*NERD ALERT* Using AI for data set trend analysis

[ChairmanFester]

"Done growing" out of context could mean the answer is "Yes, the plant appears to be done growing!" because you killed it and dead things don't come back without the right books.

 

[shooter64738]

"Done growing" out of context could mean the answer is "Yes, the plant appears to be done growing!" because you killed it and dead things don't come back without the right books.

Exactly. The user has to frame the context. Ive spent the last week trying to convince a grower that if they are going to use gemini or any ai, they have to give it context. Historical data bolsters that.

Most of these chat services dont have the bandwidth or capacity to store data for every person using it. And some use it anonymously so they have to keep retelling it everything from the start. The way @Brother_Antiocitis using it is different. Its getting context from the csv history so those results will be more reliable.

 

[trygrowup]

Exactly. The user has to frame the context. Ive spent the last week trying to convince a grower that if they are going to use gemini or any ai, they have to give it context. Historical data bolsters that.

Most of these chat services dont have the bandwidth or capacity to store data for every person using it. And some use it anonymously so they have to keep retelling it everything from the start. The way @Brother_Antiocitis using it is different. Its getting context from the csv history so those results will be more reliable.

You're showing the strawberry only,
Just saying.

 

[ChairmanFester]

For me, it seemed to concentrate on "feel" and knowing the plant without using numerical schedules or feeding just, checking signs the in-tune gardener way. Although I liked it in ways I just used it less, never quite needed it, and simply got some use out of it.
Otherwise it really hates an awful lot of my ideas, you should see how many caution paragraphs I got when designing a system to feed plants using explosives. And it won't even discuss doing anything if I use a word like "uranium".

 

[Brother_Antioch]

For me, it seemed to concentrate on "feel" and knowing the plant without using numerical schedules or feeding just, checking signs the in-tune gardener way. Although I liked it in ways I just used it less, never quite needed it, and simply got some use out of it.
Otherwise it really hates an awful lot of my ideas, you should see how many caution paragraphs I got when designing a system to feed plants using explosives. And it won't even discuss doing anything if I use a word like "uranium".

Pretty sure I had some of that nuclear weed back in the day...at least, it felt that way!

Cheers,

BA.

 

[Brother_Antioch]

Exactly. The user has to frame the context. Ive spent the last week trying to convince a grower that if they are going to use gemini or any ai, they have to give it context. Historical data bolsters that.

Most of these chat services dont have the bandwidth or capacity to store data for every person using it. And some use it anonymously so they have to keep retelling it everything from the start. The way @Brother_Antiocitis using it is different. Its getting context from the csv history so those results will be more reliable.

That was my thought too, shooter. It didn't take more than a couple of days to see that I'd have to upload the .csv and new pics every time if I wanted consistent analysis. Relying on it to go over previous hard data was just a waste of time. But, as I said, I just add a line to the spreadsheet and resave it under the same file, then export the new .csv to ChatGPT and it looks at it with a fresh set of eyes, so to speak.

Ultimately, it's a tool. Some people use a screwdriver to set screws, some use it to hammer in a nail. Can it do both? sure. Can it hammer in a nail? Not very well, no. But it's not designed to. It all comes down to understanding the limitations of any tool in order to get the most out of it.

Cheers,

BA.

 

[ChairmanFester]

With ore and a kit you can make a very precise clock, not "the atomic clock" but better than anything in an average home. I collect historic nuclear stuff though and it's about that. Some is regulated so you have to do a form and a fee.

 

[shooter64738]

You're showing the strawberry only,
Just saying.

I guess im not following. If you filter out everything except the strawberry, there is only strawberry. I dont think ai is trying gain anything from interacting with humans. Its rather stupid in most regards. It is certainly not intelligent, it just crunches data way faster than humans do it. It finds patterns and knows things that its very time consuming for a person to find and research. But it only knows those patterns because it either has a reference for it, or its been trained to do that specific thing

This to me is only a strawberry
{
"plant_id": "white_widow",
"timestamp": "2026-02-01T14:34:00Z",

"health_summary": {
"overall_status": "healthy",
"confidence": 0.92,
"notes": "no active stress signatures; white widow's typical calcium-transport instability not detected in this capture"
},

"ai_interpretation": {
"short_summary": "spectral profile matches a stable white widow phenotype with balanced pigment and water signatures.",
"detailed_summary": "visible-band absorption is consistent with strong chlorophyll density typical of white widow. red-edge slope and ndvi fall within expected ranges for a non-stressed hybrid. no spectral anomalies indicating calcium-related spotting, water stress, or nutrient imbalance."
},

"camera": {
"camera_height_m": 1.22,
"camera_angle_degrees": 45.0,
"distance_to_plant_m": 0.85,
"focal_length_mm": 3.6,
"sensor_width_mm": 4.8,
"sensor_height_mm": 3.6,
"resolution_px": { "width": 1920, "height": 1080 },
"exposure_ms": 8.3,
"iso": 120,
"white_balance_mode": "auto",
"illumination": {
"spectrum": {
"405nm_mw": 18.2,
"450nm_mw": 22.5,
"550nm_mw": 19.1,
"660nm_mw": 24.7,
"720nm_mw": 17.4,
"850nm_mw": 15.2,
"940nm_mw": 14.8
},
"intensity_ppfd": 612.0
}
},

"thermal_image": {
"sensor_type": "",
"resolution_px": { "width": 64, "height": 32},
"emissivity": 0.98,
"min_temp_c": 22.9,
"max_temp_c": 25.1,
"mean_temp_c": 24.1,
"thermal_gradient_map": "normalized_0_to_1",
"notes": "no hot spots or stomatal-shutdown signatures detected"
},

"spectral_bands": {
"405nm": { "mean": 0.12, "std": 0.01, "min": 0.10, "max": 0.15 },
"450nm": { "mean": 0.09, "std": 0.01, "min": 0.07, "max": 0.11 },
"550nm": { "mean": 0.18, "std": 0.02, "min": 0.15, "max": 0.22 },
"660nm": { "mean": 0.07, "std": 0.01, "min": 0.05, "max": 0.09 },
"720nm": { "mean": 0.32, "std": 0.03, "min": 0.28, "max": 0.38 },
"850nm": { "mean": 0.48, "std": 0.04, "min": 0.42, "max": 0.55 },
"940nm": { "mean": 0.41, "std": 0.03, "min": 0.36, "max": 0.47 }
},

"indices": {
"ndvi": 0.74,
"ndre": 0.42,
"gndvi": 0.56,
"vari": 0.21,
"nir_red_ratio": 6.85,
"red_edge_slope": 0.0031
},

"mask_diagnostics": {
"valid_pixel_count": 18234,
"mask_coverage_percent": 87.5,
"mask_quality": "good"
},

"environment": {
"leaf_temperature_c": 24.1,
"ambient_temperature_c": 23.4,
"relative_humidity_percent": 58.0,
"light_intensity": 612.0
},

"botany": {
"species": "cannabis sativa × indica hybrid",
"cultivar": "white widow",
"genotype": "hybrid, sativa-leaning",
"growth_habit": "upright",
"phenological_stage": "vegetative",
"morphology_stage": "node-4",
"days_since_germination": 21,
"leaf_count": 8,
"canopy_height_mm": 142
}
}

 

[trygrowup]

Ok

 

[amneziaHaze]

I tried using chatgpt but after a month of data one day it just started using random numbers that where not even in the list.then when you ask it why it just says oh my mistake.then you remember ai doesnt understand math and your trusting your grow to a random number generator