Not a day over 35 - but the grey hair is a giveaway!
I have written a specialist book on hydropower. When I asked ChatGPT to find it, it did so along with two non-existent Coles Kuhtze and the news (to me) that I had revised it in 2019.
I can also confirm that it is consensus driven and would have rubbished Galileo and many others like him. Don't expect it to identify breakthrough science.
I keep pondering that the human brain is said to use about 25 watts. It seems that AI has a lot of room for efficiency improvements.
Then again, AI is very fast. How many human brains for how long to sort through all the research that AI can cover in a few minutes or seconds or less?
Great photo. AI will really be something when it can do the same to actual human faces and not just photos. Still waiting for rejuvenation treatments to become real...
Outstanding post, Grandma. Thanks. Great photo, by the way. I too am in my 70s. Wrinkles scare away the kiddies, and leaves me in peace, so I edit my photos towards the dark side...
I share your conclusion about AI being "useful." It's value added to the health industry alone will justify the expense. It should also make energy SCADA systems even more useful. That its application is limited by available electricity would be criminal! If mankind is going to flourish, he will need every tool necessary! Perhaps the growth of this technology will convince utilities heavily invested in renewables that such things are nothing more than a placebo, and that new power should come from reliable sources.
I have spent a fair amount of time inside data centers. This much I do know, it's cooling that uses the most energy. Gone are the days of the IBM mainframes with a direct chilled water loop to the processor rack. We have moved on to VM warehouse now that run on larger versions of the same,servers that we all had in a rack at the office. Lots of them running in parallel. Need more power, add another machine. They have hot and cold aisles to circulate cooling air through the racks. So as the processor load goes up, so does the heat output. AI takes lots of processor power, but it's getting better.
Thank you for the insight. Most of my references do not address the cooling requirements, and my post was covering a lot of ground. So I didn't dive into that issue.
I wonder if most people are aware of the waste heat. If my phone is low on charge, I plug it in. The phone becomes warm. I can tell if it is finished charging by whether it is still warm to the touch. Primitive but effective.
Within living memory for some of us, bits were expensive. A lot of work went into the conservation of bits, some residues of which are still with us. The advance of technology reduced the cost of bits, but not uniformly. Messaging charges were a way to price bits over the air, which cost more than bits over a wire. Then it wasn't worth charging for messages. Now we don't even worry (very much) about sending full-resolution pictures wirelessly: it's not worth the trouble to compact them.
We have gotten used to the idea that a compute bit per second is also essentially costless. If our demand for bps goes up by a few orders of magnitude, maybe that will no longer be the case, at least for some amount of time.
A long-winded way to suggest that we may have to invent a way to charge for searches and figure out ways to match willingness to pay with the necessary compute power to derive a response, ranging from trivial to insightful.
Yes, Moore's law seems to apply to everything electronic, not just chips. Will this be forever? I also found Francis Turner's link (in his comment above) very interesting. "Actively Incinerating Cash."
Been using Grok. So superior to google that I really don’t care if it uses more electricity. I’m getting tax advice for next year this morning.
One problem, I’m going to want to use it at night when the wind isn’t blowing. Glad we have an endless supply of natural gas, and maybe nuclear someday.
The energy consumption of today’s AI systems can be drastically reduced — while significantly improving functionality — if we integrate them with genuine human abilities like true intelligence, intuition, and emotion.
This is where IntraCavi comes in: a non-invasive interface that interacts directly with the human brain via five cranial nerves.
The broader HAIXUS approach proposes a shift in logic:
We don’t need to reinvent intelligence from scratch — we can amplify the one we already have.
By connecting the human brain to multiple AI systems, we unlock a new form of cognition — distributed, enhanced, and resource-efficient.
💡 No chips. No implants. Just humans — in symbiosis with intelligent systems.
The issue with ai is that’s it’s artificial, as in not real - fake. It’s also just lazy way to compile multiple correlations and regressions. One huge problem is the data force. And you believe everything from the internet? Then why even use ai? If you don’t understand the question you ask you need to do the work to learn about the subject, not just throw in a bunch of garbage to get some poorly validated input.
I'd say AI saves a lot of time and effort at the lower end of the AI levels as defined by Powell, thus making the higher levels of cognition for human reasoning. That of course must include checking for AI errors.
So you found a use for AI, Angwin?
As a 71-year-old outdoorsman, I'm grateful that wrinkles don't hurt…
Not a day over 35 - but the grey hair is a giveaway!
I have written a specialist book on hydropower. When I asked ChatGPT to find it, it did so along with two non-existent Coles Kuhtze and the news (to me) that I had revised it in 2019.
I can also confirm that it is consensus driven and would have rubbished Galileo and many others like him. Don't expect it to identify breakthrough science.
Thank you. I agree.
AI is certainly consuming gigawatts of power and billions of dollars of cash.
The question is whether it produces anything people are willing to pay for - so far, AFAICT, it isn't
https://ombreolivier.substack.com/p/ai-actively-incinerating-cash?r=7yrqz
I keep pondering that the human brain is said to use about 25 watts. It seems that AI has a lot of room for efficiency improvements.
Then again, AI is very fast. How many human brains for how long to sort through all the research that AI can cover in a few minutes or seconds or less?
Great photo. AI will really be something when it can do the same to actual human faces and not just photos. Still waiting for rejuvenation treatments to become real...
Outstanding post, Grandma. Thanks. Great photo, by the way. I too am in my 70s. Wrinkles scare away the kiddies, and leaves me in peace, so I edit my photos towards the dark side...
I share your conclusion about AI being "useful." It's value added to the health industry alone will justify the expense. It should also make energy SCADA systems even more useful. That its application is limited by available electricity would be criminal! If mankind is going to flourish, he will need every tool necessary! Perhaps the growth of this technology will convince utilities heavily invested in renewables that such things are nothing more than a placebo, and that new power should come from reliable sources.
Thank you again for your great work!
Meredith, that's exactly how I remember you when we met a couple of years ago! This AI might be more clever than you give it credit.
Thank you!
I have spent a fair amount of time inside data centers. This much I do know, it's cooling that uses the most energy. Gone are the days of the IBM mainframes with a direct chilled water loop to the processor rack. We have moved on to VM warehouse now that run on larger versions of the same,servers that we all had in a rack at the office. Lots of them running in parallel. Need more power, add another machine. They have hot and cold aisles to circulate cooling air through the racks. So as the processor load goes up, so does the heat output. AI takes lots of processor power, but it's getting better.
Thank you for the insight. Most of my references do not address the cooling requirements, and my post was covering a lot of ground. So I didn't dive into that issue.
I love your substack https://substack.com/@kilovar1959
Maybe you can address this issue in a post?
I will see what I can toss together.
District heating? I know that in winter my son's gaming PC heats half our house. More of a problem in summer.
I wonder if most people are aware of the waste heat. If my phone is low on charge, I plug it in. The phone becomes warm. I can tell if it is finished charging by whether it is still warm to the touch. Primitive but effective.
Maybe 8n Europe, district heating never really caught on in the US.
I love the AI picture of you! Very cool Electric Grandma 😎
Within living memory for some of us, bits were expensive. A lot of work went into the conservation of bits, some residues of which are still with us. The advance of technology reduced the cost of bits, but not uniformly. Messaging charges were a way to price bits over the air, which cost more than bits over a wire. Then it wasn't worth charging for messages. Now we don't even worry (very much) about sending full-resolution pictures wirelessly: it's not worth the trouble to compact them.
We have gotten used to the idea that a compute bit per second is also essentially costless. If our demand for bps goes up by a few orders of magnitude, maybe that will no longer be the case, at least for some amount of time.
A long-winded way to suggest that we may have to invent a way to charge for searches and figure out ways to match willingness to pay with the necessary compute power to derive a response, ranging from trivial to insightful.
(Nice picture, BTW!)
Thank you, Dave!
Yes, Moore's law seems to apply to everything electronic, not just chips. Will this be forever? I also found Francis Turner's link (in his comment above) very interesting. "Actively Incinerating Cash."
It is a great picture. Enhanced, and great!
Been using Grok. So superior to google that I really don’t care if it uses more electricity. I’m getting tax advice for next year this morning.
One problem, I’m going to want to use it at night when the wind isn’t blowing. Glad we have an endless supply of natural gas, and maybe nuclear someday.
Thanks for your continued contribution to the sober discussion about energy!
I had no idea you were that hot!
AI does the superficial part. But did nothing about her being hot in soul!
Excellent. I have also written about this with the same kind of sentiment, people still matter. I appreciate you Meredith, Bass
This is awesome, Meredith...Thanks for sharing!
From a Grandma and classical violinist, who is concerned that "Beethoven is Rolling Over"!!
The energy consumption of today’s AI systems can be drastically reduced — while significantly improving functionality — if we integrate them with genuine human abilities like true intelligence, intuition, and emotion.
This is where IntraCavi comes in: a non-invasive interface that interacts directly with the human brain via five cranial nerves.
The broader HAIXUS approach proposes a shift in logic:
We don’t need to reinvent intelligence from scratch — we can amplify the one we already have.
By connecting the human brain to multiple AI systems, we unlock a new form of cognition — distributed, enhanced, and resource-efficient.
💡 No chips. No implants. Just humans — in symbiosis with intelligent systems.
The issue with ai is that’s it’s artificial, as in not real - fake. It’s also just lazy way to compile multiple correlations and regressions. One huge problem is the data force. And you believe everything from the internet? Then why even use ai? If you don’t understand the question you ask you need to do the work to learn about the subject, not just throw in a bunch of garbage to get some poorly validated input.
AI has been helpful to many people. It has more limitations than some people are willing to consider, but it isn't "throwing a bunch of garbage."
I'd say AI saves a lot of time and effort at the lower end of the AI levels as defined by Powell, thus making the higher levels of cognition for human reasoning. That of course must include checking for AI errors.
Agree.