I worked in computer operations, in data centers, way back in the 90s and early 00s. I had no idea what monsters had been leashed upon the world, nor what they would become. Great cartoon, Clay. We're all learning as we go along.
I don't store anything in the cloud. I really don't understand why people do when there are portable hard drives, flash drives, etc. to store information. And like anything else, they're vulnerable to be being hacked.
Don't download any apps, disabled or uninstalled all internet access, on the phone. I admit to having a solitaire game to play in doctors' offices. Do take photos, but I copy photos to the computer, and delete them from the phone, save for a couple. I hope the cloud, if they show up there, likes my doggies. Microsoft sends me notes saying my computer isn't linked to the cloud. Too bad, so sad - for them. And I always delete all cookies, etc. when I shut the computer off. My computer, very boring. :-)
Love it, Clay. A new data center has been proposed in Albany. It was announced late last week. People have already lined up to protest it. It is supposed to be a very small part of a larger development. We shall see. In general, I am not in favor of them turning up everywhere.
In 1981 when we moved to our first housing in this state it was the third floor (past servants’ quarters) of a Sanford White house inhabited by a spoiled and somewhat whacky old woman whose brother in another town provided partial oversight. We soon needed to get a post office box because she invariably tossed away my husband’s copies of the Journal of Artificial Intelligence due to objecting to the name.
Artificial Intelligence is the name of at least a dozen forms of AUTOMATION. Yes, that is what they are, types of automation.
There are some categories which have very narrow applications and work marvelously well with few errors. Think of ones which perform chores like figuring out protein folding variations.
There are other forms which raid materials on the internet which are incredibly error-prone. Those are the types people usually think of when they think “AI”.
Imagine a human who does not weigh resources. Easily done, since we all know some who think that everything is an “opinion” rather than some things being evidence based.
Now, realize that automation CAN NOT error check itself. (It simply takes too much time and is too costly.) so while specific repeated errors can be reduced in volume, the core problem of not even knowing that it is erring remains. That is what you get with forms of AI like art creating ones (which require hours of work by human artists to correct and turn in fun little videos), facial recognition (which reflects societal biases, troubles with shadowing, and proportions of reference images so makes more mistakes with minorities), and Large Language Models which is the category of automation that most people think of as AI.
Large Language Models not only have the aforementioned shortfalls, but are written to please the user. Not protect the user, not be accurate for the user, but to please. So, they cobble together fake references, false plots, new characters, etc. to please essay writers who never bothered to read books. They have given incorrect and fatal dosing information to at least one drug user. They have terrible results on tests of making medical diagnoses (with one exception at one of three category levels when it was compared with two physicians in different soecialties than what was being tested and who were given difference task definitions than the LLM — but even then the two physicians beat it at the next two levels).
There is now an error checking mathematical approach, but many consider it still too costly and slow.
Given all of these factors many expect LLMs to crash economically in a fashion echoing the first railroad investment boom in England. That basically became a practice round for what mostly did not work and bankrupted many. So, many people expect the current LLMs, given their many problems, to become a failed first try, but a later try, ironing out the problems, to succeed.
It's going to be great recession redux. Without government support because the R's have overextended debt to the point where the interest is now the biggest item in the budget.
Right now i am personally more concerned about people who are getting false medical “advice” or having procedures rejected by insurance. In 15 states RFK Jr’s health dept is having Medicare recipients’ access to certain procedures decided by AI which is programmed to pay the operating companies for saving money by denying procedures. Patients need to know that they can request that a human physician employed by the gov’t review the denial and that the vast majority do then get coverage.
In Sci Am there is a new medical piece about a nonexistant skin condition AI has diagnosed.
All true Clay, data centers are being oversold and overbuilt. That said, the things being built now are AI data centers, and they are the monsters sucking all the oxygen out of the room.
LLM cannot meet the promises being made for it. Nor can it make money in any reasonable way. There's always a "path" to being profitable, but the path for LLM's is nearly nonexistent. Much of the LLM hype is intended to boost Musk's IPO for SpaceX, which has absorbed the xAI entity. The plan seems to be to hype it insanely over the next two weeks, grab the IPO money, then absorb Tesla next year, driving stock prices far enough past actual value so that Musk can claim to be the first trillionaire. For a few days, anyhow.
This is all part of numerous bubbles that are preparing to burst. Blockchain coins are a bubble, one of the few that can actually destroy money totally, rather than just dropping. The "private" stock market which runs without oversight scares me. The trillion dollar interest payments on the national debt are insane and unsustainable. The LLM (Large Language Model) AI hype is absolutely a bubble, and it is a massive bubble indeed. NVIDIA, which is supplying most of the chips for this mess, will fall apart if they have to return to just being a video card vendor, and it's likely to happen. NVIDIA is currently about 13% of the NASDAQ, 8% of the S&P 500 and 2.5% of the DOW, all in one single company. AI linked stocks make up 55% of the value of the DOW. The AI bubble alone is enough to give any rational person a heart attack, and we've also got blockchain and housing and a dozen other crazy things going on.
The coming crash when these bubbles pop is going to be insane. And Trump is doing his best to inflate them all beyond reason.
Stand by for the SCROTUS to give data centers the same rights as human beings.
Data centers and AI.
I worked in computer operations, in data centers, way back in the 90s and early 00s. I had no idea what monsters had been leashed upon the world, nor what they would become. Great cartoon, Clay. We're all learning as we go along.
I don't store anything in the cloud. I really don't understand why people do when there are portable hard drives, flash drives, etc. to store information. And like anything else, they're vulnerable to be being hacked.
If you have photos on your phone or any data, you are probably using the cloud.
Don't download any apps, disabled or uninstalled all internet access, on the phone. I admit to having a solitaire game to play in doctors' offices. Do take photos, but I copy photos to the computer, and delete them from the phone, save for a couple. I hope the cloud, if they show up there, likes my doggies. Microsoft sends me notes saying my computer isn't linked to the cloud. Too bad, so sad - for them. And I always delete all cookies, etc. when I shut the computer off. My computer, very boring. :-)
Please keep educating me...I am weak on this subject.
Me, too - completely clueless!!
Love it, Clay. A new data center has been proposed in Albany. It was announced late last week. People have already lined up to protest it. It is supposed to be a very small part of a larger development. We shall see. In general, I am not in favor of them turning up everywhere.
“Mr. Bubbles” speaks for me, too! Unsurprising, the backlash to these data centers cropping up is growing daily.
Love all the bright colors, Peezy peeking out, black cat’s expression, and the perfect tune for the video.
I’m glad Goldy isn’t in that disgusting water.
BTW, that water used to cool data centers is typically released, at temperatures too high for survival of local life forms, into rivers and streams.
In 1981 when we moved to our first housing in this state it was the third floor (past servants’ quarters) of a Sanford White house inhabited by a spoiled and somewhat whacky old woman whose brother in another town provided partial oversight. We soon needed to get a post office box because she invariably tossed away my husband’s copies of the Journal of Artificial Intelligence due to objecting to the name.
Artificial Intelligence is the name of at least a dozen forms of AUTOMATION. Yes, that is what they are, types of automation.
There are some categories which have very narrow applications and work marvelously well with few errors. Think of ones which perform chores like figuring out protein folding variations.
There are other forms which raid materials on the internet which are incredibly error-prone. Those are the types people usually think of when they think “AI”.
Imagine a human who does not weigh resources. Easily done, since we all know some who think that everything is an “opinion” rather than some things being evidence based.
Now, realize that automation CAN NOT error check itself. (It simply takes too much time and is too costly.) so while specific repeated errors can be reduced in volume, the core problem of not even knowing that it is erring remains. That is what you get with forms of AI like art creating ones (which require hours of work by human artists to correct and turn in fun little videos), facial recognition (which reflects societal biases, troubles with shadowing, and proportions of reference images so makes more mistakes with minorities), and Large Language Models which is the category of automation that most people think of as AI.
Large Language Models not only have the aforementioned shortfalls, but are written to please the user. Not protect the user, not be accurate for the user, but to please. So, they cobble together fake references, false plots, new characters, etc. to please essay writers who never bothered to read books. They have given incorrect and fatal dosing information to at least one drug user. They have terrible results on tests of making medical diagnoses (with one exception at one of three category levels when it was compared with two physicians in different soecialties than what was being tested and who were given difference task definitions than the LLM — but even then the two physicians beat it at the next two levels).
There is now an error checking mathematical approach, but many consider it still too costly and slow.
Given all of these factors many expect LLMs to crash economically in a fashion echoing the first railroad investment boom in England. That basically became a practice round for what mostly did not work and bankrupted many. So, many people expect the current LLMs, given their many problems, to become a failed first try, but a later try, ironing out the problems, to succeed.
It's going to be great recession redux. Without government support because the R's have overextended debt to the point where the interest is now the biggest item in the budget.
Right now i am personally more concerned about people who are getting false medical “advice” or having procedures rejected by insurance. In 15 states RFK Jr’s health dept is having Medicare recipients’ access to certain procedures decided by AI which is programmed to pay the operating companies for saving money by denying procedures. Patients need to know that they can request that a human physician employed by the gov’t review the denial and that the vast majority do then get coverage.
In Sci Am there is a new medical piece about a nonexistant skin condition AI has diagnosed.
What is Mr. Bubbles supposed to be?
All true Clay, data centers are being oversold and overbuilt. That said, the things being built now are AI data centers, and they are the monsters sucking all the oxygen out of the room.
LLM cannot meet the promises being made for it. Nor can it make money in any reasonable way. There's always a "path" to being profitable, but the path for LLM's is nearly nonexistent. Much of the LLM hype is intended to boost Musk's IPO for SpaceX, which has absorbed the xAI entity. The plan seems to be to hype it insanely over the next two weeks, grab the IPO money, then absorb Tesla next year, driving stock prices far enough past actual value so that Musk can claim to be the first trillionaire. For a few days, anyhow.
This is all part of numerous bubbles that are preparing to burst. Blockchain coins are a bubble, one of the few that can actually destroy money totally, rather than just dropping. The "private" stock market which runs without oversight scares me. The trillion dollar interest payments on the national debt are insane and unsustainable. The LLM (Large Language Model) AI hype is absolutely a bubble, and it is a massive bubble indeed. NVIDIA, which is supplying most of the chips for this mess, will fall apart if they have to return to just being a video card vendor, and it's likely to happen. NVIDIA is currently about 13% of the NASDAQ, 8% of the S&P 500 and 2.5% of the DOW, all in one single company. AI linked stocks make up 55% of the value of the DOW. The AI bubble alone is enough to give any rational person a heart attack, and we've also got blockchain and housing and a dozen other crazy things going on.
The coming crash when these bubbles pop is going to be insane. And Trump is doing his best to inflate them all beyond reason.