AI videos are getting very good. How do we detect their trickery?

@Mendel

We are f*d. I mean, this is going around for a long while now. And also there are many youtubers who are real, but have partly AI generated content. And how about all the AI bots everywhere? Can you tell who is and who isn't? I can guarantee you wont be able. And if you can, then that will very soon change.

It is interesting the lady from "how to cook that" is pointing this AI out, but in my view we are beyond the obvious and easy to spot AI content. It will only get worse and I think we already have videos and youtubers out there that are 100% AI and NOBODY can spot it.
AI being hard to distinguish from real people doing real things has one principal drawback that I can see ...people no longer can trust the media, and therefore find it harder to trust each other. The implications of that are so severe that it makes people turn on the industry itself. (Indeed, the mid-term elections showed AI initiatives to encounter strong opposition in most places that had them on the ballot.) I believe AI will prove to be a fairly short-lived phenomenon, a blip in our history, but for now it's a tool that promotes mistrust among us, whether deliberately or not.
 
Bridgwater Carnival 2026 (UK):


Source: https://www.youtube.com/watch?v=cnXYVhdNPtQ

Actual footage of Bridgwater Carnival 2026:


Source: https://www.youtube.com/watch?v=8YFQIaIGAMg


Source: https://www.youtube.com/watch?v=KUwtimtKnOw


Source: https://www.youtube.com/watch?v=nEkSPZGWxsI


Google search results filtered for YouTube videos lists AI sources first:

gsearch.png

source: https://www.google.com/search?q=bri...zS6DimXnFJYnpRYm5EF5JGZgGK9ZLSlUDAORynGtEAAAA
 
I believe AI will prove to be a fairly short-lived phenomenon...
I've been saying the same thing about the internet, smart phones & vaccines!!

Okay...just kidding on that...but I've been trying to figure out AI's future role for the last couple of years.

I definitely see plenty of promise, in certain areas...but perhaps even more danger.

Having a know-nothing like Trump tell us to ignore the ("fake") warnings, and let the companies self-regulate, was making
me very concerned that the powers that be are not going to take the warning signs seriously enough, until it's possibly too late.

Then the mid-term happened, 4 days ago, and it looks like maybe the grown-ups WILL work towards meaningful guardrails.

Honestly, right now, I wouldn't bet one way or the other...
 
AI being hard to distinguish from real people doing real things has one principal drawback that I can see ...people no longer can trust the media, and therefore find it harder to trust each other. The implications of that are so severe that it makes people turn on the industry itself. (Indeed, the mid-term elections showed AI initiatives to encounter strong opposition in most places that had them on the ballot.) I believe AI will prove to be a fairly short-lived phenomenon, a blip in our history, but for now it's a tool that promotes mistrust among us, whether deliberately or not.
I do remember walking through a Sears store in the early 1990s and seeing something-something sears.com on a sign and wondered why Sears needed a web page... (Had been using dial-up BBSes and CompuServe for several years by then.)
 
I do remember walking through a Sears store in the early 1990s and seeing something-something sears.com on a sign and wondered why Sears needed a web page...)
Almost 30 years ago, to the day (11/3/96) :D the Simpsons had a quick sight gag along these lines.
The family had moved, so Bart & Lisa needed to transfer to Cypress Creek Elementary School.
As they walk in, a sign is seen that includes the (fake, since Fox wasn't a school, & couldn't use .edu) URL http//www.studynet.edu.

That was the whole joke: That this was a fancy school that even had an--obviously unnecessary, thus pretentious--website!
(John Swartzwelder [who wrote "You Only Move Twice,"] could not have known that very, very soon, all schools would have websites!)
1996 11 03 Simpsons Cypress Creek Elementary School URL.webp
 
It's
The first time a search engine offers an option to filter out ai-generated content, hopefully consumer response is going to drive that option everywhere.
If nobody can find it, you can't make money with it.
In my opinion AI has got somewhat popular because it provides something of the value of the old internet to people, which is a plain text answer to a question without "sponsored" results.

You ask the question, you get an answer.

Google had become increasingly less and less useful as you had to scroll past more and more sponsored links and other presentation of paying companies links rather than the original algorithm which presented useful answers 1st.

In fact I kept wondering when they were going to start dropping this kind of thing in to the AI results, they are likely do so by slanting the AI's training, but if so its pretty much invisible to the user unless you ask for specific products and is denied by the AI when you ask, but they seem to still be offering the plain AI text version when you search form something that's not a specifically a product

Search Bluetooth headphones and you get a normal result of sponsored links to product pages

Search why are Flamingos pink and you get an AI response

Google themselves know it's happening and it's costing them revenue but they are playing the long game, because the cat is out of the bag.

https://www.bbc.co.uk/news/articles/cm9qzdqpxy48o
 
I believe AI will prove to be a fairly short-lived phenomenon, a blip in our history,
I keep thinking back to the time when 3D was going to be the "next big thing". Most major film studios were making and releasing films in 3D, older movies were being remastered in 3D, broadcasters were starting to produce live events in 3D, and TV manufacturers were telling us "one day soon everyone will own a 3D TV!"

And the general public said "Nah....We don't want it". And that was the end of that.

Obviously the situation with AI is different because there is an interactive component to it, as well as a vastly different business model (at the moment). But I, like you, am starting to wonder if in the end the reaction from the general public to AI will be "Nah...We don't want it".
 
I believe AI will prove to be a fairly short-lived phenomenon, a blip in our history,
But I, like you, am starting to wonder if in the end the reaction from the general public to AI will be "Nah...We don't want it".
Many already do, but it's not an all-or-nothing thing. Updates on AI have been very significant - irreversibly so. The most significant usage is in various forms of business where it's replacing human workers (and, in some cases, vastly speeding things up, like with math).

But even just in the sense of people using AI to research stuff and make funny images, it's not going away.

Article:
In the new survey, 62% of U.S. adults say they interact with AI at least several times a week.

1794183298927.png


AI for making videos? It's only going to get better. There's a financial incentive to use it, but it also lets you make things that would e impossible for the average person. Yes, the public version makes mistakes, but that's also going to get a lot better. Just look at the state of the art that's now available to the public.


Source: https://www.youtube.com/watch?v=1yLAbK5F3E8


Source: https://www.youtube.com/watch?v=cZOBC-hvfDs


Source: https://www.youtube.com/watch?v=3TzBJB8K0xQ


Sure, some people will reject AI for ethical reasons. But the time of rejecting AI because it just doesn't look right is coming to an end.
 
Many already do, but it's not an all-or-nothing thing. Updates on AI have been very significant - irreversibly so. The most significant usage is in various forms of business where it's replacing human workers (and, in some cases, vastly speeding things up, like with math).
Some of us "interact" with AI because our browsers give us no option. A search for a topic comes up with an AI response before we can get to a site with a more thorough discussion of the matter.
 
Sure, some people will reject AI for ethical reasons. But the time of rejecting AI because it just doesn't look right is coming to an end.
First example (SciFi) seems to have had the benefit of professional post-production. Some actors feel very uncanny.
Second example (Scarecrow) gets a bit of leeway because it has a clearly CGI protagonist without facial expressions.
Third example (Fukushima) "doesn't look right" in many ways, e.g. the continuity error with the car door not closing, and other problems.

I've seen better characters in Chinese 3D animation.
 
I keep thinking back to the time when 3D was going to be the "next big thing". Most major film studios were making and releasing films in 3D, older movies were being remastered in 3D, broadcasters were starting to produce live events in 3D, and TV manufacturers were telling us "one day soon everyone will own a 3D TV!"

And the general public said "Nah....We don't want it". And that was the end of that.

Obviously the situation with AI is different because there is an interactive component to it, as well as a vastly different business model (at the moment). But I, like you, am starting to wonder if in the end the reaction from the general public to AI will be "Nah...We don't want it".
A true 3D TV is technically challenging, the 3D glasses are a poor substitute for it. For me at least (my sight is not exactly good) looking a 3D film is wearing.

Instead AI has become so powerful a tool that I can't see it disappearing at all, just as the 'dark satanic mills' didn't. You have seen examples here on metabunk of its power: allowing (almost) anyone to build in one day a software who in a recent past would have taken months to a skilled engineering team to complete (if ever...). When I read @Mick West's post "... when one of my bots detected a background noise pattern..." (sorry I can't find the reference) I thought I then heard a voice saying "Mr. Spock, may it be the Altarandians making contact with us?" "It's the only logical explanation, Captain Kirk!"

I think AI is an extremely powerful tool. As a hammer can be used to build a house or to devastate it, AI can be used wisely or foolishly. Differently from all other tools, though, AI can also use itself as a tool, and this is creepy and risky (like an animated hammer, fantasy-style, would be).
 
I believe AI will prove to be a fairly short-lived phenomenon, a blip in our history, but for now it's a tool that promotes mistrust among us, whether deliberately or not.
I'd like to think some aspects of it will be short lived, but the reality is the war has only just started. My problem with "AI" is that it's a term currently overused which has plenty of smoke and mirrors marketing & hype attached.

It's often anthropomorphised and given status it does not have (and may never have) e.g. free will, with statements along the lines of "The AI agents got together to break out of a security sandbox to hack computers on the internet". And the media loves this stuff - "we are all going to DIE!". They might be right for various reasons, but a sentient all knowing AI is not one of them IMHO. I think the concrete dangers are more tied to the usual politics of power, control and accountability during this time of oligarchy.

There is talk of an AI bubble collapse - some of that talk is generate by the billionaire tech bro class itself [0] who see themselves as to big to fail, thus able to ride out such a collapse which may benefit them as the market shrinks. I'm sure such a collapse will have them politically lobbying (whining) for the same privatise the profits, socialise the losses mantra that followed the 2008 economic collapse.

Where I am conflicted is that there are some genuine advances in computational tools - for example aiding CRISPR technology. [1] (used in genomics https://en.wikipedia.org/wiki/CRISPR). Being kind to myself, I think I can recognise fruit from decades of computer science research in neural networks [2] since the 1960s, often using seminal programming languages of the time, LISP [3] and Prolog [4].

Having said that, we see the rise of vibe coding which is polarising computer programming. When initially introduced I remember reading the derision aimed at those who indulged. Often seen as second class talent, not 'real' programmers and opening a Pandora's box. Real world consequences are such that vibe coded applications enter production where the output of the LLM coding is not understood by the people initiating the prompting [5][6].

Just seems to me that any productivity (a word that can be used to browbeat) gained in code generation is nullified, as resources - actual humans (hopefully) - are going to be needed to maintain these applications. But, cynically, I'm sure someone is going to vibe code AI to bug fix vibe coded apps. To be fair, LLMs are getting very good at finding vulnerabilities in existing IT infrastructure - and not only in a good way.

The, almost throwaway, term "AI" does not distinguish between the generative 'slop' and the application of models (which, to me, seems like the ultimate in pattern matching) built to solve a specific problem.

Sure, some people will reject AI for ethical reasons.
And artistic. There was some early effort in the advertising industry to 'shame' brands for using AI as cheap, low value and simply not cricket old chap. This seems to be one of the big fronts at the moment. I have a photographer contract with an agency who works in the book publishing industry - they do not allow any AI generated imagery (photographic or illustration) whatsoever as, according to them, their clients (most of the big publishing houses) demand all work is original and exclusive. I think this human led, interpersonal relationship based working practice is going to hold out for quite some time. Artistic integrity still has value.

But the time of rejecting AI because it just doesn't look right is coming to an end.
Another pressure point for sure. From my example above there are self publishing book authors who just want front matter that looks like something from a professional book designer - and, ethically, just because they can who is to say they should not use AI? These arguments are going to run and run.

And how long before AI generated posts & members on MB become undetectable?

[0] https://www.theatlantic.com/technology/2026/03/ai-bubble-defenders-silicon-valley/686340/
[1] https://www.nature.com/articles/s12276-025-01462-9
[2] https://en.wikipedia.org/wiki/History_of_artificial_neural_networks
[3] https://en.wikipedia.org/wiki/Lisp_(programming_language)
[4] https://en.wikipedia.org/wiki/Prolog
[5] https://vibecoders.gg
[6] https://heyainews.substack.com/p/the-price-of-skipping-engineers-two
 
I have a photographer contract with an agency who works in the book publishing industry - they do not allow any AI generated imagery (photographic or illustration) whatsoever as, according to them, their clients (most of the big publishing houses) demand all work is original and exclusive. I think this human led, interpersonal relationship based working practice is going to hold out for quite some time. Artistic integrity still has value.
I would say that it SHOULD have value, but I note how many AI created slop books are available (and apparently selling, as they remain available) on Amazon or in ads on my social media platforms.

A lot of that stuff seems to be personalized stuff ("Give your sweety a personalized book where they are the main character!") or children's books ("Give you child this amazing book about American hero Neil Armstrong, please don't notice that the LM has 5 landing legs and on page 9 there are two moons in the sky!") and the like.

A publishing house with the integrity to refuse that sort of stuff is currently losing a small slice of market to these slop peddlers: I wonder how long the integrity will hold out as and if the slop starts to break out of the niches it is currently invading...
 
The first time a search engine offers an option to filter out ai-generated content, hopefully consumer response is going to drive that option everywhere.
If nobody can find it, you can't make money with it.
I imagine it would be fairly trivial for AI to come up with ways to get around AI filters, though.

I'm not sure how AI could be a "temporary blip". It is here and it's not going away, however depressing that may be. I find that when I am watching or looking at anything now, the constant thought "This will all be done by AI in a couple of years" goes through my head. Who will spend vast amounts of money hiring actors, building sets, paying cinematographers etc when they can just vibe-shoot the whole thing? And probably vibe-screenwrite it too.

Maybe there will be a popular backlash in favour of real creativity, but I'm not sure I have that much faith in the taste of the general public. Show me one area of modern life where quality has won out over convenience and economy.

FWIW, YouTube adverts seem to be almost entirely AI-generated now. I am currently seeing lots of variations of "Sweet old man/woman is retiring from making hand-crafted knick-knacks, get them now before they're gone!"
 
This will all be done by AI in a couple of years" goes through my head. Who will spend vast amounts of money hiring actors, building sets, paying cinematographers etc when they can just vibe-shoot the whole thing?
Yes, I've had the same thoughts. Thankfully (for now anyway) I enjoy laughing at the outcomes. I may be crying instead during the next decade though. I have faith that once the shiny new thing effect disappears, the need to experience emotions crafted by real humans will win out.

And probably vibe-screenwrite it too.
Already happening. I saw a post from Amanda Seyfried this morning saying she is getting scripts written with detectable AI. I hope that the acting unions and the profession in general will be able to prevent total Slopgeddon at all levels of the industry in and outside of Hollywood level budgets.

Maybe there will be a popular backlash in favour of real creativity, but I'm not sure I have that much faith in the taste of the general public.
It's going to be interesting seeing the response to all of this. I'm hopeful as I see quite a lot of backlash already - mainly from insiders and movie fans, but it's all we have to push back with. Don't really know if it's going to be enough. I know what I'll be spending my money watching.

I expect that if AI takes over creativity like this (and I refuse to accept it is a done deal, roll over and accept it looser scenario) then maybe we will see 'Made with generative AI' labels being required by law. Like smoking labels: 'This movie was made using generative AI and could be harmful for your mental health. Long term effects include, but not limited to, detachment from objective reality'.

I wonder how long the integrity will hold out as and if the slop starts to break out of the niches it is currently invading...
I'd imagine a parallel publishing universe where true writers and artists live in the rat infested remains of abandoned cities, eking out a miserable existence exchanging physical MSDOS floppy disks by trusted hands to avoid detection by the AI overlords. In the year 2035 a version of WordPerfect for DOS costs more than a bar of gold.
 
I'd imagine a parallel publishing universe where true writers and artists live in the rat infested remains of abandoned cities, eking out a miserable existence exchanging physical MSDOS floppy disks by trusted hands to avoid detection by the AI overlords. In the year 2035 a version of WordPerfect for DOS costs more than a bar of gold.
That's strongly reminiscent of the scene in the old "Max Headroom" TV series (taking place "twenty minutes into the future") where two characters are wandering exactly that kind of blasted-city locale. One says "What's that thing" and the reply is "It's a book. It's a non-volatile storage medium. Everybody should have one."
 
My daughter is 15 and considering what she wants to do for a living. She's a very good artist and loves graphic design and animation, but I feel like that whole field is rapidly becoming, if not obsolete, at least a shadow of what it was until recently. I'm torn between wanting to encourage her and wanting to shout "No! Don't do it! Train to be a plasterer or a plumber or something!"

FWIW, she is very anti-AI, and so are most of her friends of the same age. An encouraging sign of resistance or just youthful idealism?
 
One of the responses to AI will be the revival of live performances of all sorts. Don't watch a video of a concert, attend it in person. Because the live in-person event demonstrates actual effort, actual capability and a greater degree of actuality in whatever performances are attended.

Live events also demonstrate the value of the event, justifying the cost of attendence. Why should I still pay for a video that has been broadcast and viewed millions or billions of times, the maker already made their money off it?
But a live performance, charging for that is justified because someone is putting new effort into that specific event.
 
One of the responses to AI will be the revival of live performances of all sorts. Don't watch a video of a concert, attend it in person. Because the live in-person event demonstrates actual effort, actual capability and a greater degree of actuality in whatever performances are attended.

Live events also demonstrate the value of the event, justifying the cost of attendence. Why should I still pay for a video that has been broadcast and viewed millions or billions of times, the maker already made their money off it?
But a live performance, charging for that is justified because someone is putting new effort into that specific event.
Given that "live" concerts with digitally generated performers have been around for decades, I don't share your confidence. I don't even think that things have even fundamentally changed much in the last decade or so, it's just that the avatars being used are now more realistic. Difference in polygon count doesn't change the gameplay. The nozzle was probably irreversably uncorked when Hatsune Miku released her 100000th song.
External Quote:

Who is Hatsune Miku?


She is a singing voice synthesizer featured in over 100,000 songs released worldwide.
Hatsune Miku means "The first Sound from the Future".
Facts
* over 100,000 released songs, 170,000 uploaded YouTube videos, 1,000,000 created artworks
* over 900,000 fans on Facebook
* performing sold-out 3D concerts worldwide with performances in LA, Taipei, Hong Kong, Singapore, Tokyo
-- https://web.archive.org/web/20130530182127/http://www.crypton.co.jp/miku_eng

Yup, you're reading that '2013' correctly. And, yes, 100000 songs in 6 years is one every 30 minutes, assuming the program never sleeps (but anyone with a PC can inject her singing into a song, this isn't a single nozzle).
 
One of the responses to AI will be the revival of live performances of all sorts. Don't watch a video of a concert, attend it in person. Because the live in-person event demonstrates actual effort, actual capability and a greater degree of actuality in whatever performances are attended.
Unfortunately, not always. With pre-recorded tracks, sequencers and autotuners even a live event can be as fake as an AI product.

Stick to the basics, no AI here :)
1794259502871.png
 
And artistic. There was some early effort in the advertising industry to 'shame' brands for using AI as cheap, low value and simply not cricket old chap. This seems to be one of the big fronts at the moment. I have a photographer contract with an agency who works in the book publishing industry - they do not allow any AI generated imagery (photographic or illustration) whatsoever as, according to them, their clients (most of the big publishing houses) demand all work is original and exclusive. I think this human led, interpersonal relationship based working practice is going to hold out for quite some time. Artistic integrity still has value.
I suspect this is mostly due to copyright issues. It doesn't do for a publisher to sell content that anyone else can appropriate freely.
Article:
The Office affirms that existing principles of copyright law are flexible enough to apply to this new technology, as they have applied to technological innovations in the past. It concludes that the outputs of generative AI can be protected by copyright only where a human author has determined sufficient expressive elements. This can include situations where a human-authored work is perceptible in an AI output, or a human makes creative arrangements or modifications of the output, but not the mere provision of prompts.

Article:
The rise of generative AI has presented a global challenge to copyright law regarding authorship and human creativity. This has implications for the ability of creators to control and monetise their work, in addition to incentives to use innovative technologies. Currently, the EU lacks specific rules on the copyrightability of AI-generated works, but existing case law of the Court of Justice of the European Union and developments at Member State level demonstrate a strong need for human creativity. The European Parliament has also advocated a human-centric approach while calling for further analysis of how intellectual property rights are applied to content created with the use of AI tools. Most countries, including the United States, maintain a human-centric approach, limiting copyright to works with significant human involvement in the creative process. However, not all the international approaches and practices are aligned, and this briefing gives examples from a selection of different jurisdictions. Certain common-law countries, including the United Kingdom, grant copyright to 'computer-generated' works without a human author while, in a slightly contradictory manner, assigning authorship to the person responsible for the arrangements for the creation of the work. Chinese courts show mixed rulings, sometimes granting copyright based on sufficient human effort in selecting and editing prompts. Cases like the artwork called 'Suryast', produced with the help of AI, reveal conflicting copyright decisions across the world, with AI being recognised as a co-author in some countries. Uniquely, Ukraine has adopted a 'sui generis' right to provide alternative protection for AI-generated images, distinct from traditional copyright. As a result, there may be a need for greater coordination globally to reduce legal uncertainty and to determine the level of human involvement needed for copyright protection.
 
Given that "live" concerts with digitally generated performers have been around for decades, I don't share your confidence.
But those aren't live performances or live concerts.

I like to go to these giant rock/heavy metal festivals, specifically Aftershock in California. Every year it gets bigger, and it was immense this past year, 4 stages, 180,000 in attendance. There are quite a few of these festivals around the country (many put on by Danny Wimmer Presents), and if you look at the Wikipedia articles about them, they are all growing every year. People are hungry for live, real music, I would say increasingly so.
 
But those aren't live performances or live concerts.
Most of them are, and the ones that aren't are being treated as replacements for live concerts. Given the prevalence of lipsynching for even more decades, the actual liveness doesn't really seem to be that important a quality. People are watching *performances*. Some performances are contractually obliged to *not* be played live (e.g. the not-plugged-in RHCP at the Superbowl). To improve the experience, apparently.
 
My daughter is 15 and considering what she wants to do for a living. She's a very good artist and loves graphic design and animation, but I feel like that whole field is rapidly becoming, if not obsolete, at least a shadow of what it was until recently. I'm torn between wanting to encourage her and wanting to shout "No! Don't do it! Train to be a plasterer or a plumber or something!"
I have a 15-year-old nephew, and for several years he's wanted to be a video game programmer. I don't really see that as a viable job now. AI can 10x or 100x the productivity of a game creator. There just won't be a need for junior programmers. Heck, I'm not that sure how long senior programmers will be useful.

FWIW, she is very anti-AI, and so are most of her friends of the same age. An encouraging sign of resistance or just youthful idealism?
I think a lot of young people recognize that AI is robbing them of the possibility of the life paths their parents had available, and it's not at all clear if it's leaving them with anything. Really, nobody knows. It's all speculation.

Anthropic released a new model today Fable 5.2, so I put the question to it on its highest setting. I feel it's a little behind the times (as it says itself, The evidence lags the tools.)

Short answer: My read is that it's still viable, but the entry-level rung is shrinking and nobody, me included, can forecast the early-2030s market he'd enter. I'd tell him to keep making games and aim at the underlying skills, not the job title.

What the data shows (mostly US)

  • Games are unstable, so far mostly for non-AI reasons. Layoffs have run between roughly 8,500 and 15,600 a year since 2022, and 2026 stood at 10,977 by October 20, already past 2025's full-year total (yearly totals, 2026 count). Those figures come from a tracker kept by one analyst, Amir Satvat. GDC's 2026 survey (about 2,300 respondents, skewed toward North America and Western Europe) found a two-year layoff rate of 28%, with nearly half of those laid off still without a new job (GDC, 80.lv, GameDev Reports summary). Restructuring, budget cuts, market conditions and cancelled projects top the reasons companies gave.
  • Entry-level software is where the squeeze shows up. Stanford's ADP payroll study put 22-to-25-year-old developers nearly 20% below their late-2022 employment peak by mid-2025. Its August 2026 revision finds a 19% shortfall for young workers across AI-exposed occupations, growing over time, driven by reduced hiring more than layoffs, and absent among experienced workers (Stanford Digital Economy Lab).
  • Whether AI is the cause is disputed. By the paper's own description these are "early, descriptive indicators", not causal estimates. A June 2026 New York Fed analysis traces about 64% of the rise in young-graduate unemployment to remote work, on the grounds that the rise started before generative AI spread (NY Fed). Yale's Budget Lab has yet to detect a clear AI effect in the overall jobs data (Yale Budget Lab).
  • The evidence lags the tools. The Stanford payroll data stops at June 2026, the month Fable 5 first shipped, so the measured effects reflect older, weaker models.
  • Official forecasts are positive but being cut. BLS's August 2026 projections have software developers growing 10% over 2025–35, roughly three times the all-occupation average (BLS). Last year's cycle said 15.8% (BLS, 2024–34), so about a third of the forecast growth vanished in one revision (my arithmetic). The separate and much smaller "computer programmers" category, about 110,800 jobs, is headed for a 7% decline (BLS).
  • Capability is climbing fast, on tidy tasks. METR's "time horizon" (task difficulty expressed as how long the job takes a skilled human) has followed an exponential trend. Real work is messier than the benchmark, by METR's own account, and its task suite can't reliably measure anything past about 16 hours (METR).

I'm one of these tools, so I'm not a neutral witness. My assessment is that the scarce skill is shifting from writing code to specifying it and judging whether the result is right, and I don't know how long humans keep that edge.

What I'd tell him

  1. Keep building games. A finished game exercises programming, math, performance work and scoping at once, and those skills carry outside games. That's my judgment, not data, though your own post-games work is an example.
  2. Use AI as a tutor, not a ghostwriter. In an Anthropic randomized trial of 52 mostly junior engineers learning an unfamiliar library, the follow-up quiz came out at 50% for the AI-assisted group against 67% for hand-coders, and debugging took the biggest hit. Asking the AI to explain, instead of delegating to it, went with much better scores (Anthropic). Caveats: small sample, one short task, and Anthropic makes me.
  3. Get fluent with the tools anyway. One anecdotal hiring signal: Krafton and Nexon are still recruiting entry-level cohorts and now test applicants on AI-tool skills (Seoul Economic Daily).
  4. Pick a broad degree (CS, math, physics, engineering) over a games-specific one, and keep the cost down. This is opinion; I found no outcome data comparing the two. CS still pays for those who land a job: recent CS graduates show about 7.0% unemployment in the New York Fed's by-major table, but 19% underemployment and an $87,000 median early-career wage, against 5.2% and 42.5% for recent graduates overall last quarter (NY Fed). The by-major figures are 2024 data, so the comparison is rough.
  5. Don't treat indie as the fallback. Steam took in roughly 20,000 new games in 2025, and third-party estimates of the median release's gross run from about $250 to about $1,700 depending on method and year (Steam Page Analyzer, Levvvel, Mezha).

None of this closes a door, so he can re-decide at 17–18 with two more years of data.

What's viable in general

I can't give a list I'd defend for someone starting work in the 2030s. Today's official picture:
  • Healthcare: Nearly a third of all new jobs projected through 2035 sit in the two healthcare occupational groups, and the single fastest-growing occupation is nurse practitioner at +41% (BLS).
  • Skilled trades: These land almost entirely in the low or moderate tiers of BLS's new AI-exposure ranking, 45 of 47 construction occupations by a homebuilders' association count (NAHB). That ranking compares occupations with one another and doesn't predict which jobs disappear (BLS).
  • Software: still on the growth side, per the BLS figures above.

Speculation on my part: the common thread is work that needs physical presence, a license, or a legally accountable person. Whether robotics erodes the first of those during his career is unknown.
 
Unfortunately, not always. With pre-recorded tracks, sequencers and autotuners even a live event can be as fake as an AI product.
Whilst I do think live concerts with real human musicians will be an important connection point in music against AI, the use of miming & auto-tuning is a real problem. Auto-tuning is even being applied to back catalogs of artists like the Beatles.

There is currently a bit of a controversy in music circles with "MIDI Vinilli", a bass player who seems to have been faking his performances for years. The YouTube channel 'Wings of Pegasus' is a popular 'live' music debunker.

 
I suspect this is mostly due to copyright issues. It doesn't do for a publisher to sell content that anyone else can appropriate freely.
Absolutely. Also there are legal ramifications with aspects of fraud, breach of contract, not to mention reputational damage. It's probably going to go to court.
 
People are watching *performances*
That's a form of theatre then. Which would be fine with me if thats what I know in advance I'll be getting.

Most of the objections seem to come from the audience's expectations that they have paid good money (and some) to see an artist perform live - warts and all. Not quite hitting a note are what gives music its human soul. But, egos exists, marketing and image exists, artists may be under pressure (or as you say actual contract) to be seen as 'perfect' at all times.

the actual liveness doesn't really seem to be that important a quality
For some maybe, for others it's critical to both audience and artist - perhaps genre specific.

Earlier this year I was photographing a concert by two Austrian jazz musicians at a small venue in Edinburgh - you could grab a pint from the bar and literally walk into the performance area and listen close up if you wanted to (not advised obviously). The direct live sound was almost as loud as the sound reinforcement in the venue (low level fill for the rear of the club), so nowhere for any performance shenanigans to hide. I think the audience would walk out otherwise.

It's a bit of an insult to audiences that reject AI music to then support what you believe to be real live music for it to be basically theatrical mime. [*]

[*] There are times when pre-recorded backing tracks or pre-programmed electronic music instruments make sense as part of the music when they are part of the artists work - in line with fans expectations.
 
hat's a form of theatre then.
I know what you mean, and you are right, but there is an issue in that the word "theatre" (or "theater" as it is normally spelled in American English) has multiple meanings. Leaving aside the less relevant ones like "a room for surgeries to happen in" or "territory in which a war is being fought," I'm thinking of the art form of live dramatic performances on stage and of the way you used it, "something fake-ish done for effect or attention," more or less.

The art of theatre is not going to be taken over by AI in the way movies and television might be, at the least you need live actors for live theatre! Modern theatre does incorporate more tech-related tools in the production, projections and lighting effects and the like, and AI might intrude into some of that. But unless/until we get AI powered robots that pass for human, the stage actor has a job as long as people value the product.

Screen actors have a fight on their hands, though. While the Tom Cruises are reasonably safe (people want to see THE Tom Cruise) nobody cares much to see the guy hired on a day-player contract to play Deputy #2, with one line and no staple needed for his copy of the script, if that role is an AIctor, as long as it is not too intrusive I doubt folks will care. Look at how readily folks accept impossible CGI stunt work, with the CGI performer smashing through violence that would not only kill the stunt performer, but would be absolutely unsurvivable by the character (which to me is boring, as opposed to the old school daring-stunt-performer risks life and limb to accomplish the not-quite-impossible.) That sort of stuff makes suspension of disbelief very difficult, for me, and not-quite-right AI actors would do the same. But, if AI gets past uncanny valley type issues, we may be in trouble in the screen acting biz -- and if the day players and bit parts actors go, where does the next generation of stars get their chance to break in and learn the craft?

(In passing, by the time I was in high school many theatre-folk in the US had switched from the "-er" to the "-re" spelling, to set what they do apart from other meanings of the word -- and possibly to seem just a bit more pretentious...)
 
A few thoughts

In the same way that inexpensive "reality shows" displaced scripted programming across network TV, AI generated content is most likely going to displace more expensive human generated content across most entertainment channels. All commoditized business models follow this path. The current incarnation of the Internet has swept away all other barriers to entry making profit margins the metric that determines who survives at scale and who does not.

Many people with actual talent will continue to try making meaningful content. Finding them will depend on algorithms now hidden within the AI generated search results, AI mediated advertising, and the recommendations of friends who may not understand how many layers of machine interactions stand between what they want and what they get. The human creators will increasingly struggle to make a living as artists.

You might be able to circumvent cheap AI content via a subscription model where the creators, publishers, and consumers collaborate in deciding what artistic qualities do or do not belong in their particular community theater, local radio station, or other venue. The result is likely to be bespoke experiences that only wealthy patrons can afford. The mass market will go to the lowest cost, ad supported channels.

Regulation is not likely to save us. Most of the emails sent every day are still spam. We did not get rid of robocalls, we just stopped answering unknown numbers.

Not trying to be alarmist but economic history strongly suggests that as the 'cost of compute' continues to fall, the only things you can be certain are physically real are those sharing your immediate physical space. At least until the humanoid robots arrive.

/rant
 
My daughter is 15 and considering what she wants to do for a living. She's a very good artist and loves graphic design and animation, but I feel like that whole field is rapidly becoming, if not obsolete, at least a shadow of what it was until recently. I'm torn between wanting to encourage her and wanting to shout "No! Don't do it! Train to be a plasterer or a plumber or something!"

FWIW, she is very anti-AI, and so are most of her friends of the same age. An encouraging sign of resistance or just youthful idealism?
Exactly the kind of situation I was once in, but my father told me he wouldn't pay for me to study art in college because "You have to work in a field where you can support yourself". I hated the fact that he was right ...but he was right. A degree and a job in chemistry did NOT keep me from painting and doing art shows on weekends, and financial security is something I'd never want to be without.
 
I have a 15-year-old nephew, and for several years he's wanted to be a video game programmer. I don't really see that as a viable job now.
Although corporate game development is shrinking rapidly, indie game development is booming. He should consider pursuing it as a hobby rather than a career path. That way, he can do what he enjoys while supporting himself through a career he likes as well.
 
In the same way that inexpensive "reality shows" displaced scripted programming across network TV
That comes as news to me. Reality TV always seemed so contrived that it had to be deliberately scripted:
External Quote:

Reality TV Shows That Are Actually Scripted

Here are 15 "Real" Shows Producers Don't Want You To Know Are Scripted!

15. House Hunters

... However, it was reported back in 2010 that the participants on the show already knew which house they were going to purchase before filming even started. In fact, having already closed on a house was a requirement to do the show!

14. Storage Wars

... However, according to David Hester (one of the show's most popular contestants), most of it is staged. So much so that Hester has filed a wrongful termination lawsuit against he show back in 2013. He alleged that the producers would purchase multiple units in bulk and then plant valuable items into the ones that their characters won. Likewise the producers have been known to egg on the contestants into creating drama, including a case in which a fight between two of the contestants turned into an all-out brawl.

13. The Apprentice

... While the competition part of the show was shown to be real enough, the "firing" segment of the show was scripted for dramatic effect. In 2004 it was revealed that none of the contestants were ever truly evicted. Instead, they simply had to stay in a different section of the tower with the rest of the season's losers. In fact, the contestants' suitcases that they took with them were usually empty (save a few necessities), and the elevator ride and taxi transportation were purely for show.

12. Catfish

An episode will always feature the crew confronting the catfish and expose them for who they really are. ... Allegedly the producers of the show run the application process in reverse; the people who are doing the catfishing actually are the ones who apply to be on the show!
-- https://screenrant.com/reality-tv-shows-scripted-fake/

They even know that we know that there's bugger all "reality" in most "reality TV", and have started to weasel their way around that fact - there are now Emmy and Critics Choice awards specifically for "Structured Reality" Programs:
External Quote:
In 2016, the category Best Reality Series was separated into two categories – Best Unstructured Reality Show and Best Structured Reality Show.
https://en.wikipedia.org/wiki/Critics'_Choice_Television_Award_for_Best_Structured_Reality_Show
External Quote:
In 2014, Outstanding Reality Program was separated into two categories – Outstanding Structured Reality Program and Outstanding Unstructured Reality Program.
https://en.wikipedia.org/wiki/Primetime_Emmy_Award_for_Outstanding_Unstructured_Reality_Program

So, what about the "unstructured" ones, you ask? Well, according to that Emmy page and https://en.wikipedia.org/wiki/Critics'_Choice_Television_Award_for_Best_Unstructured_Reality_Show many of the shows listed are the ones in the screenrant and other such "it's fake" listicles. So even "unstructured" doesn't mean "unscripted", just less-heavily-handedly-scripted.

I think. I don't know. I've never seen even one episode of any of the programs mentioned in any of those articles apart from /Top Gear/ and /Bullshit!/ who won some awards - but are they even "reality TV" anyway?
 
I'm thinking of the art form of live dramatic performances on stage and of the way you used it, "something fake-ish done for effect or attention," more or less.
Yes, 'Performative' was what made me think of that word. Musicians acting out a performance rather than *being* the performance in the moment - if that makes any sense to anyone apart from myself. Nothing deeper and maybe not the best word to use.

The art of theatre is not going to be taken over by AI in the way movies and television might be, at the least you need live actors for live theatre! ... But unless/until we get AI powered robots that pass for human, the stage actor has a job as long as people value the product.
Good point. I think some productions have used some sort of 3D holographic tech (as part of the story telling) to add actors into a live scene (and IIRC the off-stage actor was live at the time). Otherwise projecting actors is called Cinema :)

But, if AI gets past uncanny valley type issues, we may be in trouble in the screen acting biz -- and if the day players and bit parts actors go, where does the next generation of stars get their chance to break in and learn the craft?
I'll remain hopeful that auteur filmmakers will still respect human craft. And there should be theatre [1] work to hone those acting skills as/if the movie industry contracts.

[1] We spell it the right way here! :)
 
A degree and a job in chemistry did NOT keep me from painting and doing art shows on weekends, and financial security is something I'd never want to be without.
Sound.

When my daughter was younger she really got into computer animation & graphics. I remember saying something like 'If you learn Photoshop you will have a job for life'! Boy, was I wrong. And this is what - 10 years ago, an eon in technology years. She became a stylist, so until walking, talking AI robots take over that job, she should be ok for a bit.
 
Back
Top