I’ll proceed in layers:

1. What Griffith *really* means by “cheapest electricity”
2. Why rooftop solar has structural advantages that are very hard to beat
3. What it would take for another technology to displace it
4. A sober assessment of fusion, deep geothermal, and others
5. The most realistic futures: replacement vs. coexistence
6. A bottom-line probability judgment

---

## 1. What Saul Griffith actually means by “cheapest”

Griffith is **not** saying rooftop solar is the cheapest *bulk* electricity.

He is saying it is the cheapest **marginal, delivered, retail electricity** for an end user.

That’s a crucial distinction.

When you put solar on your roof, you avoid:

* Transmission
* Distribution
* Grid losses
* Retail markups
* Capacity charges
* Some regulatory overhead

So the comparison is not:

> “Can fusion generate electricity cheaply at a power plant?”

It is:

> “Can anything beat electricity generated exactly where it is used, with no transport?”

That sets a very high bar.

---

## 2. Why rooftop solar has structural, not temporary, advantages

### A. It exploits a universal free input

Sunlight:

* Arrives everywhere
* At predictable intensity
* Without fuel cost
* Without geopolitical risk

Any competing technology must:

* Either beat **free**
* Or deliver something solar cannot (e.g., 24/7 firm power)

---

### B. It rides exponential manufacturing, not infrastructure buildout

Solar improves via:

* Learning curves
* Manufacturing scale
* Materials science
* Automation

This is **Silicon Valley–style progress**, not **utility-style progress**.

Fusion, geothermal, fission, CCS:

* Improve through projects
* Regulation
* Capital deployment
* Construction learning

Those scale much more slowly.

---

### C. It benefits from avoided-cost accounting

Rooftop solar is often installed when:

* A roof is being replaced
* A home is being built
* Electrical upgrades are happening anyway

That means:

* Incremental cost is lower than headline cost
* Payback is often measured against *retail* prices, not wholesale

This advantage never goes away.

---

## 3. What would it take to make solar “just a transition technology”?

For rooftop solar to lose its position, **one of three extraordinary things must happen**:

### Condition 1: Electricity must become so cheap everywhere that location no longer matters

If grid electricity fell to, say:

* $0.01/kWh retail
 Then:
* Rooftop solar wouldn’t be worth installing

But this would require:

* Generation near-zero cost
* Transmission near-zero cost
* Distribution near-zero cost
* Financing near-zero cost

That implies a world where:

* Capital is almost free
* Infrastructure is trivial
* Regulation is frictionless

That world would already be post-scarcity.

---

### Condition 2: A technology must deliver firm, local, ultra-cheap power at household scale

To beat rooftop solar, a competitor must be:

* As local
* As modular
* As cheap
* As safe
* As maintenance-free

That’s an *extraordinarily* narrow target.

---

### Condition 3: Solar must hit a hard physical or economic wall

Solar would need to:

* Stop improving
* Become materially scarce
* Face insurmountable land/area limits
* Be eclipsed by something strictly superior at small scale

There is no evidence of this yet.

---

## 4. Assessing the main contenders

### A. Fusion

Fusion’s strengths:

* Dense energy
* Firm power
* Potentially cheap fuel

But for rooftop displacement, fusion fails on multiple axes:

* Centralized, not distributed
* Gigantic capital intensity
* Long development cycles
* Requires grid delivery
* Political/regulatory complexity

Even **perfect fusion** would likely produce:

* Very cheap *wholesale* electricity
* Still not cheaper *delivered* than rooftop solar in many cases

Fusion competes with:

* Gas
* Nuclear
* Coal
 Not rooftops.

**Probability fusion displaces rooftop solar:** *Very low*

---

### B. Deep geothermal (EGS, superhot rock)

This is more interesting.

Strengths:

* Firm, 24/7
* Local in some regions
* High capacity factor

Limits:

* Highly site-specific
* Requires drilling
* High upfront capital
* Still grid-connected

Deep geothermal could:

* Displace fossil baseload
* Reduce need for storage
* Lower wholesale prices

But it does not:

* Appear on every roof
* Avoid distribution costs
* Scale modularly to households

**Probability it displaces rooftop solar globally:** *Low*
**Probability it complements and reduces storage needs:** *High*

---

### C. Advanced fission (SMRs, microreactors)

Similar story:

* Capital-heavy
* Centralized
* Regulatory overhead
* Long timelines

Even microreactors are:

* Orders of magnitude more complex than PV
* Not consumer-installed technologies

They compete with:

* Grid generation
 Not rooftop solar.

---

### D. Space-based solar

Ironically, this is still **solar**—just with extra steps and losses.

If it worked:

* It would still deliver via the grid
* Still incur distribution costs

It does not beat rooftop solar at the margin.

---

### E. Something genuinely unknown?

This is the strongest counterargument:

> “What if a technology we don’t yet imagine appears?”

To displace rooftop solar, it would need to:

* Be modular
* Be safe
* Be cheap
* Be local
* Beat sunlight + silicon + mass manufacturing

At that point, it would likely look like:

* Some form of *even cheaper solar*
* Or *direct ambient energy harvesting*

In other words, it wouldn’t negate Griffith’s claim so much as **generalize it**.

---

## 5. The most realistic future: solar as the floor, not the ceiling

The most likely outcome is:

* Rooftop solar sets the **price floor** for electricity
* Other technologies compete to:

 * Provide firm power
 * Provide seasonal balance
 * Serve dense urban/industrial loads
 * Reduce system-wide costs

In that world:

* Solar is not “transitional”
* It is **permanent infrastructure**
* Other tech stacks on top of it

Much like:

* TCP/IP didn’t get replaced by faster networks
* It became the substrate

---

## 6. A probabilistic judgment

If we’re forced to put rough odds on it:

* **Rooftop solar remains cheapest marginal electricity for end users through 2050:** ~80–90%
* **Another tech beats it everywhere, including rooftops:** <10%
* **Solar becomes secondary but still widely used:** ~10–20%
* **Solar becomes obsolete:** ~1–2% (catastrophic or post-scarcity futures)

Griffith’s claim is not that solar is *the best* electricity—
it’s that **nothing else can so easily undercut “free photons + local conversion.”**

---

## 7. One-sentence synthesis

Saul Griffith’s assertion is likely correct in all plausible futures short of post-scarcity civilization: while other technologies may dominate bulk, firm, or industrial power, rooftop solar’s combination of free input, local delivery, modularity, and manufacturing learning curves makes it extraordinarily unlikely to be displaced as the cheapest marginal electricity for end users.

PicoMite, Armmite F4, SensorKits, MMBasic Hardware, Games, etc. on fruitoftheshed
 
     Page 3 of 3    
Print this page


To reply to this topic, you need to log in.

The Back Shed's forum code is written, and hosted, in Australia.
© JAQ Software 2025