That's an interesting take and localised use and adaptation may be more realistic than ambitious projects (e.g. large-scale monitoring at high resolution, eDNA, realtime monitoring) that produce very large datasets requiring either massive storage or massive automated computing for analysis.
11.03.2026 18:14
π 1
π 0
π¬ 1
π 0
PhD horse: starts out as a detailed drawing from the tail and gets progressively worse so that the head and forleg are stick or childlike drawing.
It's the phd horse, but for AI
11.03.2026 18:05
π 1
π 0
π¬ 0
π 0
OTOH with enough farm-specific data, AI could aid in precision agriculture (targeting inputs, farm planning).
OTOH a chatbot for farms is a waste of resources when farms need quality ag. advisors and tailored approaches for sustainability.
A question of approach and resource allocation.
11.03.2026 18:00
π 1
π 0
π¬ 2
π 0
I suck at stats, but monte carlo came to mind when I read your earlier text.
I keep thinking about an EU project some years ago where one WP was supposed to develop a "virtual agricultural advisor" or some such. In the last year of the project it appeared to me to still be completely aspirational.
11.03.2026 18:00
π 1
π 0
π¬ 1
π 0
But, lack of "opinion diversity" is an awfully important shortcoming where the range of views (not the necessarily the frequency) is a main finding of the qualitative research.
Ours are probably fundamentally divergent approaches and philosophies, both with potential applications and shortcomings.
11.03.2026 10:48
π 1
π 0
π¬ 1
π 0
"While these "artificial personas" give very *plausible* responses, a recurring problem is that LLMs underestimate opinion diversity"
I've seen this using farmer responses as an example. Yes, LLMs can produce plausible responses - after all, it's based on on the literature or the data inputted. 1/2
11.03.2026 10:48
π 1
π 0
π¬ 1
π 0
Dementia on display
10.03.2026 04:10
π 1
π 0
π¬ 0
π 0
And since the learning and validation has to take place, the generated literature review needs to be reviewedβ¦
Poor parameters will result in a poor lit review, regardless of whether it is LLM or a person. Science isnβt a linear process. Who and how will the refinement happen- who checks the work?
09.03.2026 10:02
π 1
π 0
π¬ 0
π 0
Could AI reliably carry out quantitative research like sending and analysing online surveys? Maybe in the future. Could it do qualitative research where dynamics and creativity matter? Not any time soon.
09.03.2026 09:56
π 4
π 0
π¬ 0
π 0
Janne, I do farmer interview based research. Sometimes as focus groups.
It is a 2-way street (or intersection w focus groups). There is a whole interactive dynamic that affects the research.
Currently, LLMs donβt even reliably transcribe or translate the interviews.
09.03.2026 09:53
π 5
π 0
π¬ 1
π 0
Iβm increasingly seeing this in various types of project outputs, where outputs consist of AI generated images and vague, smooth text bereft of insight or innovation. Itβs based on prompts and theft of existing materials.
09.03.2026 09:43
π 10
π 2
π¬ 0
π 0
alt text:
09.03.2026 07:16
π 48
π 27
π¬ 1
π 2
Comprehensive. Just missing the malice/ sociopathy, kompromat, child r*pe and trafficking ring, and the suspected money laundering and other shady financials.
09.03.2026 07:18
π 3
π 0
π¬ 0
π 0
A dying empire determined to take everyone with it.
Comgress could stop this madness. Instead⦠this.
08.03.2026 15:20
π 3
π 0
π¬ 0
π 0
Maria Anna Mozart composed masterpieces. Wolfgang Amadeusβ name survived, hers vanished.
Janet Sobel pioneered drip painting. Jackson Pollock was the genius.
This International Womenβs Day, letβs restore their stories and give them the credit they deserve!
08.03.2026 08:36
π 112
π 28
π¬ 0
π 0
Iranβs people were facing a horrible shortage of water before the war began. If we are destroying desalination plants and setting fire to Teheran we are committing unfathomable crimes.
08.03.2026 02:19
π 25311
π 8111
π¬ 781
π 328
The skyline of Tehran has a small shaft of light to the west. But the rest is darkness, shrouded in black cloud from burning oil depots and refineries
Daybreak brings only more night in Tehran. A city blackened by burning oil.
Hospitals, schools, aid centres, freshwater desalination plants and now fuel storage depots.
These are all the sites you would target if your wish was to perpetuate a genocide.
(πΈ Al Jazeera)
08.03.2026 08:06
π 260
π 158
π¬ 5
π 8
NEW: Watch this report just in from CNNβs Frederik Pleitgen. It is raining oil in Tehran this morning.
A city of 10 million is being actively poisoned.
(π₯ CNN)
08.03.2026 07:59
π 873
π 623
π¬ 43
π 91
The burning oil refinery in Tehran tonight.
This ancient city of 10 million people is being annihilated by the US and Israel.
(π₯ BBC Verify)
08.03.2026 01:19
π 1514
π 771
π¬ 100
π 133
Trae is always on point.
06.03.2026 06:20
π 3
π 0
π¬ 0
π 0
Probably no good solutions, but I hope you find some way to address it safely for everyone.
05.03.2026 21:12
π 1
π 0
π¬ 0
π 0
ASPCA?
05.03.2026 21:02
π 1
π 0
π¬ 1
π 0
I suspect this type of evaluation will be easy to game: Student submits course literature to LLM and tells it to write an essay based on the assignment instructions.
The educator uses the same instructions to set parameters and tells the programme to evaluate. And round and round we go π
05.03.2026 16:47
π 1
π 0
π¬ 0
π 0
It *can* identify, to some degree, whether set parameters are met or missing.
But there is no person judging the argument, evaluating the use of the literature, noting students' unique perspectives and development.
05.03.2026 16:45
π 1
π 0
π¬ 1
π 0
I wonder what students think about AI being used to evaluate their work and give them feedback.
I had to try this out on some group work as part of a course in university pedagogy. I felt deeply sad for the *appearance* of evaluation while knowing the algorithm can't actually evaluate our work.
ππ§ͺπ
05.03.2026 16:36
π 0
π 1
π¬ 1
π 0
Daily reminder that a few folks in Congress voting to stop this madness could stop this madness.
03.03.2026 22:31
π 0
π 0
π¬ 0
π 0
And, again, I am saying EV wasnβt right for me in my circumstances and by my criteria. I have literally no opinion on other peopleβs choices.
03.03.2026 22:20
π 0
π 0
π¬ 0
π 0
am happy with running a car on biogas made from waste, which reduces emissions by over 90% compared to fossil fuel. This is a huge savings without contributing to an electric fleet that will be obsolete in 10 years with no good upgrade or repair options.
03.03.2026 22:09
π 0
π 0
π¬ 1
π 0