Artificial Intelligence In Our Lives

[eyoungren]

Some food for thought about these AI data centers

Given that Caltrans can erect freeway overpasses in a single night and that they offer a commercial service, I have no doubt it could be built in two months.

AZDOT on the other hand…

 

[Luftkopf]

I’ll question just how good AI is as an economical therapist as it currently exists. I have limited exposure to AI chat, but it seems, it’s only as good as its programming, what I’ve seen tends to try to stay neutral, and complement your prejudices/preferences. I could see if AI was programmed to act like a therapist might, with the same knowledge at it’s disposal, and an understanding of human psychology, and the means to apply goid techniques to help individuals cope, it could be helpful. But aren’t these sysytem primarily designed to just regurgitate information. How much, thinking, logic, do they actually apply, it’s just a program looking for key words to react to?

They’re not even designed to regurgitate information, just string together plausible sentences!

 

[profcutter]

It’s all about short term gains, never long term sustainability. If you can lay off a massive number of workers, your stock price goes up, and you get a fat bonus. The Japanese auto makers realized this in the 1980s, when GM and Ford were laying off workers and rushing to build plants outside of the US, Toyota, Honda, and Nissan moved their production to the US. Why? Because if you have no one paid a solid auto worker salary, who’s going to buy your cars? Also, workers at those plants would be assumed to have some loyalty to the company they worked for. With AI, which as we know is a misnomer, jobs are getting replaced in almost every sector, what academics call the “ensh!tification” of the economy. Cost cutting, jobs reduction, and elimination of social programs and the welfare safety net. Mediocre products, higher prices. The idea that we’re headed for another Industrial Revolution is pretty accurate, and not in a beneficial way.

 

[splifingate]

But, but, but ... If someone has mental health issues, how can such a someone be expected to set the correct, strong boundaries for usage of an AI counselor? That is illogical. Ergo, AI should never be used as a therapist.

Love this. IAWTP!

I know a few people who have come to heavily rely upon their personal "phone agents" for comfort and counsel. It disturbs me, but it also makes me uncomfortable to potentially become an agent of change, especially since I would probably be accepting the role of the one who replaces the aigent.

 

[Howard2k]

They’re not even designed to regurgitate information, just string together plausible sentences!

No, that's not really true these days.

LLMs are designed to effectively do weighted word prediction, yes, and stringing together plausible sentences was an early metric, for sure. This is research going back decades. This is done based on both syntax and knowledge.

But the real benefits of AI are not this alone, the real benefits of AI are when this ability to "understand" a statement or request is leveraged in order to carry out other specific commands.

To highlight a really simple example - an LLM can respond to "tell me how to cook eggs" because that's easy and trivial information. But where an LLM provides additional value is when it can contextualize a prompt to do other actions beyond a simple description. In the case of an LLM running with a search engine it might be able to determine that "cook eggs" might mean frying, boiling, poaching, scrambling etc. While "cook eggs in water" might mean boiling or poaching, but probably not frying. By trying to understand (or more correctly, simulate understanding of) the query it can then leverage other tools, including research tools. And when the LLM is searching the web for information on cooking eggs, it is also able to analyze the information that it found to see if it's relevant. LLMs/AI today are fundamentally not just stringing together plausible sentences, and that hasn't been for quite a while. It's the ability to use an LLM as a part of a toolset, not purely as a hammer, where the value and benefits lie.

LLMs do regurgitate information, and when it comes to trivia it can be poor. Right tool for the right job still applies though. Don't expect your downloaded 8GB model to match a cloud based AI that is also scraping the Internet in real time.

We've already seen material medical breakthroughs come from AI. It could be argued that every step forwards in technology is a step towards techno-geddon. I get it. But do we just stop trying to advance?

 

[johnsawyercjs]

I’d like to think people can’t be that dumb. Let’s say you execute some sort of ‘Kingsman’ style plan and reduce the global population of plebs like us or replace our buying power with AI.

You know what you’ve just reduced in size and scale? Your customer base who provide your income.

The AI bros' plan seems to be to no longer need a customer base, but instead to create robots that make everything the bros need. Of course nearly everyone else will have an issue with that, and that may be why these bros are building bunkers.

 

[Huntn]

I agree with Seth.
AI getting deployed into creative fields is one of the the most depressing usages.

View attachment 2630510

This is automation. Do people actually think that automation makes them better writers? No, if what AI produces is better then how does that make them better, it’s just going to replace them. 😉 As far as I know there has been nothing created by AI that is an original thought, that surpasses the ability of the best humans, it’s just regurgitating what it has been Fed, with the one advantage, a computer-like memory.

 

[johnsawyercjs]

Humans evolved to take on complex tasks, including developing the skills to understand the universe, requiring us to develop the scientific method, math, physics, engineering, etc., the methods and principles for which we’ve been keeping active in our heads rather than farming it out, since until now there was nothing to farm it out to.

Now a world is being created in which many or most knowledge workers will be sidestepped by AI. Fewer and fewer people will need to keep knowledge in their heads that was carefully gathered and hard-won over millennia. When many or most complex knowledge tasks are handed off to AI, what happens to the human need to figure out these things? I don’t think it’ll disappear, since it’s a biological imperative, but it may be that the opportunities for expressing it will shrink so much that it may be to our detriment. Fewer Isaac Newtons, Albert Einsteins, Jane Goodalls...

 

[Howard2k]

Now a world is being created in which many or most knowledge workers will be sidestepped by AI. Fewer and fewer people will need to keep knowledge in their heads that was carefully gathered and hard-won over millennia. When many or most complex knowledge tasks are handed off to AI, what happens to the human need to figure out these things? I don’t think it’ll disappear, since it’s a biological imperative, but it may be that the opportunities for expressing it will shrink so much that it may be to our detriment. Fewer Isaac Newtons, Albert Einsteins, Jane Goodalls...

Or maybe the opposite. Maybe more of these people would have had more free time on their hands to further develop their knowledge, skills, and abilities.

 

[Huntn]

No, that's not really true these days.

LLMs are designed to effectively do weighted word prediction, yes, and stringing together plausible sentences was an early metric, for sure. This is research going back decades. This is done based on both syntax and knowledge.

But the real benefits of AI are not this alone, the real benefits of AI are when this ability to "understand" a statement or request is leveraged in order to carry out other specific commands.

To highlight a really simple example - an LLM can respond to "tell me how to cook eggs" because that's easy and trivial information. But where an LLM provides additional value is when it can contextualize a prompt to do other actions beyond a simple description. In the case of an LLM running with a search engine it might be able to determine that "cook eggs" might mean frying, boiling, poaching, scrambling etc. While "cook eggs in water" might mean boiling or poaching, but probably not frying. By trying to understand (or more correctly, simulate understanding of) the query it can then leverage other tools, including research tools. And when the LLM is searching the web for information on cooking eggs, it is also able to analyze the information that it found to see if it's relevant. LLMs/AI today are fundamentally not just stringing together plausible sentences, and that hasn't been for quite a while. It's the ability to use an LLM as a part of a toolset, not purely as a hammer, where the value and benefits lie.

LLMs do regurgitate information, and when it comes to trivia it can be poor. Right tool for the right job still applies though. Don't expect your downloaded 8GB model to match a cloud based AI that is also scraping the Internet in real time.

We've already seen material medical breakthroughs come from AI. It could be argued that every step forwards in technology is a step towards techno-geddon. I get it. But do we just stop trying to advance?

We have no choice but to advance. The challenge is how to not destroy a functional society when the motive to advance is solely/mostly based on self, greed, profits, at others expenses.
If asked how can this be at others expense? Easy, the system is not designed to keep productive citizens who have been disenfranchised (by technology or exported jobs) a realistic avenue for support and being retrained. You’re on your own kid, because Corpos want no part of this, they used to train new hires, now if I understand it correctly, they want to hire only experienced workers, so it can be inferred training detracts from profit margins. They just want to improve their bottom line, and if an AI agent can do a half assed job for free, goodbye to the expert human doing the job.
Government is the only entity that can referee, control this. It’s not part of the average corporation’s business model.

 

[Huntn]

Humans evolved to take on complex tasks, including developing the skills to understand the universe, requiring us to develop the scientific method, math, physics, engineering, etc., the methods and principles for which we’ve been keeping active in our heads rather than farming it out, since until now there was nothing to farm it out to.

Now a world is being created in which many or most knowledge workers will be sidestepped by AI. Fewer and fewer people will need to keep knowledge in their heads that was carefully gathered and hard-won over millennia. When many or most complex knowledge tasks are handed off to AI, what happens to the human need to figure out these things? I don’t think it’ll disappear, since it’s a biological imperative, but it may be that the opportunities for expressing it will shrink so much that it may be to our detriment. Fewer Isaac Newtons, Albert Einsteins, Jane Goodalls...

This has been explored frequently in science fiction writing. When civilization forget how to do everything because it’s being done for them, by the Head Computer and then something breaks, or when Skynet decides human are the problem. 🙃

Something real, all the trillions of manufacturing $ shipped to Asia, those countries now know how to manufacture thanks to our capitalist overlords. Besides a great way to boost profits, disenfranchise citizen workers, an excellent way to export pollution, not fix our global warning problem.

Some engineers here, probably still know how things are done, from an oversight and decision making, but let’s say WWIII breaks out, we have no one here ready to hop on domestic manufacturing.

 

[edubfromktown]

Qwen is surprisingly capable! I tried the Qwen 3.5 35B model locally. I think everyone wanting to use LLMs extensively should try running one locally first (and before you say they need good hardware, they don't – even a PowerBook G4 can run a LLM), play around with the settings like temperature and top k, and get a basic understanding of how they work first.

I upgraded to qwen3-14b over the weekend and amazingly it hovers just under memory pressure going yellow (28 of 32 GB RAM utilization).

I added the /no-think LM Studio flag to speed it up close to what I had experienced with qwen2.

 

[Luftkopf]

No, that's not really true these days.

LLMs are designed to effectively do weighted word prediction, yes, and stringing together plausible sentences was an early metric, for sure. This is research going back decades. This is done based on both syntax and knowledge.

But the real benefits of AI are not this alone, the real benefits of AI are when this ability to "understand" a statement or request is leveraged in order to carry out other specific commands.

To highlight a really simple example - an LLM can respond to "tell me how to cook eggs" because that's easy and trivial information. But where an LLM provides additional value is when it can contextualize a prompt to do other actions beyond a simple description. In the case of an LLM running with a search engine it might be able to determine that "cook eggs" might mean frying, boiling, poaching, scrambling etc. While "cook eggs in water" might mean boiling or poaching, but probably not frying. By trying to understand (or more correctly, simulate understanding of) the query it can then leverage other tools, including research tools. And when the LLM is searching the web for information on cooking eggs, it is also able to analyze the information that it found to see if it's relevant. LLMs/AI today are fundamentally not just stringing together plausible sentences, and that hasn't been for quite a while. It's the ability to use an LLM as a part of a toolset, not purely as a hammer, where the value and benefits lie.

LLMs do regurgitate information, and when it comes to trivia it can be poor. Right tool for the right job still applies though. Don't expect your downloaded 8GB model to match a cloud based AI that is also scraping the Internet in real time.

We've already seen material medical breakthroughs come from AI. It could be argued that every step forwards in technology is a step towards techno-geddon. I get it. But do we just stop trying to advance?

Maybe I was over simplistic in my response, but LLMs don't "know" or "understand" anything, and are frequently wrong. I do not believe they have no use, or that they should not be used, they can be very useful tools in the right context. Sadly the context for most genAI deployment is not that, and is more for the purposes of making people redundant.

Or maybe the opposite. Maybe more of these people would have had more free time on their hands to further develop their knowledge, skills, and abilities.

Maybe in a functional social democracy that believes in supporting its population, but not in any neoliberal country. those people are now busy working three jobs to try to make rent.

 

[Harddrive]

I wonder when the bubble will burst the AI data centers are using up mass quantities of memory and storage driving up consumers costs for tech so at some point they will have a product that can not be used by all do to cost .

 

[turbineseaplane]

I am of the opinion that these massive data centers should all be providing their own renewable power and water sourcing.

It’s ridiculous that a lot of them are getting built where the sunshine is unbelievably plentiful yet they aren’t building huge solar farms to support their own needs.

Worse yet, we have some of them installing un permitted gas turbine power on site.

 

[turbineseaplane]

Man oh man, this is exhausting to watch and live through.

Chatbots at the drive-thru are just the beginning

May AI take your order?

www.theverge.com

 

[Huntn]

I wonder when the bubble will burst the AI data centers are using up mass quantities of memory and storage driving up consumers costs for tech so at some point they will have a product that can not be used by all do to cost .

And there is the environmental cost so multiple competing corporations can increase their mountains of gold…

 

[skodises]

This buildout process does speak volumes about how loudly money talks, doesn’t it? It is making an absolute mockery of the entire concept of public/local government oversight of development. Permitting is definitely taking a distant back seat to devil-take-the-hindmost profit-seeking.

There should be some careful forensic accounting work done in many cases here, once the dust settles, because it seems highly likely (at least to this reporter) that many formerly underpaid public servants might be experiencing unprecedented windfalls. As they say: not a good look…

 

[Huntn]

This buildout process does speak volumes about how loudly money talks, doesn’t it? It is making an absolute mockery of the entire concept of public/local government oversight of development. Permitting is definitely taking a distant back seat to devil-take-the-hindmost profit-seeking.

There should be some careful forensic accounting work done in many cases here, once the dust settles, because it seems highly likely (at least to this reporter) that many formerly underpaid public servants might be experiencing unprecedented windfalls. As they say: not a good look…

We have been drinking the Koolaid about the benefits of commerce, and technological advances if the social elements, ramifications are ignored. I’ve got this new software that will replace 50% of jobs in a matter of years and I predict my competitors will use this to replace people. How do I remain competitive in a capitalist system?
Answer is, you don’t unless you duplicate, unless the system is changed.

And about those people, their lost purchasing power, their nightmare, effect on social stability, and the health of commerce in general? There will be some version of: care, ignore or **** them. The fallacy is that we as a group can disenfranchise huge swaths of the populace and think the good times will still keep roll along, or at least I’ll still have my pile of (now worthless) currency. 🤔

 

[ejb190]

I have a friend who works in construction in a mid sized eastern German town. Not exactly the bleeding edge of the tech revolution. They were able to reduce an office clerk to a half-time position through the use of AI (writing quotes, accounting/bookkeeping, etc.).

In my work I have contact with many people looking for work. Among them are many whose departments have been decimated, at least under the pretext of implementing AI.

So yes, it's definitely already replacing jobs, just that we are not sharing the profits with everyone so we can all work less, as people had hoped in the past century. (In 1930, John Maynard and Keynes wrote his famous essay "Economic Possibilities for our Grandchilden", predicting that by 2030 technological progress would reduce the working week to around 15 hours.)

My job involves walking fields and hunting for insect and disease pressure. My driving time is over 15 hours a week! Drones are coming, but I see them as a supplement, not a replacement to what we do. You can't spot greenhouse thrips at 15 feet and 20 mph. Most of my AI work is writing form letters and processing data - though I've found it's hit or miss for that, too.

 

[Huntn]

Really I think the issue at the moment is the fact that AI appears to be a commercial exercise, rather than an exercise in improving peoples lives.

Something like that saying online, I want an AI to do my cleaning and house work so I can do art. Or something like that.

When it comes to competitiveness, these corporations are like lemmings, if one does it, they basically all do. And because it’s all about being competitive and making $$$, just about every employee is expendable “if that’s what it takes to remain profitable.” This commerce thing will very soon have to be a government thing , that we the people demand, though the wise or carpet bagging representatives we choose, if want to keep from unraveling.

 

[profcutter]

Yeah the lack of regulation is a serious problem, and quite curious. I know it’s takes a while for the gears of government to respond to new crises, but this is absurd. It’s a case of corporate government capture where AI centers that violate every local zoning rule are approved in backdoor deals, and as pointed out, some local legislators are making a windfall for the short term as they gut environmental protections and disenfranchise their local communities.

 

[Huntn]

My job involves walking fields and hunting for insect and disease pressure. My driving time is over 15 hours a week! Drones are coming, but I see them as a supplement, not a replacement to what we do. You can't spot greenhouse thrips at 15 feet and 20 mph. Most of my AI work is writing form letters and processing data - though I've found it's hit or miss for that, too.

All those Amazon trucks scurring about driven by humans…not saying they're the best jobs, but they are jobs. Amazon want to replace them with automated drones.

 

[ReliableSource]

Have a lot of you reading this thread read Manna by Marshall Brain? Seems apropos

 

[MicroLab]

AI should be to improve our lives, not replace our livelihoods.

Fans Blame AI As Disney Decimates Marvel Studios Art Teams

Disney's latest wave of firings reportedly hit Marvel Studios' visual departments the hardest, resulting in speculation that AI is to blame

kotaku.com

View attachment 2628904

What’s so funny about this quote is that we’ve had machines that do our laundry and dishes for many decades now.