Its not the beauty of the display example, but the fact that its kites in real wind. To conflate drone feats with kites is misleading.
What’s your best autonomous kite flight example, ugly or not?
Human pilot sensing is not about “exact readings”, but high-dimensional sensor fusion.
Everyone’s flying autonomusly. Makani, Kitepower, Ampyx… take your pick.
[quote=“kitefreak, post:21, topic:180”]
Human pilot sensing is not about “exact readings”, but high-dimensional sensor fusion.
[/quote] That should be an argument for neural nets instead of deterministic algorithms.
I pick them all as inadequate. Which of the examples can respond to a METAR warning, or even just a squall line approaching?
Unfortunately for flight safety, we can’t just train Neural Nets high-dimensionally, because they can’t be validated to human pilot level of reliability. A human expert can explain their reasoning, but current neural nets are often very “opaque”. Formal Flight Software standards are so strict most AWE programmers are not even close, they don’t even document their buggy prototype code well, much less validate it to FAA “clean-room” standard.
Its going to take many years to reach the needed standards in AWE, that’s all. Its not time yet to be satisfied.
The ones with automatic launch and landing probably. Makani, Ampyx. It’s not that hard.
Luke, you are guessing wildly. No, Ampyx and Makani are not ignorantly risking their platforms with novel METAR-driven weather-sensing over-rides. They all depend on Pilot-In-Command for meta-logic control. These really are hard problems, or you could show specific solutions.
“Classic” here is who does not know standard computer programming acronyms-
Is there any larger scale awes that is intended to be manually piloted?
The largest scale AWES concepts are kPower’s Kite Network concepts, up to planetary scales in principle. The control paradigm is simple Supervised Flight Automation of mostly passive-dynamic action, not advanced AI.
Currently, the largest AWES experiments all require a PIC and VOs, for example to “sense and detect” intruding air-traffic. None of the ventures are known to be attempting full automation of PIC function, which would be very premature at present.
A critical path estimate for AI AWE seems to involve at least a decade or two of broad progress in avionics and other fields.
One more great excuse for you to make zero power. After 12 years of B.S.
KiteFreak says: The largest scale AWES concepts are kPower’s Kite Network concepts, up to planetary scales in principle.
*** DougS replies: Please show us a small demo, say 1 kW or so. As just one person, I’m very eager to see one. I’d guess others would also like to see a real-life demo what you;re talking about. Please show us any “Power-Kite” generating any meaningful or useful amount of power, at any scale. Words are not Watts.
Otherwise, how could anyone think your daily diatribes have any meaning whatsoever?
Doug, It is a “great excuse” for working at small scale to develop methods while everything lines up worldwide on many fronts, including flight AI, a cool field that just needs a lot more time and work. If you can solve AWE better and faster, please do.
Change topics to discuss non-AI power-demo issues on-topic. You are mistaken if you think the power-kite is the path of “zero power” compared to the 1000ft driveshaft idea. This topic is about AI and AWE. Any informed contribution or sincere question, on-topic, is appreciated.
Recently, @Windy_Skies showed some examples of how to use ChatGPT https://chat.openai.com/chat. I followed up.
As ChatGPT improves as we use it, I wonder if it would be possible to have even synthesized answers on some questions about the AWES methods to choose and how to use them, or about specific issues, according to the different parameters provided by the questioner (mass, power, space used, safety, choice of materials, regulations, advantages and disadvantages, lifetime, Levelized Cost of Energy (LCOE)…). Reiterating the questions with answers as they are asked would perhaps make it easier to decide on the direction to take.
In this way, an increasingly solid AWE database could perhaps be built up.
There are problems with using only the current sophistication of ChatGPT AI - -It’s basically just a big bullshit engine. Works like - look at all that’s been said up to date - What would likely be said - I’ll say that.
Maybe once they train multiple AI classes to argue and to spot when it’s factually wrong , question itself, have modesty, build proofs, stop gaslighting people …
Singularity? I dont think that necessarily is something we should look forward to. Yes we may finally get AWE, but humanity may be doomed at the same time
Perhaps then design a tool specific to AWE, having assimilated prototype testing, scientific publications, forum discussions…
That said, it would work in a ChatGPT type of interaction: the more relevant things you tell me, the more I come out with, and the more bullshit you tell me, the more I come out with.
A well-designed and well-used tool would function almost like a calculator, capable of synthesizing the multitude of data recorded, which neither an individual nor a team can do after thousands of scientific publications, patents, and in the absence of a single AWES at utility scale in permanent use.
I guess if we made a physics sim playground with enough fidelity and set goals of achieving electrical generation (available at ground) through airborne interactions… with enough evolutions scored it would eventually find some ways.
Too far off for now I reckon
AI does play a mean game of virtual human football already though
What about artificial stupidity (AS)? I was thinking about getting into it. Nobody is pursuing it! A wide open opportunity. Shhh, keep it under your hat! 
Hi Doug, Einstein is thought to have said: “Two things are infinite: the universe and human stupidity; and I’m not sure about the universe!”
I let you imagine the infinite potential of artificial stupidity (AS) based on human stupidity (HS)…
With AI i think there are two things that are conpletely wrong; the first is that AI is basically useless and stupid, and could not think. The other one is blind faith that AI can do almost anything.
AI will be very useful. It can’t do everything a human can do. The truth will probably be somewhere in the middle of these two extremes. I feel these days many people are in either one of these two camps. Better to have a nuanced point of view on these things.
In my opinion I think AI is an important step in human history. Maybe it can be as important as the internet. But I wouldnt trust an AI to do my coding.
ML (simple machine learning) is an AI which can (most likely) do particularly boring tasks in AWES very well.
Such as
If I want to read the tension / vibration detail of individual TRPT lines
Oh man that’d not be a fun human job.
But it could give very valuable data if you could quickly spot changes or recognisable patterns indicating problems like imbalance, a stretched line or worse…
Like Tesla is using AI to monitor each battery cell. That stuff makes a lot of sense to me.
15 years of empty promises from “really smart people”, and now this…
Try calling any corporation, and see if you can get to talk with a human.
“In order to direct you to the right person, Please tell me why you are calling. You can speak in plain English”
No matter what you say back, it will say:
“Uhh, I… didn’t get that” “Please tell me why you are calling. You can speak in plain English”
After 5 or 10 minutes, maybe you throw the phone at the wall.
Pick it up and it will still be saying “Please tell me why you are calling. You can speak in plain English”