The Never-Ending Hubris Of GenAI

Image: Badly edited by author

I wrote the following post almost two years ago, and as I was reading through it, it’s startling to see how little has changed.

Generative AI is still attempting to infiltrate every corner of our creative world, it’s still seeping into every piece of software we use, and it’s still only gotten to the point where it’s a relatively useful tool (when it’s not acting like a petulant child).

More so, we still have the overlords — Altman, Musk, Zuckerberg, Nadella, Amodei and co — running their mouths 24/7 telling us that AI will definitely, at some point, almost certainly, probably, maybe, could be our future. But what have they done to prove that in the last two years? Sure, they’ve burned billions of dollars in venture capital buying cloud computing and GPUs and building data servers, but what has changed? The models have only gotten incrementally better, and still don’t make anywhere near enough profit to become a sustainable businesses that can survive without hundreds of billions of investment every quarter. All that’s really changed is that we find ourselves at a point where the entire industry seems to be propped up on the delusion that the insatiable appetite for compute spend will increase exponentially, forever.

What could go wrong.

Anyways, here’s the unabridged post from two years ago which provides a fascinating reality check into how everything is much the same, just costing more money, and filling the wallets of those at the top faster than ever.

If you dip your toes into the generative AI debate, you’ll come across this common thread from the “pro camp” — artists, photographers, videographers, musicians, coders and writers should happily welcome generative AI into their work processes or use it to replace them entirely, and you’re an idiot if you think otherwise.

As the saying goes, “AI won’t replace humans bro, but humans using AI will replace other humans who don’t.”

Well, for a technology that does some stuff to an average enough level — and does other stuff inaccurately or absurdly wrong — I’ve seen little to justify this steadfast belief that the majority of creatives will a) find it helpful and b) actually want to use it. It has offered almost nothing of real significance so far except to help throw more sludge into the already overwhelmed content well. It’s resulted in more content, but all of it is of less quality than what we had before. Great. The belief that it should be widely embraced also side-steps the great rift that has been caused by the fact most of these LLMs are trained on stolen content, for which the original creator will never see a dime. Despite that, we’re all meant to eat it up anyway.

But worse than the shouting about GenAI is the god complex consuming the current crop of overlords pushing the technology.

It’s overwhelming.

It’s insufferable.

And to this point, it’s completely unjustified.

The statements of grandeur have been never-ending. In a few years, human programmers won’t exist. AI is going to become its own race. Movies will become video games, and video games will become something unimaginably better(whatever that means). If AI can replace a creator, maybe they shouldn’t have had that job. AGI is going to be hot shit and change the world. Generative AI deserves to burn through the entire world’s cash, but we promise it will be worth it.

They wax on about how their companies should command insanely high valuations despite being unprofitable — or, in many cases, despite producing nothing at all. They demand that governments allow them to keep training on copyrighted data because they can’t grow if they can’t hoover up every single morsel of data in existence (and, of course, they don’t want to pay for that data.) They lament the lack of infrastructure holding back their goals and expect others to fund it. They blatantly copy, rip off and misuse data, image rights and more, and shrug like, “we need that data to build.”

They do what they want, when they want, don’t ask for permission, and certainly don’t care to ask for forgiveness.

It shouldn’t be a surprise, though. Like most tech trends, GenAI is, at present, a self-fulfilling prophecy. It’s big because everyone keeps saying it’s big. It’s the future because everyone is telling each other it’s the future. Its use cases are endless as long as industries keep telling themselves that’s going to be the case. Trends thrive — and survive — off the possibilities and the pipe dreams. As long as that momentum is sustained, the line will keep going up, and the egos will keep inflating.

Hype cycles are built on this hubris and ego, and it appears GenAI is no different, following in the footsteps of previous ‘Next Big Things’ like crypto, NFTs, co-working spaces and more. (You know you’re trapped in a hype cycle when a guy who owns a company that makes computer chips is signing tits.) It’s these bold, brazen and frankly bullshit predictions, statements and flat-out lies that fuel the hype, that keep investors frothing at the mouth, that keep the market trickling higher and higher. It’s these fleeting yet completely vague promises of an AGI future, of a worker-less workforce, of profits and unprecedented productivity that keep businesses and companies and entire industries pumping more and more money into AI ventures.

But here’s the reality —

GenAI hasn’t earned its right to be treated as some divine technology, and the tech moguls behind it certainly haven’t earned the crown they’ve placed atop their heads. It hasn’t proved that it will transform any industry. It hasn’t proved its utility. It hasn’t even proved that it has staying power. Companies like OpenAI are losing billions of dollars a year, and its current big plan is to charge more for ChatGPT? Ed Zitron suggests that OpenAI would need to raise over $5 billion in the next 12 months to survive and that it will likely keep raising similar sums from then on. The entire industry likely needs to make $100s of billions of dollars a year to be sustainable and profitable, and to justify the frankly insane resources it commands, which are potentially dangerous to the planet, all to generate dumb as fuck images of Presidential candidates dressed as communists.

Therein lies the problem. GenAI is running out of quality data to train on and seems unwilling to license what’s left. (How about we get a slice of future profits like current investors are being offered?) Progress appears to have plateaued — if anything it’s gone backward — and new models are no longer the leap forward they once were. There are rumors that OpenAI’s big new release, Strawberry (a model that can “think” and was once deemed “a threat to humanity”), is actually only slightly better and can take 10-20 seconds longer to generate a result. The industry is already pinning its hopes on the next thing, AI agents, and of course AGI, a concept no one seems to be able to define or explain how and why it will be revolutionary to society (“just because the AI overlords says so” isn’t an answer). What happens if both of those concepts fail to establish themselves?

Most companies and enterprises that have AI-stuffed their products and services with abandon have realized the only way to make money from doing so is to raise prices. Just look at Canva, which first forced AI features onto users and is now forcing users to pay up to 300% more for them. Price hikes scream desperation (or regret). Across industries, there’s a dawning realization that adding AI isn’t a magic wand that wishes up unimaginable productivity and profits. It has to actually be, you know, something that solves a current problem and offers genuine utility to a business and its processes.

All of this is to say I’m once again tired of being told this technology is going to change my life. I’m tired of being told I’m some loser freak for not embracing, for not diluting my creative process into the sum act of entering a prompt. I’m tired of seeing the top brass of AI companies run their mouth without ever delivering anything to back it up.

If you’re going to talk the talk, you need to walk the walk.

It’s time GenAI puts up or shuts up.