Master AI (Artificial Intelligence) Discussion/News Thread

[TLEDDY]

I have tried many search engines and Perplexity has proven most accurate and reliable in my testing.
YMMV!

As I have gained experience with Perplexity, I have refined my queries. Most importantly, I have instructed it to use only original source documents - not to use news reports or other secondary sources.

I also ascertained that it looked at article summaries and conclusions, not looking at the underlying research data or critically evaluating such. I am old (84) and have decades of experience in a number of disciplines; as a result, I have a sense that, when something does not seem logically correct, I go to the original paper to confirm. Should the underlying factual data not support the conclusion(s) to my satisfaction, I report that to Perplexity. To its credit, it will review and comment, so far concurring with my findings.

I believe that AI is an excellent research tool, being able to search vast quantities of material in more detail and speed than any human resource.

That said, it reflects the current capabilities of AI…who knows the future capabilities! For now, use your own knowledge base to structure your inquiries and evaluate the responses, be a critic and use your brain.

 

[Axo1989]

I prefer sticking with the current thread. Whether you are on the side of the Boomers or the Doomers or the Skeptics or you are unfamiliar with the matter, it is worthwhile to understand the perspectives of the other sides. It is also easier to track in a single thread. We are not getting paid to research the matter, and so far I find the conversation in this thread rather desultory, which is fine by me (I have not read much of the earlier pages). So far, I have been mostly absorbing news and explanations about AI from watching a few YouTube videos. It seems to me that often the term AI is being used to refer to LLMs that use the Transformer architecture, though AI is a much broader field. It appears that LLMs may be over-hyped by the companies involved, and will end up being only a part of the eventual answer to AGI. LLMs can be faulty and yet useful at the same time, and the error-proneness may not go away sufficiently. The 'world models' currently under development sound like a more plausible route to AGI. It seems that many of the disagreements expressed in the thread have less to do with AI itself, and more to do with societal and economic issues.

Hey, my suggestion was made of the highest-grade enthusiasm.

 

[Petrushka]

On the usefulness of AI:

I have almost completed a rewrite of a 25 year old website for a small manufacturer.

The biggest challenge was photography. The products are intricate, and the original studio images are unavailable. Just the low res jpegs used on the site. Complete with dated backgrounds. I have photoshop, but the latest version will not install on my ten year old computer.

I subscribed to Krea, on the recommendation of my browser’s AI search. I subscribed to the $20 trial first, then upgraded.

I have about a hundred images, and have done about half. I tell the AI to upscale the image, enhance sharpness and detail, remove the background, and create soft shadows. On a holiday weekend there is no queue, and it takes about a minute.

The results are spectacular. At the size they are displayed, they are transformed into completely un embarrassing product photos. No grain, no jaggies, no weirdness at the transition of product and background. Even small see through areas are correctly handled.

 

[JustAnandaDourEyedDude]

Please do not take our responses negatively.

I recind my request. I guess it is the characteristics of a forum that I just don't see the point. Trolling...tribally sticking to a viewpoint no matter what example anyone else gives to at least have some pause or self question... Universalizing your experience to equal truth. Never questioning maybe this AI thing others are getting things out of it that I see the opposite... Maybe I using it wrong...maybe I may just try a tip someone offered up. Anyway I will just avoid the forums and stick to the reviews and what I can get out of it.

You are welcome to begin a new thread with a narrower focus geared towards positive discussions and usage tips on AI, and see whether that thread would take off in popularity. If you do so, please let us know so that those of us who would like to take part or observe could track the new thread too.

However, the pluses for sticking with the current thread are that it has lasted 106 pages so far, and so it has legs, and that being designated as a Master thread begun by Amir, it has higher visibility. Although the number of people who seem to be following it and posting to it seems to have thinned of late.

I have noticed the bickering that was alluded to, and attribute most of that to jockeying for status and insisting on the authoritativeness of one's own opinion. In this case, I just stay out of it, and merely skim those posts when reading. Folks sometimes allow their egos to get in the way of modifying their views in the face of counter-examples and well-argued opposing viewpoints. Ideally, genuine arguments in science are aimed at getting to the truth, and are conducted in a civil manner. I think the current thread remains usable despite the bickering. Most of us will read positive and genuine contributions to the thread with interest.

I think the negativity of some towards AI, that is based on its hallucinations, will diminish with time. The models will be improved, and people will also tire of reiterating the point about hallucinations. Some of the other disturbing aspects of AI, such as safety risks, potential massive job losses, increase of wealth inequality, diversion of resources to building out AI infrastructure, use of AI for surveillance and cybercrime, and which company is pulling ahead or winning the race, will continue for awhile to be important topics of discussion.

 

[JustAnandaDourEyedDude]

Hey, my suggestion made of the highest-grade enthusiasm.

I like your suggested title marrying AI and aliens! It definitely escapes the "desultory" category. Will it age well, though?

 

[Timcognito]

A recent AI encounter I shared recently on a PM:
My latest experience with AI was pretty surprising. Living with the ocean 100 meters to west, mountains blocking civilization 20 miles to the east and north, and empty farmland to the south the internet has been huge issue. 40 yrs of copper wire phone service, then adding DSL internet 5 mbs, two years cell based internet that was fast but often died during storms along with our phones, so we kept our DSL landline. Well after replacing our forth dead 5G hub I told my wife that I've had it and though I am very short of fondness for Mr. Musk, we are going Starlink who's basic service was on sale for $40/mo with free equipment with guaranteed 50-100 mbs or 200 mbs for $80, all we need is 100mbs.

Here's the AI part. When I got on the support line, was I transported to a tech in the Philippines as I had been with crashed 5G hubs? No, a robot with life like voice, perfect diction and razor fast answers to very complicated technical questions that I had complete detailed understanding of. I had personally installed and maintained all aspects of structured wired panel during a home remodel to the studs with miles on Cat5e and coax in every room in two buildings, phone, intercom, routers, switches in the furnace closet. Every description of my hardware particulars was understood a couple of seconds and aspects of the brand and model of my hardware with specs and solutions. I could not get myself to not treat it like a I was talking with real person, thanking for help, asking about alternatives, or which was better option. Every response was quick, detailed and to the point. The only unnerving part almost every reply or inquiry to me was preferenced by or had my name in it, "that would be a smart strategy, Tim" or " Tim, what brand and model router are you connecting to", let me look up the those specs , Tim". The bot never said It's name so I could not reply in kind or thought to ask. I left those calls feeling good about the advice and what I was doing but a little spooked interacting as if it were a another human. Instead of waiting for a call back, every call was connected in less than minute followed my new girlfriend "Is this Tim".

So far after a few minor changes my system is working perfectly as I knew it would from friends experiences. ATT was charging over $100 mo for my a land line $65 for internet and $60 mo for their separate VOIP which would have been as unreliable as their 5G hubs, I can now get it <$20 mo and satellite TV $160 mo will be $80 with a comprehensive streaming service.

So the most savvy tech with most experience can be cloned, summoned in a minute, can speak the callers dialect, understand and communicate with anyone at any skill level, have patience of Job, and never insult or be insulted, no waiting 24/7/365. Downside a lot savvy and lousy humans will lose their customer service jobs. Check it out without buying if only to speak to the perfect bot woman. The future seems amazing me.

 

[antcollinet]

A recent AI encounter I shared recently on a PM:
My latest experience with AI was pretty surprising. Living with the ocean 100 meters to west, mountains blocking civilization 20 miles to the east and north, and empty farmland to the south the internet has been huge issue. 40 yrs of copper wire phone service, then adding DSL internet 5 mbs, two years cell based internet that was fast but often died during storms along with our phones, so we kept our DSL landline. Well after replacing our forth dead 5G hub I told my wife that I've had it and though I am very short of fondness for Mr. Musk, we are going Starlink who's basic service was on sale for $40/mo with free equipment with guaranteed 50-100 mbs or 200 mbs for $80, all we need is 100mbs.

Here's the AI part. When I got on the support line, was I transported to a tech in the Philippines as I had been with crashed 5G hubs? No, a robot with life like voice, perfect diction and razor fast answers to very complicated technical questions that I had complete detailed understanding of. I had personally installed and maintained all aspects of structured wired panel during a home remodel to the studs with miles on Cat5e and coax in every room in two buildings, phone, intercom, routers, switches in the furnace closet. Every description of my hardware particulars was understood a couple of seconds and aspects of the brand and model of my hardware with specs and solutions. I could not get myself to not treat it like a I was talking with real person, thanking for help, asking about alternatives, or which was better option. Every response was quick, detailed and to the point. The only unnerving part almost every reply or inquiry to me was preferenced by or had my name in it, "that would be a smart strategy, Tim" or " Tim, what brand and model router are you connecting to", let me look up the those specs , Tim". The bot never said It's name so I could not reply in kind or thought to ask. I left those calls feeling good about the advice and what I was doing but a little spooked interacting as if it were a another human. Instead of waiting for a call back, every call was connected in less than minute followed my new girlfriend "Is this Tim".

So far after a few minor changes my system is working perfectly as I knew it would from friends experiences. ATT was charging over $100 mo for my a land line $65 for internet and $60 mo for their separate VOIP which would have been as unreliable as their 5G hubs, I can now get it <$20 mo and satellite TV $160 mo will be $80 with a comprehensive streaming service.

So the most savvy tech with most experience can be cloned, summoned in a minute, can speak the callers dialect, understand and communicate with anyone at any skill level, have patience of Job, and never insult or be insulted, no waiting 24/7/365. Downside a lot savvy and lousy humans will lose their customer service jobs. Check it out without buying if only to speak to the perfect bot woman. The future seems amazing me.

I had a similar experience - although via text chat - when my FIL was being moved to an alternate email service (his ISP was dropping email service).

Very targetted and precise help in getting everything working perfectly first time. And as you say - it is great to not be put on hold - or need to go through an endless sequence of "if you want xxx please press 3...."

I'm guessing in these scenarios it is much easier to train an LLM on just the tech stuff related to the particular service being provided.

 

[Timcognito]

I'm guessing in these scenarios it is much easier to train an LLM on just the tech stuff related to the particular service being provided.

Agreed there so many known scenarios, devices and trouble spots in a closed system, whereas asking about a skin ailment, for example, in an LLM, there is much to be interpreted from verbal or written descriptors. That is why there is a lot more work going on in VLMs (vision language models) such as in medical diagnosis, Photo, X-Ray and MRI interpretation and VLAs (vision language action) which involve robotics, with some kind mechanical outcome, sorting or steering, etc. LLMs that glean data from the internet are the low hanging fruit and the simpler model is prone to more to mistakes with too much stored unverified data versus real time measurement.

 

[Timcognito]

More on VLAs and VLMs here

The art of grasping - Berkeley Engineering

Redefining robot manipulation with creativity and code

engineering.berkeley.edu

 

[EERecordist]

International government AI safety efforts.

To me the interesting point is the comparison to aviation software development and drug testing.

In the article, the UK is spending $480M for their institute. I know NIST was starting in the area with a tiny budget, an attempt by lobbyists to block it, and I believe the defense department wants to take it over.

Inside the British Lab Hunting for Dangers Lurking in A.I.

The government’s A.I. Security Institute, staffed by alumni from OpenAI and Google, is becoming a model for countries grappling with A.I.’s emerging risks.

www.nytimes.com

 

[JustAnandaDourEyedDude]

Pope Leo iantroduces his very first encyclical which um pontificates on moral and social imperatives that ought to govern guardrails for application of AI. The co-founder Chris Olah of Anthropic also makes affirming remarks. There are several videos on YT with the full public speeches and with analysis.

 

[rdenney]

Please do not take our responses negatively.


You are welcome to begin a new thread with a narrower focus geared towards positive discussions and usage tips on AI, and see whether that thread would take off in popularity. If you do so, please let us know so that those of us who would like to take part or observe could track the new thread too.

However, the pluses for sticking with the current thread are that it has lasted 106 pages so far, and so it has legs, and that being designated as a Master thread begun by Amir, it has higher visibility. Although the number of people who seem to be following it and posting to it seems to have thinned of late.

I have noticed the bickering that was alluded to, and attribute most of that to jockeying for status and insisting on the authoritativeness of one's own opinion. In this case, I just stay out of it, and merely skim those posts when reading. Folks sometimes allow their egos to get in the way of modifying their views in the face of counter-examples and well-argued opposing viewpoints. Ideally, genuine arguments in science are aimed at getting to the truth, and are conducted in a civil manner. I think the current thread remains usable despite the bickering. Most of us will read positive and genuine contributions to the thread with interest.

I think the negativity of some towards AI, that is based on its hallucinations, will diminish with time. The models will be improved, and people will also tire of reiterating the point about hallucinations. Some of the other disturbing aspects of AI, such as safety risks, potential massive job losses, increase of wealth inequality, diversion of resources to building out AI infrastructure, use of AI for surveillance and cybercrime, and which company is pulling ahead or winning the race, will continue for awhile to be important topics of discussion.

Just a friendly suggestion: the above could be a textbook example (though certainly unintentional) of gaslighting. Those bringing up risks and potential abuses are neither crazy nor obstinate, so making them sound that way attacks who they are, not the substance of their concerns.

Rick “with all due respect” Denney

 

[Timcognito]

To me this the scariest part of AI that people get hyper-focused on issues like this, significant theoretical things that may or may not be happening or even significant threat. Similar to nuclear arms, that only the US has used and so many others have for over half a century, and ignore instead the giant underground conventional weapons proliferation, in some cases produced by the US and its allies, feeding proxy wars on several continents killing countless many civilians over many decades. Lets keep making AI the next Nuclear bomb and instead let it go ahead and kill us with a thousand cuts stealing our money and privacy. IF is the tiny word with the big meaning but not when there no history or probability of an accordance. AI is in its infancy and needs real data to get real outcomes so large horrific things need access that same kind of data IF it exists. My 2cents.

What if all the US updated hard/software, amalgamated all computer data of all US the agencies and departments to use AI follow the money to reduce redundant services, tax, subsidy and medical fraud, and measure defined outcomes of programs with real time data for their effectiveness.

 

[JustAnandaDourEyedDude]

Just a friendly suggestion: the above could be a textbook example (though certainly unintentional) of gaslighting. Those bringing up risks and potential abuses are neither crazy nor obstinate, so making them sound that way attacks who they are, not the substance of their concerns.

Rick “with all due respect” Denney

Thanks for your suggestion. I think you mis-pegged my posts. In my first reply to AudioX3, I said that whichever side someone espouses, "it is worthwhile to understand the perspectives of the other sides". I have watched several YT videos of long-form interviews of Doomers like Eliezer Yudkowsky (?), and also followed the hoopla when OpenClaw and Moltbots appeared in the wild. I listed several risk-related topics in my post as being of continuing importance (I forgot to mention chatbot-induced addiction/romance/psychosis, automated kill-chains in warfare, copyright violations and IP theft, among others as topics of interest). I am not subscribed to any single POV on where AI is headed and what is good and bad about it, and I count discussions of risks and negative consequences of AI as positive contributions to this thread. I do not see that I am gaslighting (intentionally or unintentionally) the views of such posters as being wrong or misplaced, and I am not dismissive of them in any way.

However, I did acknowlege AudioX3's complaint that the discourse has occasionally descended to the level of "you don't understand AI" and "i'm going to put you on ignore", which I termed bickering. I mentioned that I have been able to weather the bickering by not engaging in it, and that I favor continuing with this thread (which has a broad scope) rather than starting a new thread focused solely on positive aspects of AI usage. Reading that AudioX3 was bothered by people regularly mocking AI hallucinations, I said that I expected such negativity to diminish with time (by implication become less of an annoyance to him). And that other negatives of AI would continue to be important topics. In no way did I say that the people who point out AI's tendency to hallucinate when lacking data, or the incongruity of these hallucinations with the marketing and CapEx of BigAI, are wrong to do so.

 

[xanalog]

Holy hell, that looks awful. There's Star Wars fan films made for 25,000€ that look way better. And have actors.
And they expect people to pay money for this?

HiggsfieldAI's "Hell Grind" was created in just 14 days, with just a 15 member team.
A comparable traditional action-fantasy film would typically cost around $50 million, making "Hell Grind" roughly 99% cheaper to produce.
Disney's current "The Mandalorian and Grogu" booked-in at a budget of $166.5 million.

At this year's Festival De Cannes, this Adilkhan Yerzhanov' film was only shown at an "industry event" rather than within the 'official program'.
Either way and at the end of the trailer, you may realize that this [first] AI-generated sci-fi episode must put shivers in the sphincters of the Hollywood movie moguls, as was intended.
That old swamp has needed a good drain-cleaning!

 

[JustAnandaDourEyedDude]

Lets keep making AI the next Nuclear bomb and instead let it go ahead and kill us with a thousand cuts stealing our money and privacy.

I agree with you that we should not be hyper-focused on the "next nuclear bomb" concerns about AI to the exclusion of all other concerns. Thought and discussion should also be put towards forming strategies to avert lesser harms that may be sneaked in by actors with motives other than the common good. We should evaluate risk as likelihood times harm. However, when a new disruptive technology is introduced, we do not have a good handle on a full enumeration or likelihoods of all failure modes. So the initial focus is naturally on possible catastrophic outcomes that do irreparable harm rather than on lesser harms that could be mitigated after they take place.

AI is in its infancy and needs real data to get real outcomes so large horrific things need access that same kind of data IF it exists.

"Real outcomes" is ambiguous to me here. Simply giving AI control or agency over a real physical system leads to real outcomes, be they desired or undesired by the rightful human owners or operators of the physical system. If you are arguing that an AI should not be given agency over large horrific systems without training on adequate data, in order to avoid undesired outcomes, I agree with you. I would go further and require that the AI be designed and validated with further safeguards (with protocols, sensors and actuator logic) to produce no undesired outcomes, beyond the original safeguards designed into the system. The further safeguards are needed because of the AI's superhuman decision and action speed, the lack of understanding and common sense, and the inability to predict the consequences of its actions, as compared with expert human operators. It is not only "large horrific things" such as nuclear weapons systems that need AI control models to be comprehensively validated, but also anything that counts as critical infrastructure such as superdams and electric grids or mission-critical equipment such as for crewed spaceships.

What if all the US updated hard/software, amalgamated all computer data of all US the agencies and departments to use AI follow the money to reduce redundant services, tax, subsidy and medical fraud, and measure defined outcomes of programs with real time data for their effectiveness.

Does this have anything to do with a company whose name begins with P and ends with r, or with a CEO whose last name ends with k? I would consider this a positive development that aids in enforcing better compliance with laws and regulations, and that aids in better halting/punishing fraud and abuse, with the usual caveats.

Caveat: The greater degree of data aggregation requires regularly verified and more stringent safeguards against unlawful disclosure and unlawful use of the data. Caveat: All personnel with access to the data (be they authorized users or system admins or hardware maintainers) must undergo periodic background checks and obtain security clearances at the appropriate level. Caveat: Any classified data must remain within air-gapped systems certified for classified use. Caveat: Affected citizens and legal residents must have legal appeal and redress rights against erroneous conclusions drawn from AI analyses. Caveat: The AI must be designed and trained to minimize errors on such datasets, and to explain its reasoning. Caveat: Probabilistic AI processes should not lead to discriminatory outcomes (it should not happen that Person X is flagged for fraud while Person Y who took the same steps to commit fraud is not flagged). Caveat: All citizens and lawful residents should receive training and updates in applicable laws and regulations. I would prefer that people would receive augmented reality apps that would warn them in real-time when they risk breaking any laws or regulations, in view of the increased monitoring and analysis of data relevant to compliance with a very complex system of laws. I realize that ignorance of the law is not an excuse, but yet I cannot spend inordinate time to study laws and case precedent rulings in order to avoid breaking them. And I am sure there are more caveats that might occur to reasonable people.

If such a system were perfect and assured perfect compliance with the law, then we could turn our attention to the laws themselves. Ask questions about which ones are fair and just, and so on, and what might be a fairer compromise on a particular issue. Lots of argumentation among various factions of people.

 

[Audionaut]

Because AI gets stuff wrong

As I said: You have to be the expert - and AI the Assistant.

 

[rdenney]

I agree with you that we should not be hyper-focused on the "next nuclear bomb" concerns about AI to the exclusion of all other concerns. Thought and discussion should also be put towards forming strategies to avert lesser harms that may be sneaked in by actors with motives other than the common good. We should evaluate risk as likelihood times harm. However, when a new disruptive technology is introduced, we do not have a good handle on a full enumeration or likelihoods of all failure modes. So the initial focus is naturally on possible catastrophic outcomes that do irreparable harm rather than on lesser harms that could be mitigated after they take place.


"Real outcomes" is ambiguous to me here. Simply giving AI control or agency over a real physical system leads to real outcomes, be they desired or undesired by the rightful human owners or operators of the physical system. If you are arguing that an AI should not be given agency over large horrific systems without training on adequate data, in order to avoid undesired outcomes, I agree with you. I would go further and require that the AI be designed and validated with further safeguards (with protocols, sensors and actuator logic) to produce no undesired outcomes, beyond the original safeguards designed into the system. The further safeguards are needed because of the AI's superhuman decision and action speed, the lack of understanding and common sense, and the inability to predict the consequences of its actions, as compared with expert human operators. It is not only "large horrific things" such as nuclear weapons systems that need AI control models to be comprehensively validated, but also anything that counts as critical infrastructure such as superdams and electric grids or mission-critical equipment such as for crewed spaceships.


Does this have anything to do with a company whose name begins with P and ends with r, or with a CEO whose last name ends with k? I would consider this a positive development that aids in enforcing better compliance with laws and regulations, and that aids in better halting/punishing fraud and abuse, with the usual caveats.

Caveat: The greater degree of data aggregation requires regularly verified and more stringent safeguards against unlawful disclosure and unlawful use of the data. Caveat: All personnel with access to the data (be they authorized users or system admins or hardware maintainers) must undergo periodic background checks and obtain security clearances at the appropriate level. Caveat: Any classified data must remain within air-gapped systems certified for classified use. Caveat: Affected citizens and legal residents must have legal appeal and redress rights against erroneous conclusions drawn from AI analyses. Caveat: The AI must be designed and trained to minimize errors on such datasets, and to explain its reasoning. Caveat: Probabilistic AI processes should not lead to discriminatory outcomes (it should not happen that Person X is flagged for fraud while Person Y who took the same steps to commit fraud is not flagged). Caveat: All citizens and lawful residents should receive training and updates in applicable laws and regulations. I would prefer that people would receive augmented reality apps that would warn them in real-time when they risk breaking any laws or regulations, in view of the increased monitoring and analysis of data relevant to compliance with a very complex system of laws. I realize that ignorance of the law is not an excuse, but yet I cannot spend inordinate time to study laws and case precedent rulings in order to avoid breaking them. And I am sure there are more caveats that might occur to reasonable people.

If such a system were perfect and assured perfect compliance with the law, then we could turn our attention to the laws themselves. Ask questions about which ones are fair and just, and so on, and what might be a fairer compromise on a particular issue. Lots of argumentation among various factions of people.

Further caveats: 1.) before AI can be used to prosecute a crime, the prosecutor must collect real-world evidence to verify the crime, meaning that an AI conclusion is not only not admissible in court but that it is also not probable cause for an arrest (I did not say probable cause for a search warrant). There have already been serious abuses of AI affecting real (clearly innocent) people wrongfully accused and incarcerated based only on AI-based tech such as facial recognition about which the police were woefully credulous and undiscerning.

2.) AI-driven processes must be supervised by a legally responsible supervisor, the lack of which constitutes gross negligence and exposes the system owner to tort liability. The users of AI, or the people AI processes affect, must have profound legal recourse that does not require millions in a legal fund if they are harmed, with a civil legal standard of the preponderance of the evidence. I would probably throw in minimum treble punitive damages plus legal costs as a deterrent. That means Congress will have to legislate tort liability to overcome the sovereign immunity claimed by government agencies in many states. And it means that use of AI is a proprietary function, not a government function, which further minimizes a claim of sovereign immunity. It also means that overall statistics showing the betterment of mankind from use of that product can’t be used as a defense against the liability for torts committed against any one individual. Right now, tech companies are willing to throw money at victims, but that willingness will certainly end.

AI doesn’t just need technical guardrails, it needs legal guardrails. I’m sure a lawyer in the right specialty area could think of more than just those two.

Rick “IANAL” Denney

 

[Berwhale]

As I said: You have to be the expert - and AI the Assistant.

Yes, beware of fabricated specificity, for example.

 

[gaz]

HiggsfieldAI's "Hell Grind" was created in just 14 days, with just a 15 member team.
A comparable traditional action-fantasy film would typically cost around $50 million, making "Hell Grind" roughly 99% cheaper to produce.
Disney's current "The Mandalorian and Grogu" booked-in at a budget of $166.5 million.

At this year's Festival De Cannes, this Adilkhan Yerzhanov' film was only shown at an "industry event" rather than within the 'official program'.
Either way and at the end of the trailer, you may realize that this [first] AI-generated sci-fi episode must put shivers in the sphincters of the Hollywood movie moguls, as was intended.
That old swamp has needed a good drain-cleaning!

I am happy to watch movies like this, have no need for movie actors - they often get too popular and I get too familiar with them and their mannerisms - few are good enough to contain these, Morgan Freeman being an exception. I often prefer unknown actors and also enjoyed the "Final Fantasy" movies with animated, realistic humans.

And I have always thought popular actors' fees were ridiculous, would have preferred this was shared with all actors. And production budgets can also be so too - I loved "Carnivale", but the cost was so high the popularity only justified two seasons instead of the planned six; maybe they can use AI to make the other four...

Maybe we will have a world where most movies and series are made without actors, no problem for me. However I like theatre and and will always be a patron.