Bonum Certa Men Certa

Guarding and Rescuing the FSF Titanic: The Simplest Ways that AI will Change Computing

A publication from the Free Media Alliance

Overview



Unhappy feet

Summary: "AI is already used to help kill people. We should be cautious, and know that the best rules we come up with (like no doing magic outside the school grounds) won't be followed all the time."

ARTIFICIAL Intelligence (AI) enhances automation; one way to think of AI is "A lot more computing -- both good and bad." For art? Great. For surveillance? Sometimes bad. Apply it to everything -- people will. And it will be a great multiplier of things; of all computing tasks, more or less.



Not all at once. And this is not to hype it, but to describe the effect it will have -- as a multiplier:

Another way to think of AI is "enhanced computing." Because in many ways, it is fundamentally "just computer processing." Anything a computer does is "just computing." But with AI, this becomes something more; the scope of what can be touched with computing becomes richer -- for good and for bad.

Computing is very flexible, by design. We can actually say something about AI while being this vague -- it is essentially like computer processing, except that it can do a little more, it can do more with more modest requirements -- it may take a while -- but with home computing equipment you can suddenly do things that you would expect of companies like Pixar.

Certainly not at the resolution for a (feature-length) film like Pixar makes. They will still use large computing farms to get the job done in a reasonable amount of time, at least for now.

AI can possibly seem to violate Moore's law, but it won't violate the laws of physics. If we are doing 1/3 of what our CPUs can do, then AI will make it so we can do the other 2/3 as well. And we can be really amazed at the results.

"Because in many ways, it is fundamentally "just computer processing." Anything a computer does is "just computing." But with AI, this becomes something more; the scope of what can be touched with computing becomes richer -- for good and for bad."Also with "enhanced computing", things that once seemed incredibly difficult to program are now at least possible. Not necessarily "easy," but what once would take a team of 25-50 people (at least) can now be done sometimes with a team of 3. That's not a general rule, just that some things that once took many people can now be done with few, and faster than when it took more people.

Wizard-like stuff that once took a team can now be done by individuals. So the term "enhanced computing" is both telling and probably accurate.

If you want, you can say that what computers could do already 10, 20 years ago is almost like magic. We know better, but it still feels a little bit like magic.

If you think of Harry Potter -- Ollivander said of Harry's nemesis: "He too did great things. Terrible, yes -- but great." It wasn't a compliment, it was an accurate measure. Of course for a young boy who just learned he was a wizard, it's creepy enough.

AI will do great things. Some of them will be terrible -- but great. And hopefully more of them will be Harry-like than Voldemort-like.

But really, it will be both. AI is already used to help kill people. We should be cautious, and know that the best rules we come up with (like no doing magic outside the school grounds) won't be followed all the time.

No "Ministry of Artificial Intelligence" is going to be free of corruption or poor decisions -- nor would it be enough to stop all bad things that are done with or without approval. Either way, AI is here.

Perhaps the biggest difference between AI and human thought is the superficiality and bias. Humans have that sometimes, in very stupid ways, but we are more flexible. AI can magnify our stupidity -- think of the old adage about "knowing just enough to be dangerous." That's AI, and its potential to try to make computers do what we think we want -- and getting far worse versions on average.

That's going to be very common; even humans have done this now and again throughout history. AI will lead us to a greater capacity for such mistakes. Just as AI can solve things that would take 100 people to solve, it can make mistakes that would take 1000 people to create.

"Wizard-like stuff that once took a team can now be done by individuals."At least with laws, there's a judge and jury as long as it's not artificial. We are certainly building corporations that have more power than a judge and jury do. But AI could do that too.

Politically, AI lends itself to many things, but may lend itself best (or at least most easily) to fascism. Or that could be post hoc -- it's corporations and governments that are the most interested in it, so this could be describing what it lends itself to most easily by extrapolating it from the product of governments and corporations working on it. Still -- what we are developing now is like that.

People are trying to think of whether AI will be more good or more bad, and this is no argument for a neutral stance. If you look at all that computers have done both for our lives, and also to our lives, computing that is suddenly enhanced in ways that at least seem to go beyond the reach of Moore's law is exciting, but also justifiably scary.

What AI does is pattern recognition, and it can also impose patterns. This is said broadly because that's the broadness of the application -- you can find patterns similar to the way a person would, you can impose patterns similar to the way an artist would. Computers can do that without AI, but not at the same level as a person.

Today, we are designing software that can do those things faster and more tirelessly than people -- with similar (or sometimes superior) skill. Manipulating video, audio, tactile environments -- targeting, surveillance -- these are being expanded and developed all the time, not just in the future. AI may have future applications in sabotaging Free software.

Strips is a framework for creating project plans with AI. If given the outline of a project and a desired outcome, AI can be used to drive the project towards success.

"If you look at all that computers have done both for our lives, and also to our lives, computing that is suddenly enhanced in ways that at least seem to go beyond the reach of Moore's law is exciting, but also justifiably scary."If given the "desired" outcome of making a project untenable or fail, plans could be created (with or without Strips, it is just an example of a real AI planning framework and may have no direct relevance to this argument) to undermine or disrupt the viability of a business, organisation or Free software effort.

Computers have already been used for years to simulate and project outcomes of real-life processes -- the FSF has never done this, but it shouldn't surprise us if software monopolies do run such simulated campaigns.

There is an opportunity to do more testing of whether certain plans will help or hinder future efforts, with the very big warning that the previously mentioned examples of bias are still likely relevant, and engineering circular arguments that reinforce or negate the merits of a plan of action is not only possible, but could be difficult to avoid.

"Computing has always had good points and bad points -- it is very arguably not neutral, but it is nuanced."A positive of AI and AI-based planning could be to streamline and automate the creation of GNU/Linux distributions. This is about how the distro is put together, and may prove more relevant to building distros than say, package management.

The more that is done to reduce the work of building a distro, the more freedom the user will ultimately have. None of this is intended to paint AI as solely a threat, or solely a benefit. Computing has always had good points and bad points -- it is very arguably not neutral, but it is nuanced. The future is interesting, and not everything is hype.

Licence: Creative Commons CC0 1.0 (Public Domain)

Recent Techrights' Posts

Techrights' Statement on Code of Censorship (CoC) and Kent Overstreet: This Was the Real Purpose of Censorship Agreements All Along
Bombing people is OK (if you sponsor the key organisations), opposing bombings is not (a CoC in a nutshell)
 
BetaNews, Desperate for Clicks, is Pushing Donald Trump Spam Created by LLMs (Slop)
Big clap to Brian Fagioli for stuffing a "tech" site with Trump spam (not the first time he uses LLMs to do this)
[Meme] Social Control Media Bliss
"My tree is bigger than yours"
Links 24/11/2024: More IMF Bailouts and Net Client Freedom
Links for the day
Gemini Links 24/11/2024: Being a Student and Digital Downsizing
Links for the day
[Meme] The Most Liberal Company
"Insurrection? What insurrection?"
apple.com Traffic Down Over 7%, Says One Spyware Firm; Apple's Liabilities Increased Over 6% to $308,030,000,000
Apple is also about 120 billion dollars in debt
Over at Tux Machines...
GNU/Linux news for the past day
IRC Proceedings: Saturday, November 23, 2024
IRC logs for Saturday, November 23, 2024
[Meme] GAFAMfox
Mozilla Firefox in a state of extreme distress
Google Can Kill Mozilla Any Time It Wants
That gives Google far too much power over its rival... There are already many sites that refuse to work with Firefox or explicitly say Firefox isn't supported
Free (as in Freedom) Software Helps Tackle the Software Liability Issue, It Lets Users Exercise Greater Control Over Programs
Microsofters have been trying to ban or exclude Free software
In the US, Patent Laws Are Up for Sale
This problem is a lot bigger than just patents
ESET Finds Rootkits, Does Not Explain How They Get Installed, Media Says It Means "Previously Unknown Linux Backdoors" (Useful Distraction From CALEA and CALEA2)
FUD watch
Techdirt Loses Its Objectivity in Pursuit of Money
The more concerning aspects are coverage of GAFAM and Microsoft in particular
Links 23/11/2024: Press Sold to Vultures, New LLM Blunders
Links for the day
Links 23/11/2024: "Relationship with Oneself" and Yretek.com is Back
Links for the day
Links 23/11/2024: "Real World" Cracked and UK Online Safety Act is Law
Links for the day
Links 23/11/2024: Celebrating Proprietary Bluesky (False Choice, Same Issues) and Software Patents Squashed
Links for the day
Over at Tux Machines...
GNU/Linux news for the past day
IRC Proceedings: Friday, November 22, 2024
IRC logs for Friday, November 22, 2024
Gemini Links 23/11/2024: 150 Day Streak in Duolingo and ICBMs
Links for the day
Links 22/11/2024: Dynamic Pricing Practice and Monopoly Abuses
Links for the day
Topics We Lacked Time to Cover
Due to a Microsoft event (an annual malware fest for lobbying and marketing purposes) there was also a lot of Microsoft propaganda
Microsofters Try to Defund the Free Software Foundation (by Attacking Its Founder This Week) and They Tell People to Instead Give Money to Microsoft Front Groups
Microsoft people try to outspend their critics and harass them
[Meme] EPO for the Kids' Future (or Lack of It)
Patents can last two decades and grow with (or catch up with) the kids
EPO Education: Workers Resort to Legal Actions (Many Cases) Against the Administration
At the moment the casualties of EPO corruption include the EPO's own staff
Gemini Links 22/11/2024: ChromeOS, Search Engines, Regular Expressions
Links for the day
This Month is the 11th Month of This Year With Mass Layoffs at Microsoft (So Far It's Happening Every Month This Year, More Announced Hours Ago)
Now they even admit it
Links 22/11/2024: Software Patents Squashed, Russia Starts Using ICBMs
Links for the day
Over at Tux Machines...
GNU/Linux news for the past day
IRC Proceedings: Thursday, November 21, 2024
IRC logs for Thursday, November 21, 2024