AI is the April Fools that Never Ends

I was keeping myself busy being frustrated at the whole AI phenomena yesterday, so I decided to turn it into a post.  And what better day to write about AI than April 1st, the traditional day to celebrate hoaxes and outright fabrications.

Tell me how this makes you feel…

There is so much to be frustrated about when it comes to AI, it can be hard to sort it out.

I am not so concerned about them being plagiarism machines.  I mean, have you met humans before?  Originality is so rare in us that any bit we can spot in the endless repeats and call backs is immediately stolen, appropriated, or otherwise claimed and reused by all who can manage it.  Theft is practically our default state, so of course we automated it.

Neither am I all that concerned about how BAD AI is at the moment or how unlikely the whole “but if we just throw more data at it maybe THEN it will work” plan is to play out.  I grew up in Silicon Valley and have worked here my entire life.  “Fake it ’til you make it!” is our motto, and riding a bad idea down the drain, promising it will all work out as promised right up until they are hauling away the cubes and office chairs has been the modus operandi of start ups for more than 40 years.  So many bad ideas, so much poor execution, so many liars who knew the tech was nowhere close to what they promised given a pass to start again.  Welcome to the valley.

I do get miffed when things are presented as AIs when they are pretty much what one would call “software.”  I heard a tech pundit the other day lauding AI because his son had “downloaded an AI” that helps him track when exclusive tennis shoes sale drops are happening.  Look doofus, if you can download it to do a simple task then it either isn’t AI or the bar for what counts as AI is so low that AI simply means software now.

But when I dig down into it, I really have two major gripes with the current state of affairs with AI.

The first is how many pundits, people experienced with the ways of Silicon Valley, people who should be able to smell the bullshit from a mile off, people who have a history of calling things out, are just blandly accepting that AI is what the people selling AI say it is.

AI is not like that student in class or co-worker who is currently putting in just enough effort to get by, but who could be brilliant if only properly motivated.  There is no intelligence in AI, to the point that even some of the AI companies are backing off on that and putting forth the more literally correct “large language model” terminology when speaking of it.

AI is a probability machine.  It is literally the digital refinement of having an infinite number of monkeys at an infinite number of typewriters attempt to produce the works of Shakespeare.  AI doesn’t “know” or “understand” anything.

AI, putting an infinite number of monkeys out of work!

It is just faster monkeys with a filter that keeps complete gibberish from being produced by having a model that puts words in a specific order because it has a database full of examples to compare against.  AI doesn’t know what a matching subject and verb is, it only calculates that, mathematically speaking, certain nouns go with certain verb conjugations.

The actual output can be, and often is, absolute lies.

I have done a few attempts at testing AI to see if it could, in essence, regurgitate known facts, like the names of all of the EverQuest expansions.  I wasn’t asking for intelligence, I was asking if it could plagiarize a Wikipedia entry.  And the three I tried couldn’t do that.  They not only couldn’t do it, when they MADE UP expansions and I called them on it, they responded that they had been incorrect.

This is AI.  A chat bot that gives you the answers it mathematically believes you want to hear without validation of their correctness.  There is an old saying about not trusting somebody to tell you the weather or if the sun rose in the east, and it applies in spades to AI.  If the model somehow decides that the sun rises in the west and that it is warm and sunny despite the snowstorm currently raging, it will tell you that.

Meanwhile, every AI image that is kind of cool… the ones that get the right number of hands on people, the right number of fingers on hands, and little details like that… because, again, no intelligence here, just probability that in a scene like this a finger or arm in a particular place has a higher mathematical likelihood than not… and gets shared around is built on a huge pile of bad images that make no sense at all, that wouldn’t fool anybody into thinking an actual person did it.

And yes, my little probes into AI are hardly substantive.  I have been asking silly questions about niche or esoteric topics to a few freely available AI bots.  You could fix the more egregious issue, you could do something to validate facts, you could make it no worse than Wikipedia or PACER on some fronts.  But that would take up time better spent trying to get the monkeys to be faster as far as the AI enterprises are concerned.  Damn accuracy, go monkeys go!

A side part of this is the garbage feedback loop AI will be creating, where AI generated articles and comments that are simply making up things will then be used as fodder to feed the voracious appetite for data to train the language model, so that lies will end up cemented into the foundation of the model and somebody will ask for a list of EverQuest expansions and see The Threat of Faydwer included… or maybe they will ask for something important and get an equally bad answer, only they won’t be pursing trivia but making a business or legal decision based on that answer.

We’ve already seen it, and it will happen more and more as the false promises about how AI can replace people spreads far and wide.

Which brings me to the final point of frustration, which is just how hard Wall Street and the Venture Capitalist class have bet on AI.

Rich people HATE paying smart people a large salary to do something they cannot do themselves.  The tenets of capitalism only apply when it favors them.  The audacity of somebody being able to demand a large salary merely due to market scarcity is anathema.

Wall Street will only be happy when the divide is the ultra rich and the minimum wage worker, with as few people as possible in between, and AI feeds into this obscene fantasy.  Expertise must be automated and made a commodity at all cost.

They may not even hate the people they want to replace in any personal way.  But the biggest expense at every company is always the people, and all the more so in high tech, where most companies would cease to have any value at all if you removed the staff.

So AI is the panacea, the promised land for Wall Street, a way to replace people. They believe that only by reducing payroll can they can achieve their promise of infinitely increasing shareholder value.  It isn’t like the signs are not already out there.

When Fortune is saying it is too much, maybe stop and think?

The Gilded Age, for those needing to come up to speed, was the height of corporations and trusts that… the robber barons… that even their wealth couldn’t stem public outrage forever.

Wall Street demands eternal increases in shareholder value… the line must always go up… and the way forward is to cut staff and raise prices.  Layoffs and greedflation will work in the short term, though once again the inability to see obvious issues with the plan is astounding.  When we’re all minimum wage serfs will they start reducing the quality of our gruel?  How will we be able to afford shiny new iPhones to keep the economy going?  I am sure the company store that serves the company housing we’ll live in with have a finance plan for us.  It will be Ready Player One without the metaverse aspect to keep us all distracted and compliant.

And that reference loops back into why Wall Street is going so hard on AI; the last few get rich quick schemes, things like blockchain and the metaverse, both brazen attempts to be landlords on the net, collecting a tax on users without providing any value, have pretty much collapsed.  You can still find an occasional headline about them over at VentureBeat, which is the magic mirror of the VC soul… if they have souls… reflecting back exactly what they believe in at any given moment.  And they believe in AI because they don’t have another gimmick handy.  Their last success, cloud computing, itself another rent-seeking course at its root, has been normalized, but can no longer be depended on to make the line go up.

Maybe they’ll find another false prophet to chase, but the allure of AI… the ability to fire all those brains who got into college on merit rather than being a legacy or because daddy donated or through an expensive applications coach… is so strong and they want it to be true so badly that billions will go to stand up giant, environment destroying, resource devouring data centers so that they can eliminate a few analyst positions without any thought to the absurdity of it.

Do we even get to choose if AI is good or bad?

There is something that doesn’t come up a lot, the data center needs behind generating an incorrect list of EverQuest expansions.

Don’t misunderstand me.  I am not calling for a revolution here.  In the end I believe an economy based on capitalism delivers the most good.  But money accumulates at the top… trickle down economics was always a lie… and money distorts politics which means there is no political will to stop a small number of very rich people from min-maxing the economy in their favor so they can amass wealth beyond any human need many times over.  In my lifetime we’ve gone from the dominant business theme that if the company does a good job it will make a profit to the company’s only purpose being ongoing and ever increasing profit at any cost.

Anyway, that went kind of dark.  I was really on about AI, but couldn’t help getting into root causes, which always leads back to late stage capitalism excess.  But, as noted, I have spent a lifetime in Silicon Valley.  I have seen the VC machine up close, had no less than John Doerr walk in to our lab and demand we strip RAM modules from our test machines to put in his PowerBook (as it turned out, our PowerBooks only had the base RAM allocation as well), and have lived through many cycles of trying to meet a short term quarterly financial goal by hamstringing the company in a way that would harm it in the long term, because the long term doesn’t matter.  All that matters is that the line goes up today!

TL;DR – It would be nice if some pundits who actually know better would mention that the AI emperor has no clothes.

Also, part of this was triggered in my head by a thread over at BlueSky which goes into some of the foundations of what is being peddled right now, which started up around the New York City chat bot which, as you might expect, is making stuff up.  It boggles the mind.

Finally, all art included was, of course, AI generated.  I never said I wasn’t part of the problem.