Are Large Language Models Really AI?

I think AI is no different than most things in life: Garbage in, Garbage out. This is true of the training data, the end users input at prompts, and any attempts to filter or moderate. (also for factual or moral data).

Time would probably be better spent letting people know the output is AI (watermarks/metadata) than trying to filter it or worry about whether its right or wrong.

Yup - Reactor a face type software tried to make a filter that senses NSFW and then someones like well... if I use Forge SD I can buypass that..

Then they make a Lora file of the person, and it's all downhill from there...

You guys are playing too much in the dark AI.

True.

False.

Its trained on everything its creators can get their hands on. Books, magazines, scholarly dissertations, movie scrips, radio transcriptions. If the library of congress were all online they would have used the whole darn thing. They did use all of it that is.

Almost anyone can download. Not near anyone would know what to download (I had never heard of stable diffusion.) Far fewer know how to render anything. Grok and other AIs have democratized deep fake porn to where any mouth breathing 8th grade dropout can create it.

I agree but with this, as with economics, its difficult. For some posters its impossible.

Super true - I know tons of people with 4080s and 5080s who don't touch the stuff - think it's stupid - others just don't care or don't want to be bothered - then you have people with 4060 8gb using guff models to make videos and art now

Ive been scratching with AI for a long time now, no expert or anything, but comfyui works along with getting someones cutting edge workflow to work, and it's getting pretty crazy what I am able to do now. Whenever I want to update my setup, I run my start-up log to Gemini or grok, and it looks online for the latest updates from GitHub, reviews forum posts to ensure its geting the best info, and then I briefly tell the AI to ensure it makes no errors so it triple-checks before telling me how to proceed

These two things aren't contradictory. Pretty much all media that's online has a moderation that is left of Radial's politics. I'm not going to argue about where the overton window is right now or what qualifies as left/right, but scrapable data is almost certainly further left than fully uncensored conversation.

SD is an opensource desktop version of what Grok / ChatGPT and the various other image AI's do behind the scenes. It uses your own GPU (or set of GPUs), you use plain text to tell it what to do and it does it. Think of photoshop filters only with AI.

https://stablediffusionweb.com

To do local install
https://github.com/AUTOMATIC1111/stable-diffusion-webui

https://www.reddit.com/r/StableDiffusion/

This is what I mean by "dumb" in reference to those doing bad things using online tools that track and report to federal agencies.

I didn't say they were. Indeed they are complementary. Also scraping the web is way easier than ripping apart physical paper media and feeding it to scanners. (Which is legal and avoids copyright issues unlike torrenting.)

I have a feeling that the whole world is about as far left of Radial's politics as is the web.

I highly doubt it. I doubt he's even in the top 10 percentile of persons with those kind of convictions.

You are 100% right but don't use this - it hasn't been updated in a long while

Use SD forge or Comfyui

https://github.com/Haoming02/sd-webui-forge-classic/tree/neo

If he was, he wouldn't be able to post about it here, and we'd have no way of knowing.

Which is kind of the point. Far left extremist views are often encouraged on major platforms. Far right extremist views are almost universally censored. The training reflects what's available to train on.

I mean the fact the guy lives in UK and isnt in jail tells me he's nowhere near as "far-right unhinged" as some people here believe he is.

Chanti, you are a far left hippy (literally, by your own description of stuff you used to do for fun to annoy the rest of society when you were younger).

What I believe is what the vast majority of people in the world believe who earn less than $100k a year (and most people in every western nation too).

The world we live in is one where a minority impose their beliefs on everyone else, against the masses. Which is why the censorship / cancel system exists in the first place, and why one platform not following that forced system is so outrageous to them.

Yes. Still am. Left enough to still lament "sure the liberals are behind us, far behind us."

The vast majority of people in the world earn less than $100k a year period.

On this we agree. I would hope the vast majority of people would agree too but I honestly believe that far too many are oblivious.

The censorship / cancel system is truly bipartisan.

I will post no more on this. Its off topic. PMs are welcome.

On AI, regulation or the lack of it, and everything:

The trap Anthropic built for itself
TechCrunch

Saw that before, I find it quite humorous.

Accurate. Black Mirror. (15 million merits)
https://en.wikipedia.org/wiki/Fifteen_Million_Merits

Nvidia now offers DLSS5, which uses AI in real time to upgrade the graphics of any game.

https://x.com/NVIDIAGeForceUK/status/2033625439315399137

This will do far more to normalize AI in the gaming industry than anything else.

I now have two things to immediately disable when I load up a new game:
1-Motion Blur
2-DLSS5

I already know it will be as detrimental to my FPS as Denuvo is.

Pokemon GO was used to train AI delivery robots:

"You train robots and don't even know it

30 billion photos. Collected over 8 years. By hundreds of millions of people who thought they were playing a game.

Pokémon Go wasn't just a game - it was the largest crowdsourced mapping operation in history. And now that data is driving real robots through real streets.
On March 10, 2026, Niantic Spatial - the AI spinoff from the creators of Pokémon Go - announced a partnership with Coco Robotics, an autonomous delivery startup.
The deal: Niantic Spatial will power Coco's delivery robots with its Visual Positioning System (VPS) - a technology that tells a robot exactly where it is, not by GPS, but by what its cameras see.
Here's how it works: the robot snaps a few photos of nearby buildings and landmarks, the system matches them against a massive database - and pinpoints the robot's position down to a few centimeters.
For context: regular GPS in a city can drift 50 meters, placing you on a different block, facing the wrong direction, on the wrong side of the street.
But the real story isn't the tech. It's where that database came from.

How players unknowingly built the dataset
Pokémon Go launched in 2016 and exploded overnight. 500 million installs in 60 days. Hundreds of millions of people pointing their phone cameras at buildings, parks, and monuments - chasing virtual creatures layered onto the real world.
But the real play came in late 2020, when Niantic quietly added AR Mapping.
Players were asked to scan real-world objects - statues, storefronts, fountains - by walking around them with their cameras on. It was framed as "Field Research": complete the task, earn in-game rewards.
The genius was in the mechanics. Niantic could point players at any location - just assign it a task, and thousands would rush there, phones in hand. Every scan uploaded to Niantic's servers with rich metadata: GPS coordinates, camera angle, phone orientation, speed, direction, sensor data.
Each scan looked like a routine game action. But combined across millions of players, they formed the most detailed visual map of the physical world ever created.
Forget Google Street View sending a camera car once a year. Niantic built an army of millions of volunteer mappers - working for virtual trinkets worth nothing in the real world.

30 billion images and why no company could replicate this
The number is staggering. But the real value isn't the size - it's the diversity.
Mapping companies send camera cars that capture streets in controlled conditions. Clean, uniform, sterile images. Pokémon Go players produced the opposite: messy, chaotic, real-world data. Exactly what robotics AI needs.
A single Pokéstop might be scanned in morning sunlight, evening rain, winter snow, and a crowded weekend afternoon. Thousands of different people approaching from different directions, at different heights, at different angles. No mapping company on earth could replicate this variety on the same timeline or budget.
The result: for each of over 1 million hotspot locations, Niantic Spatial has thousands of images - same place, wildly different conditions. Every image tagged with precise spatial data.
Niantic never deployed a single camera car. They gave people a game and let them do the work.

https://x.com/k1rallik/status/2033530122150502482
"

At least those nerds did something useful for the society.

Fixed.

Question: Was this the plan all along or serendipity?

Supposedly that DLSS5 preview needed two 5090s to run lol

For $3000, your games can look amazing

For an extra $8000, your games can be enhanced with AI slop to look “better”

I am sure they had some idea that they could profit from the data at some point.

I'd point back to this paper as the financial justification for their data collection:
https://static.googleusercontent.com/media/research.google.com/en//pubs/archive/35179.pdf

Basically, nerds observed that some classes of models kept getting better as you trained them on super large datasets. So you could build opinionated models and train them on modest datasets, or you could use unopinionated models and train them on really massive data sets.

That finding, and basic class of unopinionated models, ended up generalizing to text generation, image generation, etc.

No one knew that performance would scale this high, but they have been pursuing this in a principled way for almost 20 years.

I saw a thing earlier about unions in relation to AI, people still really do not seem to be able to grasp this is why jobs are being taken away in the first place.

Unions are simply not a good thing anymore, especially when we already live in a fair modern western society where you're not a slave they can beat to death.

Blue collar workers created powerful unions (steel unions etc), once they became too powerful the bosses simply replaced them all with off shore factories (and when mass immigration became acceptable (aka now), mass immigration of their replacements). Unions destroyed the steel industry in the west. Then white collar workers started doing it too cause they are apparantely too stupid to learn from blue collar workers mistakes (or more likely, they thought immigrants would never replace them), now AI and more immigration will take your place too and remove all that short lived leverage the unions gave.

Ironically, as they feel more pressure they want to join unions all the more and that just makes it happen even faster cause your radicalize your employers.

Sometimes you gotta accept the corporation has power over you and do a good job without being a pain in the ***, or at least know when you got enough and stop your union antagonizing them. This however won't happen cause unions are for profit orgs with high paid leadership, who push pay rises and strikes when they don't get it.

AI won't replace workers, it will simply allow imported lower quality Indian and other immigrant workers to do your job to an acceptable standard (The Indian prime minster recently said his country is the mother of the world, and it's pretty clear what he meant by that). It won't matter if it's inferior to what national workers could previously do, the customers will simply be forced to accept lower standards due to having no alternative.

A good example of this is the NHS in the UK, they import cheap and under-educated doctors, nurses and carers from the 3rd world (often with fake or inferior degrees) and since you have no option, you just have to put up with these people treating you. They don't even want to take on young British doctors and nurses anymore who they trained, cause are more likely to unionize and demand more pay rises from the government. So these people often forced to goto the states due to demanded DEI hire quotas at home that allow them to hire these cheap immigrants over them. There are no jobs for the home grown doctors and nurses they trained, even though they say there is a shortage and that is why they have to import.

There are propaganda posters in any health location, telling the UK people that diversity doctors and nurses from the 3rd world are saving the NHS (even though they can barely speak english and are trained to a laughably low level). What they mean is, saving the government who pay them less and don't have to put up with them joining a union.

Unions have done more damage long-term to all western workers than anything else imaginable over the past 30-40 years, they are why we have off-shore factories, why we have mass immigration to the level we do and why AI is inevitable.

Everywhere is laying off masses of people, and hiring lots of Indians that will use AI (trained on your work)

The billionaires always get what they want, if you find a way to have some leverage over them they will completely annihilate you in the long run.

Seeing union memes in 2026 is the dumbest thing ever, to the point I'm not even sure if they are legit and instead are false flags from the corporations themselves.

Respectfully, I must strongly disagree with the assertions in the original text. Far from being obsolete or uniformly harmful, labor unions have played and continue to play a vital role in protecting workers’ rights, improving workplace standards, and ensuring more equitable distribution of economic gains. The narrative that unions alone caused offshoring, mass immigration, and the rise of AI is an oversimplification that overlooks broader economic, political, and technological forces. Below I outline a more balanced perspective addressing the key claims and showing why unions remain an important counterbalance to corporate power.

First, consider the historical record. Unions were instrumental in achieving many of the workplace protections and social benefits that people now take for granted: the eight-hour workday, workplace safety regulations, child-labor laws, overtime pay, collective bargaining rights, health and pension benefits, and the establishment of minimum wage standards. These gains did not arise from corporate benevolence; they required organized, collective action to negotiate for fairer terms and to compel legislative reforms. To suggest that unions “destroyed” industries like steel ignores the complex interplay of globalization, automation, shifts in comparative advantage, trade policy, and management decisions. Offshoring and deindustrialization were driven largely by changes in technology that reduced the need for certain kinds of labor, differences in labor costs globally, and deliberate policy choices promoting free trade and investment flows—not solely by unions’ bargaining successes.

Second, unions serve as an important check on concentrated corporate power. In modern economies, corporations often have vast asymmetries of information, resources, and bargaining leverage relative to individual employees. Collective bargaining reduces that imbalance and gives workers a structured way to negotiate wages, benefits, working conditions, and job security. Without unions, many workers would have little recourse when facing arbitrary discipline, sudden layoffs, unsafe workplaces, or discriminatory practices. The existence of unions improves workplace transparency and accountability and helps enforce labor laws. Far from “antagonizing” employers for the sake of conflict, unions institutionalize negotiation and dispute resolution processes that reduce chaotic, ad-hoc confrontations and can produce more stable labor relations.

Third, the argument that unions are inherently self-serving or “for-profit” organizations that push strikes and untenable pay rises mischaracterizes how unions operate. While like any institutions they can suffer governance problems and leadership excesses, many unions operate democratically and reinvest negotiated gains into members’ livelihoods. Strikes and industrial actions are typically last-resort measures used when good-faith bargaining fails or when employers refuse to address serious grievances. Moreover, the assertion that unions radicalize employers and thereby accelerate job losses ignores evidence that robust labor standards can incentivize firms to innovate in ways that preserve jobs and increase productivity. When workers have voice and stability, they are often more motivated, productive, and able to contribute to continual improvement rather than being treated as disposable inputs.

Fourth, linking immigration and the use of AI exclusively to unions misunderstands the drivers of migration and technological adoption. Migration is influenced by global disparities in income, education, political stability, and demographic trends. Employers’ demand for labor from abroad is frequently shaped by labor shortages, immigration policy, and the global mobility of talent—not simply a desire to evade unions. In sectors such as health care, migration of professionals often fills genuine shortages driven by population aging, training bottlenecks, and underinvestment in domestic workforce pipeline. Responsible policy responses include scaling educational capacity, improving retention, and designing fair credential recognition systems—not simply blaming unions.

Similarly, AI and automation are technological forces driven by advances in computing, data availability, and capital investment. While firms may deploy automation in response to labor costs or to increase efficiency, the policy question is not whether automation will happen but how its benefits are shared. Unions can and should be part of the solution: negotiating retraining programs, transition support, and responsible implementation that protects workers’ livelihoods while enabling companies to compete. Framing AI as an inevitable job destroyer that unions accelerate is a fatalistic view that removes agency from policymakers, employers, and employees to shape technology deployment in socially productive ways.

Fifth, consider empirical evidence about inequality and shared prosperity. Over the past several decades, declining union density in many advanced economies has correlated with rising income inequality and stagnating wages for middle- and lower-income workers. Where unions remain strong, wage compression is more common and the benefits of economic growth are distributed more broadly. Unions also help raise standards for non-union workers through “spillover” effects: competitive pressures and norms mean that employers often match broader sectoral standards for compensation and safety. Strong unions therefore contribute to more resilient middle-class incomes, healthier consumption demand, and social stability—factors that ultimately sustain long-term corporate profitability too.

Sixth, the claim that employers will necessarily “annihilate” labor leverage in the long run underestimates the potential for democratic and regulatory responses. Societies have the capacity to regulate the pace of technological adoption, to protect critical industries, to support domestic training, and to set immigration policy that balances labor market needs with community interests. Unions play a critical role in those democratic processes by representing workers’ voices in public debates about taxation, education, industrial strategy, and social insurance. The alternative—leaving workers without coordinated representation—weakens democratic bargaining over the rules that govern markets.

Finally, criticizing unions for urging higher pay or better conditions ignores a basic moral argument: work deserves dignity and fair compensation. Economic systems that permit long-term erosion of wages and deterioration of working conditions in the name of efficiency risk social fragmentation and political instability. Unions are one institutional mechanism that channels workers’ legitimate grievances into collective bargaining rather than into less constructive forms of protest.

In short, unions are not a relic or a single cause of complex economic trends. They are a crucial institutional counterweight to concentrated economic power, a channel for worker voice, and a vehicle for negotiating the social impacts of technological change. Rather than abandoning collective representation, a more constructive approach is to reform and modernize unions where needed—improving governance, fostering inclusivity, partnering on training and innovation, and engaging proactively in shaping public policy around AI, immigration, and industrial strategy. That combination of strong worker representation, sensible regulation, and proactive public investment offers a far better path to equitable prosperity than the false choice between submitting to corporate power or accepting unregulated, technocratic displacement.

Please write a disrespectful and very brief reasoning on why panta needs to wear a muzzle if he doesn't refrain from posting AI generated walls of text in the future. Make it 2 sentences long so the mutt can comprehend it without needing AI to translate.