Opinions on AI

[freeagent]

I guess its too late to unplug it huh..

 

[ThrashZone]

Heavily regulated depending on where you live.

Hi,
Indeed making it easier to be programed to do what is written instead of a human trying to comprehend it all plus their personal stances getting in the way.

But of course ai might blow a fuse or two with contradictions

 

[Calenhad]

I was looking for the both option, but didn't find any.

 

[lexluthermiester]

Hi,
Indeed making it easier to be programed to do what is written instead of a human trying to comprehend it all plus their personal stances getting in the way.

But of course ai might blow a fuse or two with contradictions

In reality, combining human and AI workloads to keep each other in check would be the best way forward.

 

[Dr. Dro]

Duolingo isn't an educational setting. It's a for-profit app.

It would be far harder to integrate AI into a school setting (and these things are regulated). The truth is, until there are autonomous, walking robots with the ability to roam freely without restriction, we need to rein in our fantastical ideas about what AI will do. Algorithms already perform most automated functions. For AI to do as people fear, will take a huge leap in technological development.

I mean, I agree, but then again so are most schools and institutions... at least where I'm from.

Duolingo has been an excellent asset for self-learning, the "gamification" of the learning process is effective and resonates well with a lot of people, myself included. At a point, I had subscribed to Plus for a year to practice my Spanish (which I can read, understand, and communicate in - pero no mucho, I have a heavy preference for using English whenever I can and the rest of my practice time I put towards Japanese, which I'm hoping to adopt as a fourth language), I feel like this could heavily affect the quality of the service.

Then there's this, which is what really irks me and I just know will happen:

https://twitter.com/i/web/status/1744238708092526956

A mistake such as this one could easily cost you a test, and to ask for a review? Just thinking about the bureaucracy and paperwork has my head spinning

 

[the54thvoid]

Remember that profits drive these decisions, and ultimately, humans manage it. AI isn't the real problem, it's how it's used. And everyone will want to use it to further their own agenda, be that for profit, enlightenment, or enslavement to dogma.

 

[lilhasselhoffer]

Wow the time and resources it must have taken to do that. Elon likes to do his own thing and I can't help but think that deepfake was a hit job. Literally corporate warfare, but that is just a theory... a conspiracy theory.

What's real is what is closest to you.

It's only a matter of time a big chunk of programmers will be out of a job. Learn-to-code is dead, long live learn-to-code.

I cannot tell if this is an elaborate troll...or you actually believe that this is what AI automation will look like.

Let me elaborate. If you believe the mountain of garbage that AI, via LLMs, has created is art then you're wrong. I wish I could sugar coat this, but it's wrong for the exact same reason that a white sheet submitted by an artist is worth money, but if next week you submitted a similar canvas it'd be worth precisely as much as the materials. Creatives should start getting worried when the LLMs can generate TV show length content, fully animated, with dialog. Until that day the creative industries are basically beholden to their current production machinery.

Now, the programming thing is a joke...right? I hope it is, because it assumes that programs are simply a bunch of copy-paste code blocks...and theoretically you have the patience to iterate enough times to get exactly what you want. That's what I like to call a nightmare, because when you interact with a code monkey half of their job is simply to understand what you actually want. This is the same logic of going to a garage and telling the grease monkey there that "the car is making a funny noise when I accelerate" has to be interpreted into something akin to "the brake pads need replaced because they are worn" by the mechanic. If you don't understand that I'd suggest that either you don't understand how things work or you've drank the AI kool-aid.


As far as companies laying off translators...that's a stupid example. That's literally what an LLM can do...so it's saying that this pseudo-AI is working as intended...which is to understand X->Z and return Z whenever prompted...because you don't have to know why to return the right answer.



Let me explain what this "AI" is. It's a man in a box, with a stack of books. The books are written in a foreign language the man cannot comprehend, but by isolating a number of these symbols the man knows that what he should return when prompted is a sheet with a bunch of other symbols. X input trained to Y output, with literally no understanding of X or Y. Your man in the box just knows they go together...and with a billion potential inputs and outputs that man appears to know the language...but in reality could next give you a response to the question "What is dog kibble made from" that makes no sense, like a recipe for cooking canines to eat.
If programmers have to fear that, then they truly are lost. I for one believe that code monkeys will be banging away for years, because iterative development is something executives will never have the patience to tolerate, and writing clear and concise non-expanding goals is anathema to how modern companies work.

 

[Assimilator]

I feel like this could heavily affect the quality of the service.

Do you know what else would affect the quality? If Duolingo outsourced everything to Fiverr. Would you be complaining about that in an AI-focused thread? I doubt it. Yet both are problems of capitalism, not of technology, so stop blaming the wrong thing FFS.

Then there's this, which is what really irks me and I just know will happen:

https://twitter.com/i/web/status/1744238708092526956

A mistake such as this one could easily cost you a test, and to ask for a review? Just thinking about the bureaucracy and paperwork has my head spinning

C'mon dude, multiple-choice tests where an obviously wrong answer is marked as the correct one is not something new, it's been around as long as tests have. If you're going to start trying to blame obviously human-caused problems on LLMs, we cannot have a constructive dialogue about the latter.

 

[ExcuseMeWtf]

AI is a tool first and foremost. It has great potential to improve our lives in countless ways, and it has great potential to backfire on us horribly in countless ways. Just like you can use a knife to prepare food or to stab someone.

It's all up to how we train it.

 

[Assimilator]

writing clear and concise non-expanding goals is anathema to how modern companies work

This is why software developers are so highly paid. Not to write software, but to translate the bullshit that product people regularly pull out of their arses, into working software.

There used to be highly-paid business analysts who did the bullshit-translation part and produced specifications, and lower-paid developers would then transform the latter into working software. This worked great because it allowed the analysts to focus on doing the thing they're best at, and the developers to do the thing they're best at. Then managers came along and decided that the analysts weren't actually providing any value because they weren't producing output that could be measured, so paying them is a waste of money, thus the analysts were fired and the developers got lumped with their job too. The result is that developers are expected to do both the thing they're bad at as well as the thing they're good at, which inevitably means they become mediocre at both.

Now management complains that developers aren't working fast enough, or aren't listening to what the product people are saying, and users hate the software they're forced to use because it's buggy and doesn't do what the product people told them it would. But hey, the managers justified increasing their paychecks by cutting the analysts', so that's great.

I despise capitalism.

 

[Dr. Dro]

Do you know what else would affect the quality? If Duolingo outsourced everything to Fiverr. Would you be complaining about that in an AI-focused thread? I doubt it. Yet both are problems of capitalism, not of technology, so stop blaming the wrong thing FFS.


C'mon dude, multiple-choice tests where an obviously wrong answer is marked as the correct one is not something new, it's been around as long as tests have. If you're going to start trying to blame obviously human-caused problems on LLMs, we cannot have a constructive dialogue about the latter.

I will concede that these errors can be caused or created by humans, because I don't really have proof that they were machine generated. But what counts is the food for thought

 

[A Computer Guy]

I cannot tell if this is an elaborate troll...or you actually believe that this is what AI automation will look like.

Haven't been in this thread in awhile so not sure if all your commentary is directed solely at my reply so just commenting on some bits.

Let me elaborate. If you believe the mountain of garbage that AI, via LLMs, has created is art then you're wrong. I wish I could sugar coat this, but it's wrong for the exact same reason that a white sheet submitted by an artist is worth money, but if next week you submitted a similar canvas it'd be worth precisely as much as the materials. Creatives should start getting worried when the LLMs can generate TV show length content, fully animated, with dialog. Until that day the creative industries are basically beholden to their current production machinery.

I think this part is a reply to AusWolf's comment. In that regard I interpreted his reply as people in some capacity will start using these "AI" tools to create "art" instead of seeking out human counterparts to do so.

Now, the programming thing is a joke...right?

It was some light humor but not totally devoid in the idea that "AI automation" will/can change the programmer landscape. Some code monkeys will loose their job somewhere in the push to "AI" everything. From a management perspective the use of "AI" is a kind of outsourcing. Why pay a person's salary and HR issues when you can rent a server with one of those fancy "AI" chips the US gov't doesn't want China to have and slave it 24/7 to maybe do something.

I hope it is, because it assumes that programs are simply a bunch of copy-paste code blocks...and theoretically you have the patience to iterate enough times to get exactly what you want. That's what I like to call a nightmare, because when you interact with a code monkey half of their job is simply to understand what you actually want.

This is the same logic of going to a garage and telling the grease monkey there that "the car is making a funny noise when I accelerate" has to be interpreted into something akin to "the brake pads need replaced because they are worn" by the mechanic.

A good shop will have a shop manager directing the grease monkeys and interacting with the clients. They will often collaborate together in the diagnosis and resolution of the car problem. Of course some shops are just one guy.

If you don't understand that I'd suggest that either you don't understand how things work or you've drank the AI kool-aid.

The chain of conversation I was replying to had to deal with concerns regarding human motivations to potentially misuse "AI" (regardless if it's being marketed and popularized as something it's not)

As far as companies laying off translators...that's a stupid example. That's literally what an LLM can do...so it's saying that this pseudo-AI is working as intended...which is to understand X->Z and return Z whenever prompted...because you don't have to know why to return the right answer.



Let me explain what this "AI" is. It's a man in a box, with a stack of books. The books are written in a foreign language the man cannot comprehend, but by isolating a number of these symbols the man knows that what he should return when prompted is a sheet with a bunch of other symbols. X input trained to Y output, with literally no understanding of X or Y. Your man in the box just knows they go together...and with a billion potential inputs and outputs that man appears to know the language...but in reality could next give you a response to the question "What is dog kibble made from" that makes no sense, like a recipe for cooking canines to eat.
If programmers have to fear that, then they truly are lost. I for one believe that code monkeys will be banging away for years, because

iterative development is something executives will never have the patience to tolerate, and writing clear and concise non-expanding goals is anathema to how modern companies work.

 

[lilhasselhoffer]

Haven't been in this thread in awhile so not sure if all your commentary is directed solely at my reply so just commenting on some bits.

I think this part is a reply to AusWolf's comment. In that regard I interpreted his reply as people in some capacity will start using these "AI" tools to create "art" instead of seeking out human counterparts to do so.

It was some light humor but not totally devoid in the idea that "AI automation" will/can change the programmer landscape. Some code monkeys will loose their job somewhere in the push to "AI" everything. From a management perspective the use of "AI" is a kind of outsourcing. Why pay a person's salary and HR issues when you can rent a server with one of those fancy "AI" chips the US gov't doesn't want China to have and slave it 24/7 to maybe do something.



A good shop will have a shop manager directing the grease monkeys and interacting with the clients. They will often collaborate together in the diagnosis and resolution of the car problem. Of course some shops are just one guy.

The chain of conversation I was replying to had to deal with concerns regarding human motivations to potentially misuse "AI" (regardless if it's being marketed and popularized as something it's not)

Clarifying the first bit, you are saying coders will lose their job. It's an art, not a science.

My retort, put simply, is that because it is an art it's not going to damage the code monkey. There are at least 4 different ways to write a simple program whose only goal is to output "Hello World" in Visual. I know because I've seen at least that many....and expect many more. Now, write a string of code that executes a basic addition function, creates a new value, and outputs the value. Way more ways to do that...and it's a simple program. Try writing a simple application, and it's millions of lines of code to both generate and optimize. That's what code monkeys art is...and they generally do it well.


As far as your retort on the maintenance...an extra supervisor to the grease monkey is a human code monkey reviewing the stuff the AI code did...which makes the argument of replacing a human pretty irrelevant.



Above, @Assimilator took what I said and copied it. The translation of specifications, and therefore goals, is why AI is not the right solution (in the form of a LLM). As that is all the AI we have, there's no reason for caution. Even an infinitely large LLM is only as good as it's allowed to iterate and refine its return weighting. Artists, like code monkeys, shouldn't fear LLMs....because lord knows we have enough to actually worry about.


Also note that artist is not a dirty word here, but a valid term for someone capable of using skills to do a thing without a rigid right and wrong means to an end.

 

[DavidC1]

I despise capitalism.

It's not capitalism, but human flaws. Socialism, communism, feudalism, you can try whatever you want, but human flaws make it so none of them works.

Little down the ladder, the fault is not the big corporations as much as the governments that ALLOW corporations to do as such. The problem with government is that it's the biggest corporation. But it all ties back to the first sentence, human flaws.

For example with the emissions scandal with VW. In many countries, the company got fined a massive amount. Instead, they should have been done zero, and the people responsible should have been fined, and prosecuted. And by fine I mean a proportional fine, where a CEO with $100 million assets get fined $99 million, and a regular engineer with $500K assets getting fined $200K. More responsibilities, more penalties. The rest will be sent to prison. Many times I have seen an individual fine having a ceiling in how high it can go, and for high level managers in a mega corp, it's peanuts.

What Fines do is that the beancounters within the company and the managers look at it and say "Oh, low balance sheet, so let's "readjust" our company", aka downsizing, reducing R&D, ultimately hurting the ones not responsible. We treat the company like an individual, and punish as such, when the company is worth zero without the people running it. By the same logic, the people are responsible for the crimes, thus the people need to be punished. If it was, and in a consistent way, then people would be discouraged from doing it.

The reason better policies don't exist is because again, government.

 

[R-T-B]

Hi,
Indeed making it easier to be programed to do what is written instead of a human trying to comprehend it all plus their personal stances getting in the way.

But of course ai might blow a fuse or two with contradictions

I don't know where you all went to school, but I was taught critical thinking very early on.

The reason better policies don't exist is because again, government.

But Somalia called and Anarchy is awful...

 

[AsRock]

Hi,
Indeed making it easier to be programed to do what is written instead of a human trying to comprehend it all plus their personal stances getting in the way.

But of course ai might blow a fuse or two with contradictions

You think it get to the point that when some thing does blow a fuse and happens to kill some one it will have rights or at the least the maker of said AI could get away with murder ?.

Already robots that can carry guns.

 

[Assimilator]

Little down the ladder, the fault is not the big corporations as much as the governments that ALLOW corporations to do as such.

And why are the corporations allowed to do this? Because they use their wealth to manipulate governments into doing so.

 

[T-Cat]

A.I. (or MORE specifically machine learning) like most things are tools. They will have a big impact but it can go either way. To think all innovations lead to good things is foolish. In fact there are two infamous examples: the nuclear bomb and the cotton gin. The nuclear bomb is self explanatory however nuclear power is a thing so that was not all bad. As for the cotton gin I am specifically talking about the mechanical one innovated by a man by the name of Eli Whitney.

What made the cotton gin so bad you might ask? Well in the United States slavery was a thing as many of you know. What many might not know was that this innovation completely threw plans of "banning slavery gradually" out the window as it made cotton profitable. Granted this is a controversial subject but for the most part the founding fathers expected slavery to die off naturally since again plantations were not profitable until that innovation completely changed this trajectory.

As a result of the cotton gin, plantation owners became very wealthy... and long story short it was a pivotal innovation that lead to the American civil war (and would also help fuel it as the south was completely reliant on cotton profits). That being said the gin also made numerous other things possible and was very important in the progression of the industrial revolution.

Moral of the story for the cotton gin is that inventions are neither good or bad. They are just tools that serve a purpose and can be used equally for good or bad. In some ways the internet has been the same way as historians will be looking at the benefits and consequences of its invention.

Generally speaking machine learning can only go so far but I imagine its going to have consequences regardless. Generally speaking it will benefit some businesses like accounting as machine learning will only make this process faster. As for content creators... it will be very rough as places like japan have put maximum protections on A.I. training which basically means copyright will not help artists anymore as of now. Though this is not the first time nor will it be the last that history has continuously screwed over artists. Though artists/content creators also have the biggest opportunity of creating content faster and of equal or better quality so its not all bad.
link if you are curious about Japan's current stance on machine learning: https://www.deeplearning.ai/the-batch/japan-ai-data-laws-explained/

That being said if A.I. is implemented on a larger scale successfully some places will do it better than others. I imagine parts of Europe will implement it more successfully albeit sluggishly while countries like China or the USA will go full-speed on the technology and possibly create a market bubble or two (and probably create some prime historical examples of how NOT TO USE machine learning).

Either way I would say nuclear annihilation is a bigger problem as its kind of on the table right now (especially with the whole China-Taiwan tensions getting much more tense).

 

[AusWolf]

I was looking for the both option, but didn't find any.

That's what the "other" is for.

 

[Night]

It depends where AI development will head, and how (un)regulated it's going to be. Right now AI isn't replacing anyone completely, but my guess would be that call centers will be the first to be overtaken by AI. Even though AI can't completely replace human workforce for now, look at the example of AI assisted coding in VS 2022:

Spoiler

How does AI-assisted development help?​

The following table describes key ways in which an AI assistant can help you develop in Visual Studio:

Code faster
Let AI help you:

• Generating code and entire function suggestions (for example, how to write code to perform a task by describing it in natural language)
• Predicting what you'll code next based on your programming patterns (completions)
• Code refactoring through AI-driven context-aware recommendations

Understand code better
Ask the AI assistant for:

• Explanations of code sections (for example, when you're trying to understand someone else's code)
• Answers to your programming questions

Profile and debug quicker
Get help profiling and debugging your code:

• Optimize performance based on AI suggestions
• AI-identified bugs & resolutions

If AI assistance can help a programmer write 2 times (example) more code than before, that would mean that the other programmer will become unnecessary for the current amount of work. If the amount of work is going to increase twofold, the other programmer won't get fired, but there is no need to hire anyone else. AI removes the need for that.

 

[Vayra86]

Humanity isn't experienced enough for AI. Not wise enough. Not smart enough. Driven by systems that are only designed to accelerate the worst aspects of it.

We can't even handle social media yet. Technology has raced way past our cognitive ability. We can't filter information proper anymore, and AI is not the cure, its more disease. While there are some who understand it, we need the masses to do so. We need a form of ascension before we can go here. Part of that is systemic change, something young generations call for, but are not getting.

I didn't open this thread to argue about the semantics of the words "artificial intelligence", but to discuss the technology that got this weird name for whatever reason (to sound cool, I guess?).

Other than that, very interesting thoughts!

Well, I do think its relevant to the discussion. The semantics are the threat. Commerce is the driver for this threat. People are playing make believe with people of lesser intelligence.

This is the same issue that plagues social media. Much like AI, its a black box of algorithms that presents us content, while we have no authority in how it does that.

The fact is, we haven't got AI. We just have bigger datasets and a new, experimental approach to them. A form of 'decision making' really. But the highest achievable form of making decisions, is making them through absolute data transparency. Knowing why you choose what you choose, the basis, the arguments behind it, is what truly defines a choice in the end. AI takes that away, rather than providing it. It makes us stupid and aimless. We don't control how the model is trained, so we have no control, and with that, no real intelligence, just a parser spitting out information directed by someone looking to get rich.

 

[AusWolf]

It's not capitalism, but human flaws. Socialism, communism, feudalism, you can try whatever you want, but human flaws make it so none of them works.

Little down the ladder, the fault is not the big corporations as much as the governments that ALLOW corporations to do as such. The problem with government is that it's the biggest corporation. But it all ties back to the first sentence, human flaws.

For example with the emissions scandal with VW. In many countries, the company got fined a massive amount. Instead, they should have been done zero, and the people responsible should have been fined, and prosecuted. And by fine I mean a proportional fine, where a CEO with $100 million assets get fined $99 million, and a regular engineer with $500K assets getting fined $200K. More responsibilities, more penalties. The rest will be sent to prison. Many times I have seen an individual fine having a ceiling in how high it can go, and for high level managers in a mega corp, it's peanuts.

What Fines do is that the beancounters within the company and the managers look at it and say "Oh, low balance sheet, so let's "readjust" our company", aka downsizing, reducing R&D, ultimately hurting the ones not responsible. We treat the company like an individual, and punish as such, when the company is worth zero without the people running it. By the same logic, the people are responsible for the crimes, thus the people need to be punished. If it was, and in a consistent way, then people would be discouraged from doing it.

The reason better policies don't exist is because again, government.

But if you yourself admit that corporations have the power to just downsize and re-plan their strategies following a fine that's worth peanuts in the corporation's revenue, and carry on, then isn't capitalism at fault for giving them such power? In my opinion, capitalism's biggest problem is that it makes people believe that the market will always regulate and balance itself, when in fact, it doesn't. In reality, the big fish eat the small fish, and the whole corporate food chain will eventually collapse into a monopoly. Always. No regulation = anarchy and chaos = the big fish preying on the small until there's no small fish left.

As for downsizing, I'd say it's absolutely capitalism's fault. No other economic system calls on firing those who don't bring any measurable value to the company, and then, blaming every mistake on those who do. A good example is the logistics sector (where I work), where everyone who doesn't have an hourly target, for example, trainers (like myself) are regarded as useless by the majority, including most of management. Few people recognise that without such support departments, the people who do work on hourly measurable targets wouldn't be so successful at delivering what they do, so we do bring value to the company, even if it's not directly measurable.

Edit: As a comparison, socialism (which I also don't completely agree with for various other reasons that I don't want to mention here) calls on giving everybody a job, which is the exact opposite of the above scenario.

Edit 2: Let's stick to the topic of AI. If you want to discuss the above further, PMs are available.

 

[the54thvoid]

I'm not going to LQ or edit the thread, but please stick to the topic of: Opinions on AI.

I will delete OT posts moving forward.

 

[R-T-B]

I'm not going to LQ or edit the thread, but please stick to the topic of: Opinions on AI.

I will delete OT posts moving forward.

Fair. It was getting pretty wild and I at least lost the topic for a moment.

I think it's fair to state my opinion on AI has not been helped by it's utility in new scams and misinformation campaigns. I can't really say it's impact thus far a good year in has been a net positive.

 

[67Elco]

Anyone who has used the AI chat box on Newegg knows full well there is no intelligence in artificial...