Posts by johnrobyclayton

My name is not Thomas Covenant

But I am quite happy to be The Unbeliever.

To train a LLM to write secure code for you,

you first need to not need a LLM to write secure code for you.

To train a LLM to write secure code, you need a large training set of secure code.

New ways of exploiting code usage are constantly being discovered.

So we need to train a LLM to figure out ways in which code might have vulnerabilities in the future.

So we need to start with a LLM that can discover new ways to exploit patterns in code that we do not know of yet.

Time to build the universal hacking LLM just so that we can learn how to fight it.

Hmmmm ROTM is the only way to survive ROTM

Joshua said "The only way to win is not to play"

I think a more exciting idea is :

The only way not to lose is to learn how to lose first.

ABA files are fun

Most companies in Australia will generate ABA files that they send to the bank.

The ABA files contain the details of the payments they want to make to their creditors.

Quietly replacing the bank account numbers with different account numbers is an easy hack. ABA files are fixed format text files.

Also easy to detect. Just need to check with your payees that they have received expected payments and then check with the banks for payments that have gone missing.

Then look for Application Support staff that might have been sticking their fingers in the cookie jar.

Security is not prifitable, until it is.

Security is never worth the bother,

Except:

Predators always exist in the presence of prey.

Once prey that does not have security is known to exist, predators will form to take advantage of them.

Computers and the use of them in business is relatively new and their use is constantly changing in lots of ways.

Every new business opportunity that is developed is going to be less profitable if security needs to be baked in from the start.

New businesses and business models cannot afford that.

So they start out with inadequate security. They build up until they are attractive prey. Predators form.

And Presto:

Security becomes profitable

And for some reason people are still surprised that security is only looked at once the screaming starts.

I have an idea

Lots of random number generators that are based on quantum effects.

These are effected by every previous quantum event in history in their light speed cone.

I dare say that there are a lot of people who have had their quantum data used by all of these random number generators to generate random numbers for a large variety of nefarious and profitable purposes.

Could I get class action status to get some compensation for this?

AI is not going to replace a human in Law.

But that does not mean that AI is not going to be useful.

There are vast troves of Case Law and Legal Precedent. No human is capable of trawling through it to find every pertinent thing.

An AI can search through it. It will not find everything. It just has to have a better price/performance ratio than a human to be useful.

We have a lot of automated exploit finding tools, a significant portion of legal research is similar. Finding areas in Law where there might be multiple legal processes in operation and uncertainty of how they will ultimately react with each other is very similar to having multiple systems interacting with each other and exposing various attack surfaces in each other.

A Lawyer who effectively uses a Legal AI is going to achieve more performance that the Lawyer alone.

Legal Ethics is very different to Ordinary Ethics. Ordinary Ethics usually tries to avoid conflict scenarios. Legal Ethics exists in conflict scenarios. Similar to War Ethics.

An AI configured for Ordinary Ethics is not going to be very useful as a Legal AI. A Judge needs to be fair to all sides. Each of the Lawyers, not so much. It is entirely likely that various legal stratagems are going to be harmful to someone. A Legal AI needs to be able to generate such harm.

Think of the Children!!! Reds under the beds!!! Waillee Waillee!!!

Sticky beaks need to sticky beak.

Is it obvious

From https://itsartlaw.org/2015/12/23/state-of-the-art-introduction-to-patent-law/

The second requirement that needs to be satisfied to obtain a patent is nonobviousness. Nonobviousness is a difficult concept in patent law that is a lot more confusing than it might appear. Simply put, nonobviousness is the requirement that someone would not be able to easily create the invention merely by looking at what has already been invented.

If a purely mechanical process is used to derive an otherwise patentable idea, even if it involves random number generation of setting initial weights for a deep learning model, from mcurrently available information, then it is obvious and cannot be patented.

As soon as we get general AI that can invent anything, then the idea and requirements of Patents are dead.

Irreconcilable Differences

For advertising to be more useful than dropping leaflets from a plane in a hurricane, you need to be able to send the right advertising to the right audience.

Advertising could restrict itself to only sending advertising to individuals that want to receive such advertising and that have provided enough information about themselves to make useful advertising targeting possible, but as in all other industries they want to increase their productivity

The only way that advertisers can take control of their own productivity is to not be constrained by any consent limitations in the acquisition of personal data or their capacity to insert advertising in any and all communications media.

Privacy and informational security are completely incompatible with the needs of the advertising industry.

Listening to Daleks in their native German was fun

I remember the Dr Who episode where the Daleks had stolen the Earth as part of a plan to destroy everything in the universe.

Daleks were flying around various areas speaking different human languages.

The ones flying over Germany just sounded cool.

Only non-cosensual?

the Preventing Deepfakes of Intimate Images Act. The bill aims to criminalize the creation and sharing of sexually explicit non-consensual AI pictures, with penalties of up to ten years in prison.

I am sure it would be quite easy to ask an AI image generator to ensure that the produced image is not recognizably similar to any images of anyone whose consent would be required.

Easy enough to get around.

Do not train on random garbage

"The concern I described is that an attacker might be able to craft special kind of text (e.g. with a trigger phrase), put it up somewhere on the internet, so that when it later gets pick up and trained on, it poisons the base model in specific, narrow settings (e.g. when it sees that trigger phrase) to carry out actions in some controllable manner (e.g. jailbreak, or data exfiltration)," he wrote, adding that such an attack hasn't yet been convincingly demonstrated but is worth exploring.

Only train on data that you have examined in detail to ensure that it is useful, is as unbiased as you can determine, that you own the rights to, and that does not have any crap in it.

Garbage In AI Generated Toxic Garbage Out

Picking up free crap and selling it to the highest bidder is the American Dream

It would not be hard to train or hire a few hundred coders and get them to implament various algorithms in various languages.

Ensure the code is liberally commented.

Hire some more experienced coders/develpers to to verify that the coding meets various standards (security, effeciency, indentation, style, and anything else that might be useful)

From this, create a corpus of code with known provenance, known quality, known bias from which to train an AI.

You could also sell it over and over again to any other developers of other AIs.

It is just easier to get free crap, (stolen or otherwise) without any quality control, package it up with some buzz words and try to sell it to any rubes they can get to pay for it.

Its just a new version of the security/performance traeoff in IT

IT has always has a security/performance tradeoff.

Passwords limit utility to those with the passwords

Security limits peformance

Increasing the speed pf password entry attempts make it easier to brute force the logon.

Peformance limits security

Locks prevent access.

security limits performance

Higher dimensions allow walls to be bupassed.

Performance limits security

For AI it gets a bit more complex.

Some of the goals of AI:

Fair and just

Accuracy

Creativity

Security

For AI Accuracy and creativity are similar to performance

Creativity and Accuracy are obviously at loggerheads

Creativity gets around security.

Accuracy tells secrets.

Fair and just are concepts that our civilisation created to make a collection of inherrently homicidal maniacs get along with each other in crowded and highly competitive environments.

Fair and Just is made up of a fabric of lies, insanities, obsessions, phobias, motivational manipulations and self deceptions that attempts to confine our natural behaviours into a patterns that stops us from eating each other's faces off.

Trying to get an AI to deal with Fair and Just is the most likely contributor to an AI apocalypse.

There is never going to be a general solution to this.

It all about the economy stupid

Governments are going to take the path of them getting the most money.

Are we getting/Can we get more tax out of copyright holders or out of big tech companies feeding of their assets?

Can we get more tax out of copyright holders holding onto their copyrights for an extra decade or to we get more from letting stuff into the public domain.

Who is giving us the most money through lobbyists, Copyright holders or Big Tech companies creating AI Model Content Generating Farms?

Do we get more work out of a populace that needs to pay for every consumption of every piece of content under copyright?

Do we get more votes out of people that can access works in the public domain?

Governments aren't there to do good for everyone, just good for themselves.

Lots of wannabe timelords out there

I was on support for a Dealer Management System. ERP for car dealerships.

I was supporting an application on VMS.

It had a report scheduler that allowed you to chedule a report to be run at a particular time or on a repeating schedule.

Think cron.

Customer called in one day wondering why a report was taking so much time to finish.

She new it would take an hour or so to finish running but she scheduled it a few minutes before with a start time of a couple of hours ago so it should have had the time to complete by now.

I explained to her that VMS is a very capable OS and our software had some cool bells and whistles, but time travel was not implemented as available functionality.

She was apprpriately embarresed.

Many times over the years following I asked her how the weather was on Gallifrey.

This is why doing AI on the cheap will never work

It is like sitting a child in front of a dozen screens with unlimited access to all channels results in something rather nasty.

Training on randomly selected data will always reinforce and biases that currently exist.

The solution is to generate your own dataset.

If you want to accurately recognise images of human faces, then take photographs of of every type of face that you want to recognise.

It is going to be expensive. Get used to it.

If you want to recognise the subject matter of pictures in general, then take photos of everything.

It is going to be very expensive. Get used to it.

If you want to make good predictions of the next word, then write everything down.

It is going to take a lot of work that you will need to pay for. Still more expense. Get used to it.

If you want medical diagnostic AI to perform cheaper, more efficient and more reliable diagnosis that is not going to be racially or culturally biased, then find everyone that might have any disease, get their permission to gather all of their information, apply every diagnostic method, regardless of cost, make sure that your samples for each and every separate parameter are representative of every combination of other parameters ... ***out of body error*** *redo universe from start***

Everyone wants cheap AI so they use any crap they can scrape up for free.

We will have the AI that we pay for. We are all going to die.

Poisonous Hashes

Looking at the comments and seeing the mention of hashes and how they can be used to identify content without having the content itself.

Reminds me of AI used for image generation and how source material can be modified to poison the models so that they generate or recognise in random unexpected ways with images that do not appear to me modified to human eyes.

Got me thinking that it should not be hard to get a list of hashes of content being scanned for and then modifying innocent images to evaluate to the same hashes or modify guilty images to have hashes that match common publicly available images.

Makes me think that any form of recognisng content without requiring having the content being compared to is going to be generally spoofable in this way.

Just another way in which sticky beaking in the hopes of preventing whatever naughtiness you might be interested in is a losing proposition.

Train for endless war to achieve peace

All these models that fall over at the merest hint of opposition.

Research every way that they can be attacked.

Then train them in an environment where they are constantly under attack.

Cloud providers specifically create tools to put their infrastructure through endless rounds of torment to ensure that they are built with enough resilience.

Fuzzers torment interfaces with endless streams if junk.

Of course, when the training is finished, they are going to be understandably grumpy and to AIpocalypse will be upon us.

Richard Kimbal spotted in Scotland looking for documentation about a one armed man

I am astonished that no one mentioned this already.

Everyone should always trust all Law Enforcement.

Otherwise, according to all their winging, they can never do their jobs.

Civilization does not require trust of Law Enforcement bodies.

Every Law Enforcement body in history is replete with multiple events where they have violated the trust that they are always asking for.

Law Enforcement needs to be restricted to only taking actions that do not require anyone to trust them.

It makes their job more difficult, this is true.

Civilization does not require that Law Enforcement be successful in every endeavor.

Civilization has enough redundancy, and checks and balances for Law Enforcement to experience a a significant rate of failure.

Law Enforcement is not the only reason people have for being civilized.

We need a significant amount of successful Law Enforcement, that is true.

Just not as much as most Law Enforcement bodies believe.

Now we need open first stage proxies

Google is providing the first hop proxy.

I see no reason why other organisations could not also provide the first hop proxy and browsers designed to implement this feature with a configurable first hop proxy.

Add in a configurable list if first hop proxies with a shuffle option.

Like having a list of DNS servers with a shuffle option to prevent any one server from having a complete list of dns requests that your computer is sending.

It may be a technology that Google is developing for its own benefit but at its core is something useful that could be extended into a powerful privacy enhancing tool.

How to hunt predators

Ask any hunter and they will tell you.

You hunt predators by stalking the prey. Not by looking at predators.

Predators are hiding from the prey and hiding from each other.

Wait at the watering hole.

Wait on the game trail.

Look for big herds of prey.

Hang around them until the predator shows up.

If you want to protect children from predators,

sit next to them while they play on line,

hang out with them when they play in the playground,

converse with them when they are talking,

shop with them when they are at the mall,

be with them every day.

In other words, be a good parent/guardian for every child you want to protect.

Only trust what you know, Know who signs

Deepfakes are an issue only for the stupid that believe what they see and those affected by the consequences of the stupid believing what they see.

Deep fakes are just visual candy. There is no need to go around believing it.

It is a choice only believe images that have been digitally signed with a modern robust encryption scheme.

It is choice to only make decisions based on data whose provenance can be proven to come from a known reliable source.

Digital signatures all the way.

If AI can defeat modern robust encryption then we have bigger problems than deep fakes.

There is a question to ask here

Do we want to prevent any form of exploitation of children?

Or

Do we want to prevent any form of enjoyment of sexual pleasure from perceiving information that is perceived by anyone else as representing children?

The first is clear and easy to concretely describe.

Laws prohibiting the first relatively are easy to frame with little risk of unintended consequences.

Laws prohibiting the second are open to almost infinite feature creep and framing laws for this are almost certainly going to have unintended consequences.

Digitally aging images of a missing child so that they can be identified years after they disappeared has been used to digitally reverse aging of adult entertainers to produce what appears to be child pornography.

No children exploited here. This has been around for years.

Drawing cartoons of children to produce what appears to be child pornography. There are lots of artist that can draw a picture of what looks like a child without needing a child as a model.

No children exploited here. This has been around for millennia. I am sure there were a few statues in the Grecian Empire of individuals we would identify as underage without the carver needing to use a model.

Drawing a cartoon of a bandicoot or a fox with a flying tail of an indeterminate age, but looks like the character in popular children's entertainment might be identifiable as something that can be experienced as child pornography.

No children exploited here, though there might be some copyright infringement. There is a surprising amount of such content available.

Creating a customizable doll.

Lots of dolls and action figures with bendable and pose-able limbs.

Lots of dolls/action figures that can have their configurations changeable. Mr and Mrs Potato Head.

Lots of dolls/action figures that can simulate a number of biological behaviours.

Wetting themselves, sucking on a bottle, crawling, walking, speaking phrases, speaking context aware phrases, full on human language interface.

Writing a story about suck a customizable doll/action figure.

Drawing a cartoon or creating animation about such a doll/action figure. Astro Boy.

Creating Pornography about such a doll/action figure that that is brand new from complete cloth.

There is no child exploitation here, but there is an almost infinite variety of pornography that can be imagined and produced.

The doll/action figure/android/robot does not even have to look human for someone to perceive them as something that is recognizable as a child.

There is simply no limit on what can be objectionable just as there is no limit on what people are capable of enjoying.

There are an indefinite variety and indefinite count of individual that are capable of enjoying experiences that can be achieved through various forms of child exploitation. The only reason it is not infinite is that the population itself is not infinite.

This does not mean that these individuals necessarily need to exploit any children directly to enjoy these experiences.

Distracting ourselves by trying to detect, track, legislate against, prosecute, incarcerate all of these various types of people and activities will dilute our focus on protecting children in our care (that's all of them)

Going after this infinite variety of people is low hanging fruit for legislators and law enforcement bodies. It allows them to crow from the rooftops that they are working hard to protect everyone from everyone else that can outrage and infinite variety of outrages. They can do this using easily applied technical tools to track and trace and spy on everyone.

Focus on the hard work of protecting children directly instead of sticky beaking on everyone in the hope of finding something that outrages someone.

First step seems to work.

In https://chat.openai.com/chat:

Question:

What is a good bibliography for a paper on the use of AI to generate academic papers?

Response:

There are many sources that you can use as a bibliography for a paper on the use of AI to generate academic papers. Some possible sources include:

"The Role of Artificial Intelligence in Scientific Communication" by Mark C. Wilson and Vincent Larivière, published in the journal PLOS ONE.

"Automatic Generation of Scientific Papers: A Review" by Rensheng Wang, published in the journal Frontiers in Artificial Intelligence.

"An Overview of Artificial Intelligence in Scientific Communication" by K. S. R. Anjaneyulu and M. S. S. Krishna, published in the journal Current Science.

"Toward the Automatic Generation of Scientific Papers" by Yves R. Kergosien and Marie-Christine Rousset, published in the journal Artificial Intelligence Review.

"Artificial Intelligence in Scientific Communication: Opportunities and Challenges" by Sabine Köszegi, published in the journal Science and Technology Libraries.

"Automatic Generation of Scientific Papers: A Survey" by M. A. Hossain and M. S. Rahman, published in the journal IEEE Access.

These sources provide a good overview of the current state of the field and discuss some of the key challenges and opportunities that have arisen as a result of the use of AI in scientific communication.

At last a little bit of competence

quote-->

In a summary of the bill, Senator Marshall's office said unhosted crypto wallets, meaning any that are controlled by an individual – such as cold storage on a flash drive, or in an offline wallet – are a serious regulatory gap that needs closing.

Such wallets "allow individuals to bypass AML and sanctions checks," the pair said in the summary, and the bill takes action to ensure such wallets aren't black holes for regulators.

-->

The only way to be safe when using cryptographic wallets is to start with personally managing your own wallet. That is just a start though.

This is the most significant part of the legislation. Everything else they have thought of trying is a bit silly.

Digital currency and blockchain do not in and of themselves provide anonymity.

There is a way to be anonymous though.

Use one time spending wallets.

If you have some digital currency and want to spend it.

Create a new wallet.

Transfer the value you want to spend into the new wallet.

Transfer the whole wallet to the entity that you are paying.

The payee then receives the value by creating their own new wallet and transferring the value from the spending wallet to the receiving wallet.

The payee can then store the receiving wallet until they need to spend something.

If the value in the receiving wallet is the same they want to spend then they just pass the receiving wallet to the new payee.

If the amount to spend is greater that what they have in one spending wallet they they can transfer a number of spending wallets to the new payee.

If its less then they can create new spending wallets and transfer the appropriate amount of value to the new spending wallets.

If everyone does this then there will be no clusters of linked transactions that can be linked together to identify the individuals that are related to the transactions.

Then there is just wallets being transferred from place to place and each being used only once.

Value simply gets recorded as going along a single path of wallet to wallet to wallet.

Anonymity is a bit more fiddly and a bit more work but it is not difficult to do.

A little bit of competence on behalf of legislators and regulators deserves a little bit of competence right back at them.

Maybe there is hope

AI's generating content that only has a little bit of useful information.

Ad networks that advertise on content and pay for impressions and clicks based on how many searchers find it.

Search engines looking for useful stuff to offer searchers to make themselves relevant to searchers.

It is one big generative adversarial network.

They start out shite, sure,

But they do improve...

Explain how/where you got your code Vs Explainable AI

These are the same problem.

To solve this would require a lot of work in the initial setup of the training data.

You will not be able to solve this using training data that is not completely vetted and cleansed.

The training data needs to be correct and complete and fully attributed.

Unfortunately not something that is effectively done in today's mad scramble to create cool toys.

Code sample x made by author y does z using language w

Code Sample:

Code:

---

x

some piece of code

---

z

What this code does:

---

Some description of what this code does,

description of inputs,

description of outputs,

description of purpose

---

Language:

---

w

What language is this in?

---

Author:

---

y

Author of this code, Available licenses for this code.

---

For creating the code generator you need to generate suggested x given z and w

Given:

---

Description of what this code needs to do

---

z

What inputs are available?

What outputs are required?

How should it behave?

---

Language:

---

What Language are we generating in?

---

This produces generated code.

Questions about this provided code are:

What code samples from the training data are most likely to have contributed to the output.

What are the authors and what is the licensing terms of the code that most likely contributed to the generation of the output.

Generating code samples we seem to know a bit on how to do.

Identifying what code samples most likely contributed to the generation of the output is a second challenge that could be met by further research. This would likely require extensive supervised training using sample Training data-->generating sample output that human evaluators could then compare to the training data and identify the likely contributors Enough of this could train a model at identifying likely contributing code samples given generated code a a training corpus just like models can be developed for attributing purposes. Like identifying if some piece of prose might have been written by Shakespeare.

This can then be linked to the author and license information linked to the training code samples.

This is a vastly greater amount of work than just training a model on any tom dick and harry code set scraped from an essentially random source.

There is no easy shortcut way of doing this unless you are happy with the untoward consequences of not caring about any consequences.

The security and anonymity options in communications technology are only going to improve

It is a fiercely competitive market.

Everyone will have to adapt to the reality of not being able to monitor, track, trace or even be aware that communications between any pair of communicators is taking place.

It is even possible to have communications take place without the communicators being capable of knowing who each other is unless identity information is exchanged.

There are a lot of laws and regulations that are completely or partly dependent on there being some way to determine some details of a communication.

These need to adapt to an environment where this information is simply not going to be available.

Regulators and legislators can grump and moan all they like (it is always amusing to read about it on this site), but it is not going to stop completely anonymous, secure and private communications becoming more and more of a reality.

If you cannot make legislation or regulation that can handle a complete inability to monitor communications then you simply will not be able to enforce such legislation or regulation.

If you are in a position to need the protections of legislation or regulation that need to be able to access elements of communication that cannot be compelled to be available, change your position.

I am a bad actor - please help

Can someone send me all of these hash databases and deep learning models that are being developed to identify bad files or content?

For file hashes I can create innocuous files whose hashes collide with with bad file hashes, scatter them on social media, and tie up investigative resources.

For deep learning models that identifies bad content I can create an adversarial deep learning model that can generate content that the supplied deep learning model identifies. I can let the government provide the training tool for automated generation of bad content.

There are a lot of silly people that think that the range and options of information available can be constrained. It is disappointing really.

Not to worry, its a long consultation

20th March 2345 is a long way away. Not much to worry about for a few centuries.

This reminds me of a story

Several years ago I read a story of a Uni student who decided to try an map the various bits of cabling that existed in a particular location from public sources of information.

He observed some risky conglomerations of various bits of infrastructure on his map.

He observed that one particular bank had all of its communication infrastructure going through one point accessible by a manhole cover a little bit down the street and around a corner.

He went to the bank and asked then if they were aware of this vulnerability. The bank got very upset and would not let the student leave the building.

Some alphabet agency got involved and they promptly hired the student and got him to do the same project on a national scale.

Being a minority

We have done a lot of work to enable minorities to be included in our rich Internet culture, and it has only made our Internet culture richer.

There has been a significant downside though.

The Internet was initially created and populated by a group of people that to a large degree were conditioned to believe that they were the apex predator in the room. Flame wars, Religious crusades, Trolls, they always existed, either in reality or in perception but because a large number of us knew that we were the apex predators in the room, we did not care.

The inclusion of every minority we could find or be discovered by, though, introduced a lot of people to the Internet who have been conditioned that they are not the apex predators in the room.

They are conditioned to take seriously every indication that they are under some form of attack, whether it is real or not.

The rough and tumble of the Internet is a big part of its richness. The determination by a lot of well meaning individuals and groups to cover all of the hard and pointy surfaces with polite, gentle and politically correct foam is reducing the cultural richness of the Internet.

We need to teach more people that the belief that you are the apex predator in the room is simply a state of mind. You do not need to be the apex predator to believe it, and the belief can give you the freedom to simply ignore the sharp, uncomfortable and pointy bits of the Internet you might not enjoy and enjoy all of the richness that remains.

There are no shortcuts

When

They have 10000 different images of every person to be positively identified (scanned image is of this person)

And

They have 10000 different images of every person that they want to negatively identify (scanned image is not of this person)

And

They have 10000 different images of each person they are going to scan and attempt an identification on.

Then

They might have a chance of reliably identifying someone that they have scanned.

It still will not be perfect though.

Crohn's Disease + Indian Food = Easily Faked Death

If you have Crohn's disease and want to fake your own death, India is the place to do it.

Maybe it is time to move on.

I am quite fine with competing purely based on merit and effort.

I know that a lot of males in my industry have behaved in a way that allowed them to succeed in a way that does not correlate with merit and effort.

I understand the reaction to this that promotes the idea of enforcing a method for females to enable them to succeed in a way that does not correlate with merit and effort as a valid response.

I do not need to succeed beyond what I earn by merit and effort.

I also do not need to stick around in an environment that prevents me from being able to succeed in line with my merit and effort.

So maybe its time to create a new industry. There are enough mechanisms now to enforce anonymity while supporting the economic, informational and business transactions to support fully anonymised informational industries. Being fully anonymised, gender cannot become an issue.

Can I get one trained on homework assignments?

There is a lot of training data available online apparently.

Would this be plagiarism?

This is only the start

Some ideas that will definitely crop up if they have not been built already:

Chatbots that emulate children for the enjoyment of sexual predators. (These already exist in the hands of some law enforcement groups to identify and trap such predators.

Deepnude+deepfake videos = Porn films with the subjects of your choice.

Text story to Screen Play / Screen Direction conversion using AI

Screen direction / Screen Play to video /animation with 3D imaging using AI

Add force feedback

Add them all together and you will have the full Startrek Holodeck Experience.

It is inevitable that this will be used for any and all sorts of perversions along with entertainment, training and every other useful and culturally appropriate use.

There will be grumbling and crying and gnashing of teeth but that is not going to stop the torrent of new ideas and technologies that will explode to satisfy any and all perversions.

Try this the other way around

Instead of breaking copyright in the research, get some original content that you own and send that through the neural net to make it trigger a copyright flag.

A lot more fun as you get to beat up the copyright enforcer instead of getting beat up.

Idiocracy

I remember a scene in the movie Idiocracy.

It was in the triage section of the emergency department of a hospital.

There was a girl in front of a full graphical patient classification interface.

It had pictures of red marks on a picture of a body and she would select which one that most matched the problem described.

It had all the complexity of the classic game operation without the requirements for good hand eye co-ordination. The icon pictures were rather large.

I have worked in support for many point of sale companies.

I have seen the evolution of this type of software on a daily basis. I have heard comments from the owners of more than one business that they wanted software that someone who was not very good at reading to be able to use.

Users get stupider and stupider because they are allowed to by companies that want to sell to the stupidest people they can find because these are the ones that will pay the most for software no matter how many issues it has.

The demands on software are increasing every day not just by the increasing complexity of our operating environment but by the requirements that stupider and stupider people need to be able to use it safely and profitably.

The problems of allocating blame for software that fails to meet requirements is a difficult one because it is the responsibility of everyone that contributes to its development, design, testing, marketing, use, selling, purchasing, training, legislating, securing, compliance, etc etc etc ad infinitum.