|
Although this thread does not function under the same strict guidelines as the USPMT, it is still a general practice on TL to provide a source with an explanation on why it is relevant and what purpose it adds to the discussion. Failure to do so will result in a mod action. |
On May 05 2026 23:01 GreenHorizons wrote:Show nested quote +On May 05 2026 22:08 Jankisa wrote:+ Show Spoiler +I personally think that it's something of a bubble, just like the dot com was a bubble, but we still have Cisco as a huge company despite it being one that was riding the high crest of the bubble at the time.
The need for compute is not going to go away, the development work and inference will be needed for the foreseeable future, unless there is a cataclysmic event where insane amount of economic activity goes away, the compute that a DC like this will be churning out will find a buyer, at what prices, well, that is not for me to worry about, more for the investors.
Even if AI development stops right now, new way to apply the agents we currently have will keep popping up, as critical as I am of the people leading this industry on the US side, it's pretty obvious that this is not just hype, it's going to be the most important technology of the next decade and I'm yet to find a convincing argument against that.
The "these things have 50 people working once constructed" is a typical Reddit number being thrown around for a year now, it's silly. The rough calculation, for DC operations only and at Hyperscaler size is 0.5 employees per MW, this is a GW DC so the number of employees in the DC itself would be 200-500, depending on a variety of factors.
Like I mentioned before, this monstrosity, if built, would at peek capacity be using about a quarter of the energy of my whole country, this, however, wouldn't all come from our grid, a big part of this project is an LNG power plant + fields of solar, for which there is plenty of space around the area, these both come with it's own staff, so the upper estimate of 1.500 people working in and around this project doesn't seem crazy.
For the "it just makes shit up" people, yeah, sometimes it does, that's why no serious business or person is going to give it access to critical systems and allow it to delete shit (as recently publicized) or have it do free form data analysis that no one double checks for months at the time. + Show Spoiler +
As an example, when I try to get to the bottom of an issue, I often review logs, since there is a ton of data, I feed it in to the AI. Once it does a comparative analysis it will give me timestamps and I'll go and check if the lines are in the logs, anything else would be reckless.
I'm having a hard time figuring out if this is sarcasm or not? Seems like you actually believe it, but it also seems way too ridiculous/oblivious for anyone to seriously believe. Maybe I'm misunderstanding your qualifiers? An example that comes to mind: Show nested quote +UnitedHealth uses faulty AI to deny elderly patients medically necessary coverage
The lawsuit, filed last Tuesday in federal court in Minnesota, claims UnitedHealth illegally denied "elderly patients care owed to them under Medicare Advantage Plans" by deploying an AI model known by the company to have a 90% error rate, overriding determinations made by the patients' physicians that the expenses were medically necessary.
"The elderly are prematurely kicked out of care facilities nationwide or forced to deplete family savings to continue receiving necessary medical care, all because [UnitedHealth's] AI model 'disagrees' with their real live doctors' determinations," according to the complaint. https://www.cbsnews.com/news/unitedhealth-lawsuit-ai-deny-claims-medicare-advantage-health-insurance-denials/I don't know if AI (with a known 90% error rate) deployed by a top 10 fortune 500 company making life and death decisions for people counts as a "serious business" or a "critical system" to you though.
Simberto already replied to you, horrible companies using this to be able to shift responsibility for killing people doesn't mean that they are "serious people and organizations", I mean, they are, but they aren't using this in a genuine way where they expect high quality outputs, the goal is to utilize the low quality to shift responsibility.
American capitalism and healthcare being broken does not invalidate that AI as a technology, when used by intelligent people and companies who are trying to deliver good products faster then they could before, AI is an absolute boon to productivity.
|
On May 05 2026 22:08 Jankisa wrote:
For the "it just makes shit up" people, yeah, sometimes it does, that's why no serious business or person is going to give it access to critical systems and allow it to delete shit (as recently publicized) or have it do free form data analysis that no one double checks for months at the time.
As an example, when I try to get to the bottom of an issue, I often review logs, since there is a ton of data, I feed it in to the AI. Once it does a comparative analysis it will give me timestamps and I'll go and check if the lines are in the logs, anything else would be reckless.
That's a hope that you can have at the start, but once people get used to it and it usually looks decent they'll just go with it. When Google came out we used to verify that we weren't getting nonsense results on the interwebs too, and now we have grandpas getting their news from Facebook.
|
On May 06 2026 00:00 Jankisa wrote:Show nested quote +On May 05 2026 23:01 GreenHorizons wrote:On May 05 2026 22:08 Jankisa wrote:+ Show Spoiler +I personally think that it's something of a bubble, just like the dot com was a bubble, but we still have Cisco as a huge company despite it being one that was riding the high crest of the bubble at the time.
The need for compute is not going to go away, the development work and inference will be needed for the foreseeable future, unless there is a cataclysmic event where insane amount of economic activity goes away, the compute that a DC like this will be churning out will find a buyer, at what prices, well, that is not for me to worry about, more for the investors.
Even if AI development stops right now, new way to apply the agents we currently have will keep popping up, as critical as I am of the people leading this industry on the US side, it's pretty obvious that this is not just hype, it's going to be the most important technology of the next decade and I'm yet to find a convincing argument against that.
The "these things have 50 people working once constructed" is a typical Reddit number being thrown around for a year now, it's silly. The rough calculation, for DC operations only and at Hyperscaler size is 0.5 employees per MW, this is a GW DC so the number of employees in the DC itself would be 200-500, depending on a variety of factors.
Like I mentioned before, this monstrosity, if built, would at peek capacity be using about a quarter of the energy of my whole country, this, however, wouldn't all come from our grid, a big part of this project is an LNG power plant + fields of solar, for which there is plenty of space around the area, these both come with it's own staff, so the upper estimate of 1.500 people working in and around this project doesn't seem crazy.
For the "it just makes shit up" people, yeah, sometimes it does, that's why no serious business or person is going to give it access to critical systems and allow it to delete shit (as recently publicized) or have it do free form data analysis that no one double checks for months at the time. + Show Spoiler +
As an example, when I try to get to the bottom of an issue, I often review logs, since there is a ton of data, I feed it in to the AI. Once it does a comparative analysis it will give me timestamps and I'll go and check if the lines are in the logs, anything else would be reckless.
I'm having a hard time figuring out if this is sarcasm or not? Seems like you actually believe it, but it also seems way too ridiculous/oblivious for anyone to seriously believe. Maybe I'm misunderstanding your qualifiers? An example that comes to mind: UnitedHealth uses faulty AI to deny elderly patients medically necessary coverage
The lawsuit, filed last Tuesday in federal court in Minnesota, claims UnitedHealth illegally denied "elderly patients care owed to them under Medicare Advantage Plans" by deploying an AI model known by the company to have a 90% error rate, overriding determinations made by the patients' physicians that the expenses were medically necessary.
"The elderly are prematurely kicked out of care facilities nationwide or forced to deplete family savings to continue receiving necessary medical care, all because [UnitedHealth's] AI model 'disagrees' with their real live doctors' determinations," according to the complaint. https://www.cbsnews.com/news/unitedhealth-lawsuit-ai-deny-claims-medicare-advantage-health-insurance-denials/I don't know if AI (with a known 90% error rate) deployed by a top 10 fortune 500 company making life and death decisions for people counts as a "serious business" or a "critical system" to you though. Simberto already replied to you, horrible companies using this to be able to shift responsibility for killing people doesn't mean that they are "serious people and organizations", I mean, they are, but they aren't using this in a genuine way where they expect high quality outputs, the goal is to utilize the low quality to shift responsibility. American capitalism and healthcare being broken does not invalidate that AI as a technology, when used by intelligent people and companies who are trying to deliver good products faster then they could before, AI is an absolute boon to productivity.
There was the case of an AI bot inciting violence or suicide or something like that. When you get machines performing human tasks, corporations can shift responsibility for harm caused to actual people to the consumer who sees the content tailored to them. It's a premise that removes accountability altogether for corporations but not for users.
If I build a robot that goes around shooting people, can I say it's the bots fault?
|
There was the case of an AI bot inciting violence or suicide or something like that. When you get machines performing human tasks, corporations can shift responsibility for harm caused to actual people to the consumer who sees the content tailored to them. It's a premise that removes accountability altogether for corporations but not for users.
I hope that won’t be the case. Because in my understanding currently companies are always held responsible for those inviduals’ actions and decisions who work for the company in any form. So just because companies replace some part of their work or decision making with AI. They should be still responsible of choosing AI and the results those come from that.
Edit: I realise that they start employing less and less people is a different question
|
Even if we assume that we should always blame the manufacturer of the hardware and never the user, it is really hard to say who exactly should be punished. Let's assume that an automated car goes mad and starts killing people. Who should be held responsible?
Company A - car maker
CEO? - The guy who most probably had absolutely nothing to do with the error itself. Although he might have had set conditions that led to it.
Process people - people who set rules by which others worked?
Quality control people?
Their manager?
Carline designer?
Company B - main software provider (in 95% of cases this will be different company)
CEO? Process people? Developers? Quality control? Middle way managers?
Company C - main hardware provider (again in 90% this will be a different company than A and B.)
Company D - stack supplier. Again, different company than all the above
Comapny E - chipset provider.
Company F - OS provider.
Company G - application provider.
Company H - OTA provider.
The creator of the open-source repository which contributed to issue? The guy who made the last MR?
I can also imagine, in some cases blaming industry standard designers and government regulations.
What I mean by all this is that often modern systems are so complicated, and so many people work on them, that pinpointing WHO exactly is to blame for some behavior is close to impossible. Especially if this is some corner case - once-in-several million behavior.
|
On May 06 2026 00:14 Nebuchad wrote:Show nested quote +On May 05 2026 22:08 Jankisa wrote:
For the "it just makes shit up" people, yeah, sometimes it does, that's why no serious business or person is going to give it access to critical systems and allow it to delete shit (as recently publicized) or have it do free form data analysis that no one double checks for months at the time.
As an example, when I try to get to the bottom of an issue, I often review logs, since there is a ton of data, I feed it in to the AI. Once it does a comparative analysis it will give me timestamps and I'll go and check if the lines are in the logs, anything else would be reckless. That's a hope that you can have at the start, but once people get used to it and it usually looks decent they'll just go with it. When Google came out we used to verify that we weren't getting nonsense results on the interwebs too, and now we have grandpas getting their news from Facebook.
Yeah, but the people like you and me who grew up with internet search becoming what it is still understand that you need to double check shit if you are going to base decisions on it.
Don't get me wrong, the whole hallucination issue is something that worries me, but, as I've been using, on daily basis AI since the first GPT until now, first mostly for private fucking around and then very much for professional tasks for the last 3 or so years, the hallucination rate is way down especially if you are using premium models.
That is the case both for self hosted ones, ones accessed via API and talking to them via chat interfaces, the real issue I've been seeing more and more is that it "forgets" things from relatively close time frame of the conversations, to me, it seems like the memory context is one of the most expensive aspects of it to maintain and they've (in my case Anthropic) been skimping on it, a lot.
For cheaper and models with less guardrails, I'm guessing that incentives to make shit up instead of doing the actual work took over, which, you know, is a very human thing to do.
It's all very strange, same as with the responsibility.
If my kid kills someone, who is to blame. Is it me for raising a monster. Is it his or hers mom for not instilling the correct values. Is it the system, the school, some random person traumatizing them at some point, and now, on top of all of that we also get to worry about them being radicalized by AI.
In my experience, these are very capable systems and most fuckups happen because humans are trying to save money, at least recently, and maybe that makes the whole endeavor unprofitable, but I have a feeling that optimizations will come and even if we stay at this level of capabilities the technology itself is here to stay and will transform the world.
If we get to AGI/ASI, well, then all bets are off and who the fuck knows.
|
There are plenty of people who grew up with the internet that are down one rabbit hole or another. I don’t think that growing up with the internet has really been that much of a protection. And given how much data they now have on all of us and the power of AI, I suspect it’s going to get a lot worse not better.
|
On May 06 2026 00:22 Vivax wrote:Show nested quote +On May 06 2026 00:00 Jankisa wrote:On May 05 2026 23:01 GreenHorizons wrote:On May 05 2026 22:08 Jankisa wrote:+ Show Spoiler +I personally think that it's something of a bubble, just like the dot com was a bubble, but we still have Cisco as a huge company despite it being one that was riding the high crest of the bubble at the time.
The need for compute is not going to go away, the development work and inference will be needed for the foreseeable future, unless there is a cataclysmic event where insane amount of economic activity goes away, the compute that a DC like this will be churning out will find a buyer, at what prices, well, that is not for me to worry about, more for the investors.
Even if AI development stops right now, new way to apply the agents we currently have will keep popping up, as critical as I am of the people leading this industry on the US side, it's pretty obvious that this is not just hype, it's going to be the most important technology of the next decade and I'm yet to find a convincing argument against that.
The "these things have 50 people working once constructed" is a typical Reddit number being thrown around for a year now, it's silly. The rough calculation, for DC operations only and at Hyperscaler size is 0.5 employees per MW, this is a GW DC so the number of employees in the DC itself would be 200-500, depending on a variety of factors.
Like I mentioned before, this monstrosity, if built, would at peek capacity be using about a quarter of the energy of my whole country, this, however, wouldn't all come from our grid, a big part of this project is an LNG power plant + fields of solar, for which there is plenty of space around the area, these both come with it's own staff, so the upper estimate of 1.500 people working in and around this project doesn't seem crazy.
For the "it just makes shit up" people, yeah, sometimes it does, that's why no serious business or person is going to give it access to critical systems and allow it to delete shit (as recently publicized) or have it do free form data analysis that no one double checks for months at the time. + Show Spoiler +
As an example, when I try to get to the bottom of an issue, I often review logs, since there is a ton of data, I feed it in to the AI. Once it does a comparative analysis it will give me timestamps and I'll go and check if the lines are in the logs, anything else would be reckless.
I'm having a hard time figuring out if this is sarcasm or not? Seems like you actually believe it, but it also seems way too ridiculous/oblivious for anyone to seriously believe. Maybe I'm misunderstanding your qualifiers? An example that comes to mind: UnitedHealth uses faulty AI to deny elderly patients medically necessary coverage
The lawsuit, filed last Tuesday in federal court in Minnesota, claims UnitedHealth illegally denied "elderly patients care owed to them under Medicare Advantage Plans" by deploying an AI model known by the company to have a 90% error rate, overriding determinations made by the patients' physicians that the expenses were medically necessary.
"The elderly are prematurely kicked out of care facilities nationwide or forced to deplete family savings to continue receiving necessary medical care, all because [UnitedHealth's] AI model 'disagrees' with their real live doctors' determinations," according to the complaint. https://www.cbsnews.com/news/unitedhealth-lawsuit-ai-deny-claims-medicare-advantage-health-insurance-denials/I don't know if AI (with a known 90% error rate) deployed by a top 10 fortune 500 company making life and death decisions for people counts as a "serious business" or a "critical system" to you though. Simberto already replied to you, horrible companies using this to be able to shift responsibility for killing people doesn't mean that they are "serious people and organizations", I mean, they are, but they aren't using this in a genuine way where they expect high quality outputs, the goal is to utilize the low quality to shift responsibility. American capitalism and healthcare being broken does not invalidate that AI as a technology, when used by intelligent people and companies who are trying to deliver good products faster then they could before, AI is an absolute boon to productivity. There was the case of an AI bot inciting violence or suicide or something like that. When you get machines performing human tasks, corporations can shift responsibility for harm caused to actual people to the consumer who sees the content tailored to them. It's a premise that removes accountability altogether for corporations but not for users. If I build a robot that goes around shooting people, can I say it's the bots fault?
This question is the main driving force (huhu pun) for the car industry and level 3+ autonomous driving. Much legal groundwork should be done already I imagine.
|
On May 06 2026 20:47 Harris1st wrote:Show nested quote +On May 06 2026 00:22 Vivax wrote:On May 06 2026 00:00 Jankisa wrote:On May 05 2026 23:01 GreenHorizons wrote:On May 05 2026 22:08 Jankisa wrote:+ Show Spoiler +I personally think that it's something of a bubble, just like the dot com was a bubble, but we still have Cisco as a huge company despite it being one that was riding the high crest of the bubble at the time.
The need for compute is not going to go away, the development work and inference will be needed for the foreseeable future, unless there is a cataclysmic event where insane amount of economic activity goes away, the compute that a DC like this will be churning out will find a buyer, at what prices, well, that is not for me to worry about, more for the investors.
Even if AI development stops right now, new way to apply the agents we currently have will keep popping up, as critical as I am of the people leading this industry on the US side, it's pretty obvious that this is not just hype, it's going to be the most important technology of the next decade and I'm yet to find a convincing argument against that.
The "these things have 50 people working once constructed" is a typical Reddit number being thrown around for a year now, it's silly. The rough calculation, for DC operations only and at Hyperscaler size is 0.5 employees per MW, this is a GW DC so the number of employees in the DC itself would be 200-500, depending on a variety of factors.
Like I mentioned before, this monstrosity, if built, would at peek capacity be using about a quarter of the energy of my whole country, this, however, wouldn't all come from our grid, a big part of this project is an LNG power plant + fields of solar, for which there is plenty of space around the area, these both come with it's own staff, so the upper estimate of 1.500 people working in and around this project doesn't seem crazy.
For the "it just makes shit up" people, yeah, sometimes it does, that's why no serious business or person is going to give it access to critical systems and allow it to delete shit (as recently publicized) or have it do free form data analysis that no one double checks for months at the time. + Show Spoiler +
As an example, when I try to get to the bottom of an issue, I often review logs, since there is a ton of data, I feed it in to the AI. Once it does a comparative analysis it will give me timestamps and I'll go and check if the lines are in the logs, anything else would be reckless.
I'm having a hard time figuring out if this is sarcasm or not? Seems like you actually believe it, but it also seems way too ridiculous/oblivious for anyone to seriously believe. Maybe I'm misunderstanding your qualifiers? An example that comes to mind: UnitedHealth uses faulty AI to deny elderly patients medically necessary coverage
The lawsuit, filed last Tuesday in federal court in Minnesota, claims UnitedHealth illegally denied "elderly patients care owed to them under Medicare Advantage Plans" by deploying an AI model known by the company to have a 90% error rate, overriding determinations made by the patients' physicians that the expenses were medically necessary.
"The elderly are prematurely kicked out of care facilities nationwide or forced to deplete family savings to continue receiving necessary medical care, all because [UnitedHealth's] AI model 'disagrees' with their real live doctors' determinations," according to the complaint. https://www.cbsnews.com/news/unitedhealth-lawsuit-ai-deny-claims-medicare-advantage-health-insurance-denials/I don't know if AI (with a known 90% error rate) deployed by a top 10 fortune 500 company making life and death decisions for people counts as a "serious business" or a "critical system" to you though. Simberto already replied to you, horrible companies using this to be able to shift responsibility for killing people doesn't mean that they are "serious people and organizations", I mean, they are, but they aren't using this in a genuine way where they expect high quality outputs, the goal is to utilize the low quality to shift responsibility. American capitalism and healthcare being broken does not invalidate that AI as a technology, when used by intelligent people and companies who are trying to deliver good products faster then they could before, AI is an absolute boon to productivity. There was the case of an AI bot inciting violence or suicide or something like that. When you get machines performing human tasks, corporations can shift responsibility for harm caused to actual people to the consumer who sees the content tailored to them. It's a premise that removes accountability altogether for corporations but not for users. If I build a robot that goes around shooting people, can I say it's the bots fault? This question is the main driving force (huhu pun) for the car industry and level 3+ autonomous driving. Much legal groundwork should be done already I imagine.
What's the fun in being in a car if you don't drive it. My opinion at least. But yeah, it's an angle that drives progress in other areas at least.
Seems like a lot of extra work for insurers. Maybe they needed that. One of their side income sources are drunk drivers causing accidents, now they can add machines to the mix.
|
On May 06 2026 21:15 Vivax wrote:Show nested quote +On May 06 2026 20:47 Harris1st wrote:On May 06 2026 00:22 Vivax wrote:On May 06 2026 00:00 Jankisa wrote:On May 05 2026 23:01 GreenHorizons wrote:On May 05 2026 22:08 Jankisa wrote:+ Show Spoiler +I personally think that it's something of a bubble, just like the dot com was a bubble, but we still have Cisco as a huge company despite it being one that was riding the high crest of the bubble at the time.
The need for compute is not going to go away, the development work and inference will be needed for the foreseeable future, unless there is a cataclysmic event where insane amount of economic activity goes away, the compute that a DC like this will be churning out will find a buyer, at what prices, well, that is not for me to worry about, more for the investors.
Even if AI development stops right now, new way to apply the agents we currently have will keep popping up, as critical as I am of the people leading this industry on the US side, it's pretty obvious that this is not just hype, it's going to be the most important technology of the next decade and I'm yet to find a convincing argument against that.
The "these things have 50 people working once constructed" is a typical Reddit number being thrown around for a year now, it's silly. The rough calculation, for DC operations only and at Hyperscaler size is 0.5 employees per MW, this is a GW DC so the number of employees in the DC itself would be 200-500, depending on a variety of factors.
Like I mentioned before, this monstrosity, if built, would at peek capacity be using about a quarter of the energy of my whole country, this, however, wouldn't all come from our grid, a big part of this project is an LNG power plant + fields of solar, for which there is plenty of space around the area, these both come with it's own staff, so the upper estimate of 1.500 people working in and around this project doesn't seem crazy.
For the "it just makes shit up" people, yeah, sometimes it does, that's why no serious business or person is going to give it access to critical systems and allow it to delete shit (as recently publicized) or have it do free form data analysis that no one double checks for months at the time. + Show Spoiler +
As an example, when I try to get to the bottom of an issue, I often review logs, since there is a ton of data, I feed it in to the AI. Once it does a comparative analysis it will give me timestamps and I'll go and check if the lines are in the logs, anything else would be reckless.
I'm having a hard time figuring out if this is sarcasm or not? Seems like you actually believe it, but it also seems way too ridiculous/oblivious for anyone to seriously believe. Maybe I'm misunderstanding your qualifiers? An example that comes to mind: UnitedHealth uses faulty AI to deny elderly patients medically necessary coverage
The lawsuit, filed last Tuesday in federal court in Minnesota, claims UnitedHealth illegally denied "elderly patients care owed to them under Medicare Advantage Plans" by deploying an AI model known by the company to have a 90% error rate, overriding determinations made by the patients' physicians that the expenses were medically necessary.
"The elderly are prematurely kicked out of care facilities nationwide or forced to deplete family savings to continue receiving necessary medical care, all because [UnitedHealth's] AI model 'disagrees' with their real live doctors' determinations," according to the complaint. https://www.cbsnews.com/news/unitedhealth-lawsuit-ai-deny-claims-medicare-advantage-health-insurance-denials/I don't know if AI (with a known 90% error rate) deployed by a top 10 fortune 500 company making life and death decisions for people counts as a "serious business" or a "critical system" to you though. Simberto already replied to you, horrible companies using this to be able to shift responsibility for killing people doesn't mean that they are "serious people and organizations", I mean, they are, but they aren't using this in a genuine way where they expect high quality outputs, the goal is to utilize the low quality to shift responsibility. American capitalism and healthcare being broken does not invalidate that AI as a technology, when used by intelligent people and companies who are trying to deliver good products faster then they could before, AI is an absolute boon to productivity. There was the case of an AI bot inciting violence or suicide or something like that. When you get machines performing human tasks, corporations can shift responsibility for harm caused to actual people to the consumer who sees the content tailored to them. It's a premise that removes accountability altogether for corporations but not for users. If I build a robot that goes around shooting people, can I say it's the bots fault? This question is the main driving force (huhu pun) for the car industry and level 3+ autonomous driving. Much legal groundwork should be done already I imagine. What's the fun in being in a car if you don't drive it. My opinion at least. But yeah, it's an angle that drives progress in other areas at least. Seems like a lot of extra work for insurers. Maybe they needed that. One of their side income sources are drunk drivers causing accidents, now they can add machines to the mix.
To a lot of people, cars are not for fun, but a tool to get to some place. And I'd much rather get to a place without having to concentrate on traffic all the time. Which is why i like taking trains, and like driving a lot less than riding a train.
|
On May 06 2026 21:15 Vivax wrote:Show nested quote +On May 06 2026 20:47 Harris1st wrote:On May 06 2026 00:22 Vivax wrote:On May 06 2026 00:00 Jankisa wrote:On May 05 2026 23:01 GreenHorizons wrote:On May 05 2026 22:08 Jankisa wrote:+ Show Spoiler +I personally think that it's something of a bubble, just like the dot com was a bubble, but we still have Cisco as a huge company despite it being one that was riding the high crest of the bubble at the time.
The need for compute is not going to go away, the development work and inference will be needed for the foreseeable future, unless there is a cataclysmic event where insane amount of economic activity goes away, the compute that a DC like this will be churning out will find a buyer, at what prices, well, that is not for me to worry about, more for the investors.
Even if AI development stops right now, new way to apply the agents we currently have will keep popping up, as critical as I am of the people leading this industry on the US side, it's pretty obvious that this is not just hype, it's going to be the most important technology of the next decade and I'm yet to find a convincing argument against that.
The "these things have 50 people working once constructed" is a typical Reddit number being thrown around for a year now, it's silly. The rough calculation, for DC operations only and at Hyperscaler size is 0.5 employees per MW, this is a GW DC so the number of employees in the DC itself would be 200-500, depending on a variety of factors.
Like I mentioned before, this monstrosity, if built, would at peek capacity be using about a quarter of the energy of my whole country, this, however, wouldn't all come from our grid, a big part of this project is an LNG power plant + fields of solar, for which there is plenty of space around the area, these both come with it's own staff, so the upper estimate of 1.500 people working in and around this project doesn't seem crazy.
For the "it just makes shit up" people, yeah, sometimes it does, that's why no serious business or person is going to give it access to critical systems and allow it to delete shit (as recently publicized) or have it do free form data analysis that no one double checks for months at the time. + Show Spoiler +
As an example, when I try to get to the bottom of an issue, I often review logs, since there is a ton of data, I feed it in to the AI. Once it does a comparative analysis it will give me timestamps and I'll go and check if the lines are in the logs, anything else would be reckless.
I'm having a hard time figuring out if this is sarcasm or not? Seems like you actually believe it, but it also seems way too ridiculous/oblivious for anyone to seriously believe. Maybe I'm misunderstanding your qualifiers? An example that comes to mind: UnitedHealth uses faulty AI to deny elderly patients medically necessary coverage
The lawsuit, filed last Tuesday in federal court in Minnesota, claims UnitedHealth illegally denied "elderly patients care owed to them under Medicare Advantage Plans" by deploying an AI model known by the company to have a 90% error rate, overriding determinations made by the patients' physicians that the expenses were medically necessary.
"The elderly are prematurely kicked out of care facilities nationwide or forced to deplete family savings to continue receiving necessary medical care, all because [UnitedHealth's] AI model 'disagrees' with their real live doctors' determinations," according to the complaint. https://www.cbsnews.com/news/unitedhealth-lawsuit-ai-deny-claims-medicare-advantage-health-insurance-denials/I don't know if AI (with a known 90% error rate) deployed by a top 10 fortune 500 company making life and death decisions for people counts as a "serious business" or a "critical system" to you though. Simberto already replied to you, horrible companies using this to be able to shift responsibility for killing people doesn't mean that they are "serious people and organizations", I mean, they are, but they aren't using this in a genuine way where they expect high quality outputs, the goal is to utilize the low quality to shift responsibility. American capitalism and healthcare being broken does not invalidate that AI as a technology, when used by intelligent people and companies who are trying to deliver good products faster then they could before, AI is an absolute boon to productivity. There was the case of an AI bot inciting violence or suicide or something like that. When you get machines performing human tasks, corporations can shift responsibility for harm caused to actual people to the consumer who sees the content tailored to them. It's a premise that removes accountability altogether for corporations but not for users. If I build a robot that goes around shooting people, can I say it's the bots fault? This question is the main driving force (huhu pun) for the car industry and level 3+ autonomous driving. Much legal groundwork should be done already I imagine. What's the fun in being in a car if you don't drive it. My opinion at least. But yeah, it's an angle that drives progress in other areas at least. Seems like a lot of extra work for insurers. Maybe they needed that. One of their side income sources are drunk drivers causing accidents, now they can add machines to the mix.
I guess in the end it will be a insurance question: lets say self driving cars can be insured for a premium for example. Taking the person "driving" as well as the manufacturer out of the equation for direct responsibility. I guess the same could be done for AI and data in general. Ofc somewhere on the way there will be fuckups so huge that no insurance company will cover them. But that is more of a market issue than law issue then.
|
On May 06 2026 21:21 Simberto wrote:Show nested quote +On May 06 2026 21:15 Vivax wrote:On May 06 2026 20:47 Harris1st wrote:On May 06 2026 00:22 Vivax wrote:On May 06 2026 00:00 Jankisa wrote:On May 05 2026 23:01 GreenHorizons wrote:On May 05 2026 22:08 Jankisa wrote:+ Show Spoiler +I personally think that it's something of a bubble, just like the dot com was a bubble, but we still have Cisco as a huge company despite it being one that was riding the high crest of the bubble at the time.
The need for compute is not going to go away, the development work and inference will be needed for the foreseeable future, unless there is a cataclysmic event where insane amount of economic activity goes away, the compute that a DC like this will be churning out will find a buyer, at what prices, well, that is not for me to worry about, more for the investors.
Even if AI development stops right now, new way to apply the agents we currently have will keep popping up, as critical as I am of the people leading this industry on the US side, it's pretty obvious that this is not just hype, it's going to be the most important technology of the next decade and I'm yet to find a convincing argument against that.
The "these things have 50 people working once constructed" is a typical Reddit number being thrown around for a year now, it's silly. The rough calculation, for DC operations only and at Hyperscaler size is 0.5 employees per MW, this is a GW DC so the number of employees in the DC itself would be 200-500, depending on a variety of factors.
Like I mentioned before, this monstrosity, if built, would at peek capacity be using about a quarter of the energy of my whole country, this, however, wouldn't all come from our grid, a big part of this project is an LNG power plant + fields of solar, for which there is plenty of space around the area, these both come with it's own staff, so the upper estimate of 1.500 people working in and around this project doesn't seem crazy.
For the "it just makes shit up" people, yeah, sometimes it does, that's why no serious business or person is going to give it access to critical systems and allow it to delete shit (as recently publicized) or have it do free form data analysis that no one double checks for months at the time. + Show Spoiler +
As an example, when I try to get to the bottom of an issue, I often review logs, since there is a ton of data, I feed it in to the AI. Once it does a comparative analysis it will give me timestamps and I'll go and check if the lines are in the logs, anything else would be reckless.
I'm having a hard time figuring out if this is sarcasm or not? Seems like you actually believe it, but it also seems way too ridiculous/oblivious for anyone to seriously believe. Maybe I'm misunderstanding your qualifiers? An example that comes to mind: UnitedHealth uses faulty AI to deny elderly patients medically necessary coverage
The lawsuit, filed last Tuesday in federal court in Minnesota, claims UnitedHealth illegally denied "elderly patients care owed to them under Medicare Advantage Plans" by deploying an AI model known by the company to have a 90% error rate, overriding determinations made by the patients' physicians that the expenses were medically necessary.
"The elderly are prematurely kicked out of care facilities nationwide or forced to deplete family savings to continue receiving necessary medical care, all because [UnitedHealth's] AI model 'disagrees' with their real live doctors' determinations," according to the complaint. https://www.cbsnews.com/news/unitedhealth-lawsuit-ai-deny-claims-medicare-advantage-health-insurance-denials/I don't know if AI (with a known 90% error rate) deployed by a top 10 fortune 500 company making life and death decisions for people counts as a "serious business" or a "critical system" to you though. Simberto already replied to you, horrible companies using this to be able to shift responsibility for killing people doesn't mean that they are "serious people and organizations", I mean, they are, but they aren't using this in a genuine way where they expect high quality outputs, the goal is to utilize the low quality to shift responsibility. American capitalism and healthcare being broken does not invalidate that AI as a technology, when used by intelligent people and companies who are trying to deliver good products faster then they could before, AI is an absolute boon to productivity. There was the case of an AI bot inciting violence or suicide or something like that. When you get machines performing human tasks, corporations can shift responsibility for harm caused to actual people to the consumer who sees the content tailored to them. It's a premise that removes accountability altogether for corporations but not for users. If I build a robot that goes around shooting people, can I say it's the bots fault? This question is the main driving force (huhu pun) for the car industry and level 3+ autonomous driving. Much legal groundwork should be done already I imagine. What's the fun in being in a car if you don't drive it. My opinion at least. But yeah, it's an angle that drives progress in other areas at least. Seems like a lot of extra work for insurers. Maybe they needed that. One of their side income sources are drunk drivers causing accidents, now they can add machines to the mix. To a lot of people, cars are not for fun, but a tool to get to some place. And I'd much rather get to a place without having to concentrate on traffic all the time. Which is why i like taking trains, and like driving a lot less than riding a train.
Would love to take trains but that is not possible for me.
Everyday when I get into the office my heart rate is already through the roof because of some reckless drivers (and sometimes old / slow drivers, but that is a me problem ^^' ).
A world where every car is driven by a computer under the same ruleset is a day that can't come soon enough. Traffic jams will be severly lessened. Accidents between cars will be at almost zero. Accidents with pedestrians and cyclist should be zero as well.
|
On May 06 2026 21:28 Harris1st wrote:Show nested quote +On May 06 2026 21:21 Simberto wrote:On May 06 2026 21:15 Vivax wrote:On May 06 2026 20:47 Harris1st wrote:On May 06 2026 00:22 Vivax wrote:On May 06 2026 00:00 Jankisa wrote:On May 05 2026 23:01 GreenHorizons wrote:On May 05 2026 22:08 Jankisa wrote:+ Show Spoiler +I personally think that it's something of a bubble, just like the dot com was a bubble, but we still have Cisco as a huge company despite it being one that was riding the high crest of the bubble at the time.
The need for compute is not going to go away, the development work and inference will be needed for the foreseeable future, unless there is a cataclysmic event where insane amount of economic activity goes away, the compute that a DC like this will be churning out will find a buyer, at what prices, well, that is not for me to worry about, more for the investors.
Even if AI development stops right now, new way to apply the agents we currently have will keep popping up, as critical as I am of the people leading this industry on the US side, it's pretty obvious that this is not just hype, it's going to be the most important technology of the next decade and I'm yet to find a convincing argument against that.
The "these things have 50 people working once constructed" is a typical Reddit number being thrown around for a year now, it's silly. The rough calculation, for DC operations only and at Hyperscaler size is 0.5 employees per MW, this is a GW DC so the number of employees in the DC itself would be 200-500, depending on a variety of factors.
Like I mentioned before, this monstrosity, if built, would at peek capacity be using about a quarter of the energy of my whole country, this, however, wouldn't all come from our grid, a big part of this project is an LNG power plant + fields of solar, for which there is plenty of space around the area, these both come with it's own staff, so the upper estimate of 1.500 people working in and around this project doesn't seem crazy.
For the "it just makes shit up" people, yeah, sometimes it does, that's why no serious business or person is going to give it access to critical systems and allow it to delete shit (as recently publicized) or have it do free form data analysis that no one double checks for months at the time. + Show Spoiler +
As an example, when I try to get to the bottom of an issue, I often review logs, since there is a ton of data, I feed it in to the AI. Once it does a comparative analysis it will give me timestamps and I'll go and check if the lines are in the logs, anything else would be reckless.
I'm having a hard time figuring out if this is sarcasm or not? Seems like you actually believe it, but it also seems way too ridiculous/oblivious for anyone to seriously believe. Maybe I'm misunderstanding your qualifiers? An example that comes to mind: UnitedHealth uses faulty AI to deny elderly patients medically necessary coverage
The lawsuit, filed last Tuesday in federal court in Minnesota, claims UnitedHealth illegally denied "elderly patients care owed to them under Medicare Advantage Plans" by deploying an AI model known by the company to have a 90% error rate, overriding determinations made by the patients' physicians that the expenses were medically necessary.
"The elderly are prematurely kicked out of care facilities nationwide or forced to deplete family savings to continue receiving necessary medical care, all because [UnitedHealth's] AI model 'disagrees' with their real live doctors' determinations," according to the complaint. https://www.cbsnews.com/news/unitedhealth-lawsuit-ai-deny-claims-medicare-advantage-health-insurance-denials/I don't know if AI (with a known 90% error rate) deployed by a top 10 fortune 500 company making life and death decisions for people counts as a "serious business" or a "critical system" to you though. Simberto already replied to you, horrible companies using this to be able to shift responsibility for killing people doesn't mean that they are "serious people and organizations", I mean, they are, but they aren't using this in a genuine way where they expect high quality outputs, the goal is to utilize the low quality to shift responsibility. American capitalism and healthcare being broken does not invalidate that AI as a technology, when used by intelligent people and companies who are trying to deliver good products faster then they could before, AI is an absolute boon to productivity. There was the case of an AI bot inciting violence or suicide or something like that. When you get machines performing human tasks, corporations can shift responsibility for harm caused to actual people to the consumer who sees the content tailored to them. It's a premise that removes accountability altogether for corporations but not for users. If I build a robot that goes around shooting people, can I say it's the bots fault? This question is the main driving force (huhu pun) for the car industry and level 3+ autonomous driving. Much legal groundwork should be done already I imagine. What's the fun in being in a car if you don't drive it. My opinion at least. But yeah, it's an angle that drives progress in other areas at least. Seems like a lot of extra work for insurers. Maybe they needed that. One of their side income sources are drunk drivers causing accidents, now they can add machines to the mix. To a lot of people, cars are not for fun, but a tool to get to some place. And I'd much rather get to a place without having to concentrate on traffic all the time. Which is why i like taking trains, and like driving a lot less than riding a train. Would love to take trains but that is not possible for me. Everyday when I get into the office my heart rate is already through the roof because of some reckless drivers (and sometimes old / slow drivers, but that is a me problem ^^' ). A world where every car is driven by a computer under the same ruleset is a day that can't come soon enough. Traffic jams will be severly lessened. Accidents between cars will be at almost zero. Accidents with pedestrians and cyclist should be zero as well.
I can highly recommend a bicycle for commutes <= 10 km. To the point where i would recommend moving that close to your work if you intend to stay there for a longer period of time. The quality of life improvement from being able to bike to work in half an hour is massive.
|
On May 06 2026 21:28 Harris1st wrote:Show nested quote +On May 06 2026 21:21 Simberto wrote:On May 06 2026 21:15 Vivax wrote:On May 06 2026 20:47 Harris1st wrote:On May 06 2026 00:22 Vivax wrote:On May 06 2026 00:00 Jankisa wrote:On May 05 2026 23:01 GreenHorizons wrote:On May 05 2026 22:08 Jankisa wrote:+ Show Spoiler +I personally think that it's something of a bubble, just like the dot com was a bubble, but we still have Cisco as a huge company despite it being one that was riding the high crest of the bubble at the time.
The need for compute is not going to go away, the development work and inference will be needed for the foreseeable future, unless there is a cataclysmic event where insane amount of economic activity goes away, the compute that a DC like this will be churning out will find a buyer, at what prices, well, that is not for me to worry about, more for the investors.
Even if AI development stops right now, new way to apply the agents we currently have will keep popping up, as critical as I am of the people leading this industry on the US side, it's pretty obvious that this is not just hype, it's going to be the most important technology of the next decade and I'm yet to find a convincing argument against that.
The "these things have 50 people working once constructed" is a typical Reddit number being thrown around for a year now, it's silly. The rough calculation, for DC operations only and at Hyperscaler size is 0.5 employees per MW, this is a GW DC so the number of employees in the DC itself would be 200-500, depending on a variety of factors.
Like I mentioned before, this monstrosity, if built, would at peek capacity be using about a quarter of the energy of my whole country, this, however, wouldn't all come from our grid, a big part of this project is an LNG power plant + fields of solar, for which there is plenty of space around the area, these both come with it's own staff, so the upper estimate of 1.500 people working in and around this project doesn't seem crazy.
For the "it just makes shit up" people, yeah, sometimes it does, that's why no serious business or person is going to give it access to critical systems and allow it to delete shit (as recently publicized) or have it do free form data analysis that no one double checks for months at the time. + Show Spoiler +
As an example, when I try to get to the bottom of an issue, I often review logs, since there is a ton of data, I feed it in to the AI. Once it does a comparative analysis it will give me timestamps and I'll go and check if the lines are in the logs, anything else would be reckless.
I'm having a hard time figuring out if this is sarcasm or not? Seems like you actually believe it, but it also seems way too ridiculous/oblivious for anyone to seriously believe. Maybe I'm misunderstanding your qualifiers? An example that comes to mind: UnitedHealth uses faulty AI to deny elderly patients medically necessary coverage
The lawsuit, filed last Tuesday in federal court in Minnesota, claims UnitedHealth illegally denied "elderly patients care owed to them under Medicare Advantage Plans" by deploying an AI model known by the company to have a 90% error rate, overriding determinations made by the patients' physicians that the expenses were medically necessary.
"The elderly are prematurely kicked out of care facilities nationwide or forced to deplete family savings to continue receiving necessary medical care, all because [UnitedHealth's] AI model 'disagrees' with their real live doctors' determinations," according to the complaint. https://www.cbsnews.com/news/unitedhealth-lawsuit-ai-deny-claims-medicare-advantage-health-insurance-denials/I don't know if AI (with a known 90% error rate) deployed by a top 10 fortune 500 company making life and death decisions for people counts as a "serious business" or a "critical system" to you though. Simberto already replied to you, horrible companies using this to be able to shift responsibility for killing people doesn't mean that they are "serious people and organizations", I mean, they are, but they aren't using this in a genuine way where they expect high quality outputs, the goal is to utilize the low quality to shift responsibility. American capitalism and healthcare being broken does not invalidate that AI as a technology, when used by intelligent people and companies who are trying to deliver good products faster then they could before, AI is an absolute boon to productivity. There was the case of an AI bot inciting violence or suicide or something like that. When you get machines performing human tasks, corporations can shift responsibility for harm caused to actual people to the consumer who sees the content tailored to them. It's a premise that removes accountability altogether for corporations but not for users. If I build a robot that goes around shooting people, can I say it's the bots fault? This question is the main driving force (huhu pun) for the car industry and level 3+ autonomous driving. Much legal groundwork should be done already I imagine. What's the fun in being in a car if you don't drive it. My opinion at least. But yeah, it's an angle that drives progress in other areas at least. Seems like a lot of extra work for insurers. Maybe they needed that. One of their side income sources are drunk drivers causing accidents, now they can add machines to the mix. To a lot of people, cars are not for fun, but a tool to get to some place. And I'd much rather get to a place without having to concentrate on traffic all the time. Which is why i like taking trains, and like driving a lot less than riding a train. Would love to take trains but that is not possible for me. Everyday when I get into the office my heart rate is already through the roof because of some reckless drivers (and sometimes old / slow drivers, but that is a me problem ^^' ). A world where every car is driven by a computer under the same ruleset is a day that can't come soon enough. Traffic jams will be severly lessened. Accidents between cars will be at almost zero. Accidents with pedestrians and cyclist should be zero as well.
I think expectations for automatic driving is too high for it to be implementable. It being twice as good as the average human seems like an acceptable level to me. That means there will be hundreds or accidents per year still. It is then something that can be worked on by setting the cost of an accident high enough that it is worth investing in improving it over just paying the insurance premium.
|
Northern Ireland27269 Posts
On May 06 2026 05:12 Jankisa wrote:Show nested quote +On May 06 2026 00:14 Nebuchad wrote:On May 05 2026 22:08 Jankisa wrote:
For the "it just makes shit up" people, yeah, sometimes it does, that's why no serious business or person is going to give it access to critical systems and allow it to delete shit (as recently publicized) or have it do free form data analysis that no one double checks for months at the time.
As an example, when I try to get to the bottom of an issue, I often review logs, since there is a ton of data, I feed it in to the AI. Once it does a comparative analysis it will give me timestamps and I'll go and check if the lines are in the logs, anything else would be reckless. That's a hope that you can have at the start, but once people get used to it and it usually looks decent they'll just go with it. When Google came out we used to verify that we weren't getting nonsense results on the interwebs too, and now we have grandpas getting their news from Facebook. Yeah, but the people like you and me who grew up with internet search becoming what it is still understand that you need to double check shit if you are going to base decisions on it. Don't get me wrong, the whole hallucination issue is something that worries me, but, as I've been using, on daily basis AI since the first GPT until now, first mostly for private fucking around and then very much for professional tasks for the last 3 or so years, the hallucination rate is way down especially if you are using premium models. That is the case both for self hosted ones, ones accessed via API and talking to them via chat interfaces, the real issue I've been seeing more and more is that it "forgets" things from relatively close time frame of the conversations, to me, it seems like the memory context is one of the most expensive aspects of it to maintain and they've (in my case Anthropic) been skimping on it, a lot. For cheaper and models with less guardrails, I'm guessing that incentives to make shit up instead of doing the actual work took over, which, you know, is a very human thing to do. It's all very strange, same as with the responsibility. If my kid kills someone, who is to blame. Is it me for raising a monster. Is it his or hers mom for not instilling the correct values. Is it the system, the school, some random person traumatizing them at some point, and now, on top of all of that we also get to worry about them being radicalized by AI. In my experience, these are very capable systems and most fuckups happen because humans are trying to save money, at least recently, and maybe that makes the whole endeavor unprofitable, but I have a feeling that optimizations will come and even if we stay at this level of capabilities the technology itself is here to stay and will transform the world. If we get to AGI/ASI, well, then all bets are off and who the fuck knows.
Largely agreed, as ever really.
For me it’s pretty clearly transformative tech, but the question is what is being transformed and how?
Questions it feels too few people with the actual ability to shape adoption, or block/regulate seem to be grappling with.
Having mucked around intermittently myself, flaws that were pretty obvious 6 month ago that I’d have considered limitations for certain applications really aren’t now.
I may (or may not) depending on who’s listening have a lot of source code lying around from when I did my industrial placement on my second degree.
When l left it was something I just kept for education and to look at a whole code base whose end functionality I was pretty familiar with.
Being similar curious about people saying AI will replace at least some engineers I’ve asked AI to analyse or help break down the code base just out of curiosity in the past.
It’s really progressed a hell of a lot, early doors it wasn’t really writing either optimal stuff for idiosyncrasies of a particular language, in even a small module. Like, it worked, but it was missing some tricks, and I needed to correct it quite a bit to plug it in.
It was notably better in other languages that were either more widely used, or had a ton of documentation online.
It got a lot better there, but struggled with the wider context of the system. So your solo module might now be looking good, but you’d have issues with how it intersected with others
Now? You can basically pump the whole thing in, it knows how the various discrete modules all interact together and how changes in one affect the other. It doesn’t really break things and it can track all that kind of stuff and give feedback.
The improvement is pretty mental.
When I started my internship there was a pretty critical launch of our product incoming, so not a huge amount to trust the newbie with. I wrote some new tests for the feature expansion and as that didn’t take that long offered to just clean the whole codebase of our component and standardise it.
Was decent busywork, exposed me to how it all was written and did make it a bit cleaner and readable.
I tested Claude on the same task relatively recently and it can do it very well. And not just to like a standard or convention, you can make it take whatever idiosyncrasies of whoever’s writing whatever, and tweak the codebase into that particular style.
Its a pretty rapid improvement far as I can tell
|
On May 06 2026 22:58 Yurie wrote:Show nested quote +On May 06 2026 21:28 Harris1st wrote:On May 06 2026 21:21 Simberto wrote:On May 06 2026 21:15 Vivax wrote:On May 06 2026 20:47 Harris1st wrote:On May 06 2026 00:22 Vivax wrote:On May 06 2026 00:00 Jankisa wrote:On May 05 2026 23:01 GreenHorizons wrote:On May 05 2026 22:08 Jankisa wrote:+ Show Spoiler +I personally think that it's something of a bubble, just like the dot com was a bubble, but we still have Cisco as a huge company despite it being one that was riding the high crest of the bubble at the time.
The need for compute is not going to go away, the development work and inference will be needed for the foreseeable future, unless there is a cataclysmic event where insane amount of economic activity goes away, the compute that a DC like this will be churning out will find a buyer, at what prices, well, that is not for me to worry about, more for the investors.
Even if AI development stops right now, new way to apply the agents we currently have will keep popping up, as critical as I am of the people leading this industry on the US side, it's pretty obvious that this is not just hype, it's going to be the most important technology of the next decade and I'm yet to find a convincing argument against that.
The "these things have 50 people working once constructed" is a typical Reddit number being thrown around for a year now, it's silly. The rough calculation, for DC operations only and at Hyperscaler size is 0.5 employees per MW, this is a GW DC so the number of employees in the DC itself would be 200-500, depending on a variety of factors.
Like I mentioned before, this monstrosity, if built, would at peek capacity be using about a quarter of the energy of my whole country, this, however, wouldn't all come from our grid, a big part of this project is an LNG power plant + fields of solar, for which there is plenty of space around the area, these both come with it's own staff, so the upper estimate of 1.500 people working in and around this project doesn't seem crazy.
For the "it just makes shit up" people, yeah, sometimes it does, that's why no serious business or person is going to give it access to critical systems and allow it to delete shit (as recently publicized) or have it do free form data analysis that no one double checks for months at the time. + Show Spoiler +
As an example, when I try to get to the bottom of an issue, I often review logs, since there is a ton of data, I feed it in to the AI. Once it does a comparative analysis it will give me timestamps and I'll go and check if the lines are in the logs, anything else would be reckless.
I'm having a hard time figuring out if this is sarcasm or not? Seems like you actually believe it, but it also seems way too ridiculous/oblivious for anyone to seriously believe. Maybe I'm misunderstanding your qualifiers? An example that comes to mind: UnitedHealth uses faulty AI to deny elderly patients medically necessary coverage
The lawsuit, filed last Tuesday in federal court in Minnesota, claims UnitedHealth illegally denied "elderly patients care owed to them under Medicare Advantage Plans" by deploying an AI model known by the company to have a 90% error rate, overriding determinations made by the patients' physicians that the expenses were medically necessary.
"The elderly are prematurely kicked out of care facilities nationwide or forced to deplete family savings to continue receiving necessary medical care, all because [UnitedHealth's] AI model 'disagrees' with their real live doctors' determinations," according to the complaint. https://www.cbsnews.com/news/unitedhealth-lawsuit-ai-deny-claims-medicare-advantage-health-insurance-denials/I don't know if AI (with a known 90% error rate) deployed by a top 10 fortune 500 company making life and death decisions for people counts as a "serious business" or a "critical system" to you though. Simberto already replied to you, horrible companies using this to be able to shift responsibility for killing people doesn't mean that they are "serious people and organizations", I mean, they are, but they aren't using this in a genuine way where they expect high quality outputs, the goal is to utilize the low quality to shift responsibility. American capitalism and healthcare being broken does not invalidate that AI as a technology, when used by intelligent people and companies who are trying to deliver good products faster then they could before, AI is an absolute boon to productivity. There was the case of an AI bot inciting violence or suicide or something like that. When you get machines performing human tasks, corporations can shift responsibility for harm caused to actual people to the consumer who sees the content tailored to them. It's a premise that removes accountability altogether for corporations but not for users. If I build a robot that goes around shooting people, can I say it's the bots fault? This question is the main driving force (huhu pun) for the car industry and level 3+ autonomous driving. Much legal groundwork should be done already I imagine. What's the fun in being in a car if you don't drive it. My opinion at least. But yeah, it's an angle that drives progress in other areas at least. Seems like a lot of extra work for insurers. Maybe they needed that. One of their side income sources are drunk drivers causing accidents, now they can add machines to the mix. To a lot of people, cars are not for fun, but a tool to get to some place. And I'd much rather get to a place without having to concentrate on traffic all the time. Which is why i like taking trains, and like driving a lot less than riding a train. Would love to take trains but that is not possible for me. Everyday when I get into the office my heart rate is already through the roof because of some reckless drivers (and sometimes old / slow drivers, but that is a me problem ^^' ). A world where every car is driven by a computer under the same ruleset is a day that can't come soon enough. Traffic jams will be severly lessened. Accidents between cars will be at almost zero. Accidents with pedestrians and cyclist should be zero as well. I think expectations for automatic driving is too high for it to be implementable. It being twice as good as the average human seems like an acceptable level to me. That means there will be hundreds or accidents per year still. It is then something that can be worked on by setting the cost of an accident high enough that it is worth investing in improving it over just paying the insurance premium.
I think the scenario where accidents are basically eliminated relies more on the ability of the cars/AIs to be meshed and work together.
That would, of course, require 90 % + of the cars to be driven by them, because throwing unpredictable humans into the mix will inevitably cause accidents, but in a world where the vast majority of cars are autonomous and the small numbers that aren't are people who are (most likely) professionals I think an order of magnitude reduction in accidents is expected.
Getting there is very much a different story, and honestly my reservation there would be, just like with AI, who controls this, because as soon as they reach a certain threshold and start chasing this "ideal world with no accidents" the money interests will absolutely work on outlawing human driving and when that happens squeezing the user base for every penny they have, which would at that point be everyone.
|
On May 06 2026 21:21 Simberto wrote:Show nested quote +On May 06 2026 21:15 Vivax wrote:On May 06 2026 20:47 Harris1st wrote:On May 06 2026 00:22 Vivax wrote:On May 06 2026 00:00 Jankisa wrote:On May 05 2026 23:01 GreenHorizons wrote:On May 05 2026 22:08 Jankisa wrote:+ Show Spoiler +I personally think that it's something of a bubble, just like the dot com was a bubble, but we still have Cisco as a huge company despite it being one that was riding the high crest of the bubble at the time.
The need for compute is not going to go away, the development work and inference will be needed for the foreseeable future, unless there is a cataclysmic event where insane amount of economic activity goes away, the compute that a DC like this will be churning out will find a buyer, at what prices, well, that is not for me to worry about, more for the investors.
Even if AI development stops right now, new way to apply the agents we currently have will keep popping up, as critical as I am of the people leading this industry on the US side, it's pretty obvious that this is not just hype, it's going to be the most important technology of the next decade and I'm yet to find a convincing argument against that.
The "these things have 50 people working once constructed" is a typical Reddit number being thrown around for a year now, it's silly. The rough calculation, for DC operations only and at Hyperscaler size is 0.5 employees per MW, this is a GW DC so the number of employees in the DC itself would be 200-500, depending on a variety of factors.
Like I mentioned before, this monstrosity, if built, would at peek capacity be using about a quarter of the energy of my whole country, this, however, wouldn't all come from our grid, a big part of this project is an LNG power plant + fields of solar, for which there is plenty of space around the area, these both come with it's own staff, so the upper estimate of 1.500 people working in and around this project doesn't seem crazy.
For the "it just makes shit up" people, yeah, sometimes it does, that's why no serious business or person is going to give it access to critical systems and allow it to delete shit (as recently publicized) or have it do free form data analysis that no one double checks for months at the time. + Show Spoiler +
As an example, when I try to get to the bottom of an issue, I often review logs, since there is a ton of data, I feed it in to the AI. Once it does a comparative analysis it will give me timestamps and I'll go and check if the lines are in the logs, anything else would be reckless.
I'm having a hard time figuring out if this is sarcasm or not? Seems like you actually believe it, but it also seems way too ridiculous/oblivious for anyone to seriously believe. Maybe I'm misunderstanding your qualifiers? An example that comes to mind: UnitedHealth uses faulty AI to deny elderly patients medically necessary coverage
The lawsuit, filed last Tuesday in federal court in Minnesota, claims UnitedHealth illegally denied "elderly patients care owed to them under Medicare Advantage Plans" by deploying an AI model known by the company to have a 90% error rate, overriding determinations made by the patients' physicians that the expenses were medically necessary.
"The elderly are prematurely kicked out of care facilities nationwide or forced to deplete family savings to continue receiving necessary medical care, all because [UnitedHealth's] AI model 'disagrees' with their real live doctors' determinations," according to the complaint. https://www.cbsnews.com/news/unitedhealth-lawsuit-ai-deny-claims-medicare-advantage-health-insurance-denials/I don't know if AI (with a known 90% error rate) deployed by a top 10 fortune 500 company making life and death decisions for people counts as a "serious business" or a "critical system" to you though. Simberto already replied to you, horrible companies using this to be able to shift responsibility for killing people doesn't mean that they are "serious people and organizations", I mean, they are, but they aren't using this in a genuine way where they expect high quality outputs, the goal is to utilize the low quality to shift responsibility. American capitalism and healthcare being broken does not invalidate that AI as a technology, when used by intelligent people and companies who are trying to deliver good products faster then they could before, AI is an absolute boon to productivity. There was the case of an AI bot inciting violence or suicide or something like that. When you get machines performing human tasks, corporations can shift responsibility for harm caused to actual people to the consumer who sees the content tailored to them. It's a premise that removes accountability altogether for corporations but not for users. If I build a robot that goes around shooting people, can I say it's the bots fault? This question is the main driving force (huhu pun) for the car industry and level 3+ autonomous driving. Much legal groundwork should be done already I imagine. What's the fun in being in a car if you don't drive it. My opinion at least. But yeah, it's an angle that drives progress in other areas at least. Seems like a lot of extra work for insurers. Maybe they needed that. One of their side income sources are drunk drivers causing accidents, now they can add machines to the mix. To a lot of people, cars are not for fun, but a tool to get to some place. And I'd much rather get to a place without having to concentrate on traffic all the time. Which is why i like taking trains, and like driving a lot less than riding a train.
To be fair I hate driving, but when it comes to autonomous driving... well if I am about to die in car crash, I would rather have myself to blame, rather than software glitch. Later seems so... avoidable.
|
I don't know, to me, the method of my dying would very much be irrelevant since I'd be dead.
I'm a big believer in statistics and as a driver with 20+ years of experience very little faith in my fellow human drivers, given how amazing just the "basic" tools that come with fairly budget cars (I drive a Mazda 2) like the adaptive cruise control with lane tracking is, my general inclination is to trust the technology, when the above mentioned network effects kick in, I'm fairly certain I'd have no issue giving up driving.
Of course, I'll never give up my tractor so I'll always have that, ain't no clanker gonna spray my vineyards.
|
The rayzda mentality is very common, which is why I think it’s going to be a long time until we get full autonomous cars as the common. It seems like being significantly better than people is not enough. It needs to be perfect.
I look at it a different way. My big concern driving is not me, it’s the others on a road. And I trust computers more than texting, drunk, distracted by kids, dumb, sleepy people.
And how nice would it be to be able to do something other than drive in the car! And the efficiency in cities is going to change traffic so dramatically. Sooner the better, as long as properly tested.
|
On May 08 2026 01:07 Jankisa wrote:
I don't know, to me, the method of my dying would very much be irrelevant since I'd be dead.
Thats one way of looking at it, as for me I would rather be in charge of my fate myself as much as possible.
On May 08 2026 01:07 Jankisa wrote:
I don't know, to me, the method of my dying would very much be irrelevant since I'd be dead.
I'm a big believer in statistics and as a driver with 20+ years of experience very little faith in my fellow human drivers, given how amazing just the "basic" tools that come with fairly budget cars (I drive a Mazda 2) like the adaptive cruise control with lane tracking is, my general inclination is to trust the technology, when the above mentioned network effects kick in, I'm fairly certain I'd have no issue giving up driving.
Of course, I'll never give up my tractor so I'll always have that, ain't no clanker gonna spray my vineyards.
On May 08 2026 01:48 Billyboy wrote:
The rayzda mentality is very common, which is why I think it’s going to be a long time until we get full autonomous cars as the common. It seems like being significantly better than people is not enough. It needs to be perfect.
I look at it a different way. My big concern driving is not me, it’s the others on a road. And I trust computers more than texting, drunk, distracted by kids, dumb, sleepy people.
And how nice would it be to be able to do something other than drive in the car! And the efficiency in cities is going to change traffic so dramatically. Sooner the better, as long as properly tested.
Precisely, do you think any autonomous car will be able to account for idiocy which is happening on the roads?
|
|
|
|
|
|
EPT
