• S&T Moderators: streaM Freak

Is A.I. waiting to take over ?

For sure, I can say with a heavy heart the LLM's data center puts the world in a declining rate because of the pollution acceleration it has over the worldwide sea's. In a indirectly way, this year in Romania, we've seen a drought like never had before, we had phuu uh, the Danube hitting a crosspoint between ok and not ok to not ok at all and while yes, you could techincally but not practically mention the AI needs of the water that's found in alps, i think it has some influence. But in a way, because of this tremendous amount of fund stealing from the party leaders and all that's been roughly been going on for about 37 years now, we had some minuses on some critical infrastructure point where this drought that's been seen in France too, affected us 5 time less than France, even if we import some electricity from it and Ukraine. Danune was supposed to be fixed in 3 vital points, which for now I can't recall their names but we've been worse, and we're yet to hear our mysterious prime minister highlighting any contrast off, Nicusor Dan says from 1 next month after this one, so that'd be 1 September, we're gonna see some "Red Plague aka Ciuma Rosie, PSD" figure here.

Again, moral? AI I think... from my point does have some influence over this paradoxal drought, problem is that these type of stuff aren't said public because of the alarming rate i'd go by.
 
Depends what you mean by "taking over" there very slow moving and as far as I know only inhabit parts of South America ,spending 95 percent of their lives in trees would make it seem unlikely they will be taking over anything that would directly impact us humans
 
taking over and propping up the stock market up (in a highly questionable way) -- definitely

here is an A.I overview which may have well came out of an economic professors mouth (insert proper sarcasm) Artificial intelligence is heavily propping up major stock indices like the S&P 500, accounting for roughly half of market gains and masking sluggish performance in the broader consumer economy. Driven by massive infrastructure spending, top tech giants are committing over $1 trillion to data centers and chips, creating a powerful wealth effect that keeps macroeconomic growth numbers high. [1, 2, 3, 4, 5]
 
There are two different things here: there's the actual technology of LLMs, which is a huge breakthrough and an extraordinarily useful technology in many ways and regards, and then there are the shitbag Big Tech companies run by corporate billionaire tech bros who think of personal ethics as a minor inconvenience best to try to ignore. It sounds like this is the main argument people are getting upset about, and they either are unwilling to admit how impressive the technology really is / they are willfully ignorant to it on purpose and try to downplay it because of their beef elsewhere in this argument… OR they just don't understand it.

Yo I didnt say technology makes you weaker that was someone else.
True. I apologize for the misattribution.

I am only further convinced this is a shitty decision pushed ruling class.
I mean, it's free market crapitalism. You don't have to use AI if you don't want to. Sure, it's popping up here and there, but it's still pretty easy to opt out of and not use. You're so upset over this and so quick to try to condemn anyone with any viewpoint that challenges yours whatsoever. You write that person off as stupid or ill-willed and sit there 100% convinced of your self-righteous anger. Just try to see it from a different perspective for a little while to understand the other side of the argument. It won't hurt you to try.

And isn't like I"m unaffected by AI's negative points. RAM prices really are sky high, Apple just jacked all their prices, and I see so much bullshit AI slop these days it makes me want to vomit sometimes. Also, it's encroached on some of the work I do as an artist, graphic designer, and front-end code developer. My approach has been to roll with the punches and stay agile. But what do you expect to happen as a result of your complaining about AI LLMs? Do you really think anyone is going to pass legislation that will just stop AI altogether? Do you really think any of these tech giants are going to stop pursuing all matters AI right in the middle of all the action? Are CEOs at publicly traded companies going to acquire a sudden new conscience about this and do what is against shareholders' best interests and just fold AI operations, take losses for what's been invested, get punished by their competition for not adopting new technology, lose their job and forfeit their career or change career paths due to a pang of guilt over something where the ethics are not crystal clear?

So if you're like me and don't see that happening, what's the point of being so anti-AI? Do you have a solution to this? I mean, a real working solution in which everyone will be pleased? If not, you're just bitching about shit that, for all intents and purposes, no one can change. There's no point. You're just getting yourself upset.

I am reminded of when I was young and the Internet was first becoming a thing. There were people who held out, who said that the Internet was just a fad soon to go. They swore that it was an insanely BAD IDEA to order something "online". They were convinced it was all just scams and psychopaths on the Internet. It was thought back then that if you "met" someone online, that person was guaranteed to be a psychotic serial killer. At one point, some people were opposed to "talkies" in Hollywood, and went around telling everyone that audible dialogue was going to ruin movie-making. People thought automobiles were too dangerous and disruptive to society and should be banned. Imagine that. Automated textile machinery was famously opposed by the Luddites, and there were scores of people opposed to the printing press when it revolutionized the dissemination of education and information.

But if you want to live a simpler life, no one is stopping you. Another approach would be to understand that there is a balance between embracing new technologies and preserving out humanity. The two are not mutually exclusive, as history as shown and time will reveal.

Let's look at some examples of how AI benefits large portions of society…
  • Healthcare ☞ AI helps analyze scans, identify drug candidates, summarize records, monitor patients, and extend basic medical guidance to underserved areas.
  • Science ☞ AI processes huge datasets, predicts protein structures & materials, simulates complex systems, and helps researchers find promising hypotheses faster.
  • Accessibility ☞ AI provides captions, text-to-speech, image descriptions, translation, simplified language, and assistive communication.
  • Education ☞ It offers individualized explanations, tutoring, practice, feedback, and translation, although it cannot replace good teachers or reliable sources, but this limitation is wide-known.
  • Productivity ☞ AI handles repetitive paperwork, scheduling, transcription, data entry, preliminary research, and routine coding, leaving people more time for judgment and creative work.
  • Public safety and infrastructure ☞ AI can improve weather forecasting, disaster response, traffic management, energy efficiency, equipment maintenance, and detection of fraud or cyberattacks.
  • Economic opportunity ☞ It helps small organizations and independent workers perform tasks previously affordable mainly to large institutions.
  • Communication ☞ Again: the language translation alone is incredible. Plus it helps people organize & express ideas, reducing or removing barriers, and filling gaps in technical knowledge and literacy.
It's fine if you don't like Big Tech companies. I too have some grievances with them, but you're getting angry at people like me for suggesting that the technology itself is useful. I know it's useful because I've used to it in ways that are incredibly helpful, time-saving, and yes, game-changing. Don't get mad at me about it. Get angry at the right people for their business model, not new technologies or the people who use them.

A.I is a piece of baggage noone asked to carry. happy
Hey, an actual metaphor. Look at you! Lol, sorry I didn't mean to get caught up on strict definitions and shit. My bad.

Idk, I don't have a data center plopped down right next to where I live, and I imagine that would irkTF outta me. Most likely, I think I would target that company's data center practices and lean on local politicians to pass legislation preventing these data centers from operating how they operate, demanding reform and steps be taken to protect local residents more than it has. I wouldn't target the technology behind what they're doing. If it were not AI, and it was something else data driven (which I almost guarantee you it would be), the same problem would exist. So in my humble opinion, targeting the root of the problem seems more appropriate.

If "no one wants AI" is true… "no one asked to carry" the "AI baggage"… then that means there would be no demand for the it, right? And we live in a supply and demand society. Something doesn't add up, and my guess is that you're wrong. If that supply exists, it's safe to assume that there is a demand for it. I mean, you know how economics works, don't you? I'm just stating the obvious, right? I mean this in all honesty, and I swear I'm not trying to be a jerk. Try not to kill the messenger.

Yes a humanity multiplier and human nature is generally good and generous right?
Not sure if you're being sarcastic here or what… Generally I do believe in the goodness of most people. I am also aware that people are capable of mind-blowing cruelty and, for lack of a better word, evil. Idk what point you're attempting to make, bc again, whether you're being sarcastic is lost on me at the moment.

Definitely the people at the top designing it care about what is best for most of us.....
Sarcasm? The executives of large, publicly traded companies have an ethical and legal obligation to make decisions that are in the best interest of shareholders in an attempt increase owner's equity through revenue, profits, and handling of company stock. There's also a difference between ethics and morals, but that's kind of a different subject.

Anything that errs in the direction of giving an incorrect answer
Nah, it's pretty sharp. I don't usually get wrong answers, but we should be fact-checking things ourselves more frequently anyways. Only a fool lets someone else do all their homework for them. Doubly true considering how AI is excellent at teaching things. It's odd to me how people react to this phenomenon. Just my estimate, but let's say a human is right 90% of the time and an AI is right 99% of the time. People will be impressed by that human, like someone who wins at Jeopardy for five days straight, for example. No one holds it against that person when they get an answer or two wrong; I mean, they lose money from their winnings on the game, of course, but no one is like, "Oh fuck this guy. What a stupid bastard. Fuckin' degenrate nobody asked for, taking up all this prime time television air! Fuckin' Big Trivia, bro, amirite?" But despite giving you 99 beautifully explained answers, that one or two questions it got wrong a year and a half ago proves to you that the technology should be abandoned at once…

over admitting it's unsure (or incapable of discerning whether tbings are true) makes its value incredibly low, and its tendency to create people overconfident in flawed ideas incredibly high
This is not unlike humans. I encounter people all the time with bad intel, spreading rumors or false shit, especially concerning drugs, pharmacology, drug myths, etc. Also, how many times is the news wrong, or the weather? Think about it. And I know some people, smug and self-satisfied, will answer that with some shit like: "I don't watch the news, the media is XYZ and the weather isn't accurate!", but all that means is they get their info secondhand, and they occasionally are dressed for yesterday's weather. Meanwhile, anyone who watches, will occasionally have an umbrella or sweater they did not need.
 
Yes sarcasm on both basis. I find human nature to be performative, self serving, in general path of least resistance. Morals are very flexible based on perceived legitimacy of authority (Milgram studies) --- at the CEO of A.I level it is statistically true these people are more likely to be sociopaths --- the 1% looking out for 99 is bad news everytime.

On the second case -- yea it is wild how they talk to there shareholders compared to the public, especially regarding surveillence / pricing etc --- I think they are some of the most disconnected evil bunker building fucks on earth and would have no problem fucking the middle class out of existence. In fact I believe it to be a priority and A.I to be a tool.

something is messed up supply/demand wise -- hence it propping the stock market up. I can opt out of using it but not paying for it - I disagree with that. I can opt out of oil and opt out of paying for it. That is a fair difference?

I am not upset nor surprised that such a thing is happening again; I just recognize a big tech fuck over when i see one, I think.

I am happy to admit I could be wrong here; time will tell.
 
I don’t know, but I really do feel like judgment day might be coming soon. Honestly, I think some illness will probably get me long before all this AI stuff does. may just drop dead someday, get hit by a truck or something,
 
Like I say ,I think there may be some hysteria over this issue ,we swim in the sea with fish that can bite us in half and there are plenty other mammals even domesticated ones that frequently cause fatalities
A little rational perspective is in need here I believe
 
@unodelacosa - you know, using AI to list all of the wonderful benefits of AI is akin to asking a murderer to list the benefits of killing people.

I may be one of the few people to see both the utility as well as the risks of AI.

Let's start with pointing out that AI is a misnomer. No intelligence is involved. As far as I know every chatbot available to the public just scraped the internet and scanned books for it's dataset. So one issue right away is that an AI is only as good as it's training-set and since digital publishing is essentially free and also seems to avoid legal liability, it's not a great training-set thus problematic.

The 'Claude crash' shows how over-confidence and the crippling cost of developing a general purpose chatbot will tend to make the creators far less risk adverse than they should be. Those companies owe TRILLIONS to the point where economists have pointed out that essentially 'failure is not an option' although I lived throught the dot.com bubble and it was incredible how investment companies would pour in money when they didn't have anything but a very basic understanding of what the internet was and would become. To most investors, ANYTHING novel was good enough.

I've known people who directly conned investors in the dot.com boom and walked away rich because none of the investors had the slightest understanding. They had developed a narrow-band video codec just as broadband was arriving but because it could just about provide a grainly 8FPS image 9.6kb/s (2G) they thought it would be HUGE. Nobody had mentioned that satellite phones typically had the same band-width and that a video-codec existed. It just never got converted to 2G because ENCODING took too much CPU time so there was only a tiny market... WE knew, the three guys who paid £50 million for 51% of the shares did not... I can go find the name if it interests people.

To the average idiot, AI looks really impressive. It's only when you test it in depth that you learn it will emphatically state one case but if you repeat the question in a slightly different way, it will just as emphatically state another case with just as much 'evidence'. Because AIs do not know the QUALITY of evidence. They seem it parse 'truth' as 'the case stated most frequently'. In academic work we term this 'woozling' (from Winnie the Pooh). But I can point to 14 papers all stating a case then a single 15th paper that shows the error in all of the earlier papers.

Free is NOT a sustainable price-point. So soon an income has to be at least part of planning. As I said, Claude made so many mistakes that it hit the value of any listed company who had AI as PART of it's portfolio. No pure AI company has gone public because the technology isn't mature, it makes laughably basic mistakes and for me the real 'killer' is that when an AI gets something wrong, even the creators cannot say why. So from a legal perspective, if AI is used and an error is made, the law WILL blame the person who used the AI. Europe is in fact already beginning to do this. AI generated text and images are now legally required to be marked as such and the AI companies are also required to watermark the output. It won't be perfect but I really believe AI was being sold as a way to AVOID accepting fault so at least our response is to assume fault so at least results will still have to be checked by a person - whose fault it will be.

So as I have noted. I see a future where people will subscribe to training-sets. Academic papers are all paywalled and whomever owns those is in a strong position as they can parcel up training-sets for every subject (where overlap means papers will appear in multiple training sets) but the important bit will be things like those people would fund people such as Retraction Watch so that they can sell their training-sets as being MORE accurate that the current way people access papers. Not many people check if a paper had been Retracted so automating that will then allow them to show a clear benefit to users who understand what they are buying.

Selling improved accuracy is great but I've read a LOT of papers claiming AI to be 'better' and in every case seemingly arbritary limits were set. They aren't arbritary, those tests were designed to show the AIs in their best light.

Like the recent psychology study in which chatbots were compared to real people and in which for no apparent reason, a limit of 15 rounds of conversation were allowed. Gee, I wonder if it's because it takes approximately 10 rounds of phatic communion BEFORE any real conversation takes place? So in truth the first 10 rounds are almost a script which an AI can mimic so no test AT ALL. Or a test in which only 5 rounds mattered. Students were limited in what they could discuss to exclude things the training-set didn't cover. Not inherenly lying, but nowhere nearly as powerful as someone who doesn't understand human communication might think.

Every study I've seen has limits which are never explained - so that alone should be suspicious. The fact they didn't even bother to explain the reasoning of why limits were imposed is where you start looking at the authors and wondering if that while they COULD claim no conflict of interest, authoring a high-impact article which by design overstates the QUALITY of an AI is a pathway to a much better job. None of us can answer than BUT I tracking them and I would risk £1 that the authors end up as 'consultants' a year and a day AFTER the stufy was published (to avoid the letter of conflict of interest if not the spirit).

BE Skeptical. It isn't for anyone to disprove a statement, it's up to the person making the statment to provide evidence. Hitchen's Razor. If an arbritary limitation is listed without explaination, how that paper got through peer-reviews is a question I cannot answer... but I bet an AI could; convincingly.
 
Last edited:
@unodelacosa - you know, using AI to list all of the wonderful benefits of AI is akin to asking a a murderer to list the benefits of killing people.

I may be one of the few people to see both the utility as well as the risks of AI.

Let's start with pointing out that AI is a misnomer. No intelligence is involved. As far as I know every chatbot available to the public just scraped the internet and scanned books for it's dataset. So one issue right away is that an AI is only as good as it's training-set and since digital publishing is essentially free and also seems to avoid legal liability, it's not a great training-set thus problematic.

The 'Claude crash' shows how over-confidence and the crippling cost of developing a general purpose chatbot will tend to make the creators far less risk adverse than they should be. Those companies owe TRILLIONS to the point where economist have pointed out that essentially 'failure is not an option' although I lived throught the dot.com bubble and it was incredible how investment companies would pour in money when they didn't have anything but a very basic understanding of what the internet was and would become.

I've known people who directly conned investors in dot.com but walked away rich because none of the investors had the slightest understanding. To the average idiot, AI looks really impressive. It's only when you test it in depth that you learn it will emphatically state one case but if you repeat the question in a slightly different way, it will just as emphatically state another case with just as much 'evidence'. Because AIs do not know the QUALITY of evidence. They seem it parse 'truth' as 'the case stated most frequently'. In academic work we term this 'woozling' (from Winnie the Pooh). But I can point to 14 papers all stating a case then a single 15th paper that shows the error in all of the earlier papers.

Free is NOT a sustainable price-point. So soon an income has to be at least part of planning. As I said, Claude made so many mistakes that it hit the value of any listed company who had AI as PART of it's portfolio. No pure AI company has gone public because the technology isn't mature, it makes laughably basic mistakes and for me the real 'killer' is that when an AI gets something wrong, even the creators cannot say why. So from a legal perspective, if AI is used and an error is made, the law WILL blame the person who used the AI. Europe is in fact already beginning to do this. AI generated text and images are now legally required to be marked as such and the AI companies are also required to watermark the output. It won't be perfect but I really believe AI was being sold as a way to AVOID accepting fault so at least our response is to assume fault so at least results will still be checked by a person.

So as I have noted. I see a future where people will subscribe to training-sets. Academic papers are all paywalled and whomever owns those is in a strong position as they can parcel up training-sets for every subject (where overlap means papers will appear in multiple training sets) but the important bit will be things like those people would fund people such as Retraction Watch so that they can sell their training-sets as being MORE accurate that the current way people access papers. Not many people check if a paper had been Retracted so automating that will then allow them to show a clear benefit to users who understand what they are buying.

Selling improved accuracy is great but I've read a LOT of papers claiming AI to be 'better' and in every case seemingly arbritary limits were set. They aren't arbritary, those tests were designed to show the AIs in their best light.

Like the recent psychology study in which chatbots were compared to real people and in which for no apparent reason, a limit of 15 rounds of conversation were allowed. Gee, I wonder if it's because it takes approximately 10 rounds of phatic communion BEFORE any real conversation takes place? So in truth the first 10 rounds are almost a script which an AI can mimic so no test AT ALL. Or a test in which only 5 rounds mattered. Students were limited in what they could discuss to exclude things the training-set didn't cover. Not inherenly lying, but nowhere nearly as powerful as someone who doesn't understand human communication might think.

Every study I've seen has limits which are never explained - so that alone should be suspicious. The fact they didn't even bother to explain the reasoning of why limits were imposed is where you start looking at the authors and wondering if that while they COULD claim no conflict of interest, authoring a high-impact article which by design overstates the QUALITY of an AI is a pathway to a much better job. None of us can answer than BUT I tracking them and I would risk £1 that the authors end up as 'consultants' a year and a day AFTER the stufy was published (to avoid the letter of conflict of interest if not the spirit).

BE Skeptical. It isn't for anyone to disprove a statement, it's upto the person making the statment to provide evidence. Hitchen's Razor. If an arbritary limitation is listed without explaination, how that paper got through peer-reviews is a question I cannot answer... but I bet an AI could; convincingly.
I use AI alot for electrical queries, I shouldn't I know but it does just gleam information from the Internet (reddit especially) I've found if I word the question several different ways I'll get the answer I'm looking for or to confirm I havnt made a mistake .
Now that is not good
 
BTW As for AI generated 'creative writing'. I think publishers are now adding clauses (because you cannot hold the copyright on AI-generated text) and for a writer, even if they fool existing software tests, they cannot predict that at some point software won't be developed that WILL spot it.

AI can be useful in clarifying things you DO know. Rephrasing them in a more coherent way. But even then I've not have a single case in which I didn't have to point out a VERY basic error. I assume because there simply isn't a large training-set on medicinal chemisty for the reasons I gave.

Claude screwed up by claiming to hold all UK caselaw since 1174 but it turned out that no, it had gleaned caselaw from secondary sources. So a large firm of lawyers ended up losing almost all of it's practice after in just ONE day it cited 34 examples of case law that did not in fact exist.

So with clean training-sets, pushback if a question is ambiguous and being honest when a case is uncertain all need to happen. But then it LOOKS less impressive to the average investor. Having to admit it doesn't know or isn't cetrain.

We begin by designing tools until tools start designing us.
 
Last edited:
BTW As for AI generated 'creative writing'. I think publishers are now adding clauses (because you cannot hold the copyright on AI-generated text) and for a writer, even if they fool existing software tests, they cannot predict that at some point software won't be developed that WILL spot it.

AI can be useful in clafiyinying things you DO know. Rephrasing them in a more coherent way. But even then I've not have a single case in which I didn't have to point out a VERY basic error. I assume because there simply isn't a large training-set on medicinal chemisty for the reasons I gave.

Claude screwed up by claiming to hold all UK caselaw since 1174 but it turned out that no, it had gleaned caselaw from secondary sources. So only large firm of lawyers ended up losing almost all of it's practice after in just ONE day it cited 34 examples of case law that did not in fact exist.

So with clean training-sets, pushback if a question is ambiguous and being honest when a case is uncertain all need to happen. But then it LOOKS less impressive to the average investor. Having to admit it doesn't know or isn't cetrain.

We begin by designing tools until tools start designing us.
I'm repeating myself but I can ask the same question using the same words but in a different format just to give me the reassurance that the mistake I have made by say using the wrong size of cable justifies me
There is bs7671 which if I were to reference would make it clear I've made a mistake
Ah it's been a tough night
 
@4DQSAR how would you define "intelligence"?

alasdair
Well obviously if I were to look it up in the dictionary my name would be near the top
But in all honesty if we are talking about AI it's just shit gleamed from millions of stuff on Internet
Some humans can be very intelligent but mostly they are standing on the shoulders of giants
So I can't answer your questio sorry
 
Intelligence is understanding how to classify a statement as wisdom, knowledge, information or data.

thanks.

i guess that leads to the next question: how do you define "understanding"?

from your definition, i think that's a hierarchy, right?

i.e. data is raw facts > information is data that has been organized or put in some context > knowledge is information that's been internalized > wisdom is the application of knowledge

if you accept that current a.i. can certainly classify statements along that hierarchy, you can argue a.i. is intelligent.

if you believe that a.i. doesn't understand the distinctions in the hierarchy, then a.i. is not intelligent.

which brings us back to me question: how do you define "understanding"?

and/or what's the difference between understanding (in the human sense) and a.i. pattern-matching output that looks like understanding?

alasdair
 
Well gosh alasdairm, you moved the goal-posts on me ;-) But I know you enough to know it's asked in good faith so I will do my best.

Understading is the process of applying the appropriate logic when classifying information.

Stating that a heirachy exists does not infer that appropriate logic was used to create that heirachy.
 
Last edited:
Judgment day might be coming, but honestly I can’t live without AI anymore. I start with the simple stuff — fixing grammar, cleaning up replies, correcting spelling, answering emails, doing research, even helping track down bugs at work. Thank God we still have AI in the office.
 
Top