Sam Barrett, PhD
banner
ai4geo.bsky.social
Sam Barrett, PhD
@ai4geo.bsky.social
GeoAI, Climate, Remote Sensing, Generative AI and more!
I guess though it makes compute cheap which itself enables more compute. If compute is expensive, the overall impact might be lower because of the cost. Inverse Jevons?
November 13, 2025 at 6:17 PM
College is for parties. Anything that increases the party to learning ratio is good.
November 11, 2025 at 8:48 PM
There's a possible future of personal bots and continuous learning where this could happen. Though would still be bizarre and unnecessary!
November 11, 2025 at 8:26 PM
This sounds like a fascinating genre.
November 11, 2025 at 8:23 PM
Been nodding along to smart people espousing KGs for a while not knowing what to really think but now you provoked me to consider, I think you're right.
November 11, 2025 at 3:25 PM
Because of that, I'm at least open to the idea there may be very non intuitive ways to address e.g. climate change in a world where we actually massively increase overall energy production and use.
November 11, 2025 at 8:03 AM
Are you just talking about increasing electricity as proportion of total energy production/use (replacing existing non electrified systems), or growing total energy production?
November 11, 2025 at 7:50 AM
Which is why solar replacing farm land is FAR better than growing corn for biofuels.
November 11, 2025 at 7:48 AM
This feels bitter lesson pilled
November 11, 2025 at 7:46 AM
Exactly, and also the "it isn't human shaped" (would love to know what a neuroscientist calls that...). We somehow *still* seem surprised that intelligence can be expressed in ways fundamentally jagged compared to us: a crazy mix of super human through to incompetent depending on the axis.
November 9, 2025 at 10:57 PM
Though arguably that's just generally the case for all existing human work and means of expression!
November 9, 2025 at 6:06 PM
Can you elaborate on this meta-cognitive design work? I'm definitely coming around to the idea there will be a very wide range of kinds of human-AI collaboration, and that some of them will be much more natural and comfortable for particular types of people.
November 9, 2025 at 6:06 PM
But but but... hamburgers are food! How dare you compare using the evil parrot machine to something fundamentally necessary for human survival...
November 9, 2025 at 5:56 PM
I really feel all this would be easier to talk about if AI DIDN'T range all the way from IF statements... if it wasn't an umbrella term but was more intuitive to what's going on right now.
November 9, 2025 at 5:49 PM
Indeed, but I do feel there's a missing term. Maybe it's just generative AI, but AI which is sufficiently advanced that it provokes discussions about and comparisons to (whether one agrees or not) human intelligence. Thing is it's a continuum...
November 9, 2025 at 5:48 PM
Have you considered defining intelligence as what only humans have? Reasoning as what only humans do? Understanding as what only humans can do? Checkmate AI lover!
November 7, 2025 at 8:01 AM
One thing is I doubt it's a huge accelerate. Could be it's the best Earth Observation foundation model, but since other models haven't been big accelerate yet, I'm not sure this improvement will be a watershed.
November 5, 2025 at 12:49 AM
I need to dig in but this is my domain! Skim says impressive. I'm wondering how they've managed to achieve better performance than before. Will report back.
November 5, 2025 at 12:48 AM
There are ways to use LLMs in unhealthy or detrimental ways: no shit Sherlock. Therefore any use is detrimental... I have no words.
November 3, 2025 at 9:06 PM
..
"We found asking other people to do your homework for you leads to less learning, therefore asking people for things leads to less learning".
November 3, 2025 at 9:06 PM
Plus, the is a near infinite variety of ways to use LLMs, much like there are near infinite possible ways to communicate with other people via language. Yet of you made some sweeping statement about general cognitive effects of interacting with humans based on a single narrow example in one study..
November 3, 2025 at 9:06 PM