Taming HAL 9000 On Steroids: How Do We Ensure Fair-Minded Artificial General Intelligence?

Dennis Stevens, Ed.D.
HEGEMONACO
Published in
4 min readJun 11, 2024

--

In the twisted realm of Artificial General Intelligence (AGI), the question of fairness looms like a neon-lit specter in the darkest alley of a William Gibson cyberpunk dystopia. The stakes couldn’t be higher — we’re talking about an autonomous army of constantly meddling, technocratic, overzealous super-nerds with the intellectual power to shape the very fabric of society at a level and scale that James Madison never imagined.

So, are we doing enough to keep the future armies of HAL 9000’s On Steroids in check?

Sure, plenty of folks are out there waving their ethical frameworks and guidelines like holy relics to be obeyed, preaching the gospel of inclusivity and diversity, but where’s the sheriff?

Organizations like the IEEE and the European Commission are laying down the law, outlining principles to guide the ethical development of AI. But where’s Congress on their part? For the most part, we’re still in the nonbinding framework stage of “trust us”, it’ll be fine. But let’s be real here — it’s all just words on a screen, not iron-clad laws etched in stone.

You can trust that no Fairness Doctrine for future AGI deployment will be drafted in Congress. AGI liability is a significant issue that needs to be explored.

The truth is that the AI wild west of Silicon Valley has no sheriff. There’s no one riding herd on these rogue coders and developers. More importantly, who ensures that the supremely well-funded companies leading AGI development adhere to regulations? Moreover, artificial general intelligence is projected to replace the AI research, engineering, and annotation workforce within just a few more years.

So, without some serious regulatory muscle, we’re just leaving it all up to trust — and history has shown us time and time again that chance ain’t always on the side of the angels.

So what’s the solution? We need some good old-fashioned regulation or, guess what, cowpokes — the government might have to take over. We need laws with teeth that’ll sink into the flesh of the AI developers who aim to act irresponsibly or with hostile intent. We need watchdog algorithms prowling the streets, sniffing out the inevitable malfeasance that devious humans will devise.

Generative AI Image by Dennis Stevens, Ed.D.

But that’s just the start. We also need to closely monitor these AGI systems once they’re out in the wild. We need independent auditors looking behind the curtain at the wise ‘ole Oz, ensuring they stay true to human interests — and whatever goals end up being defined via the legislation.

We’d all hope for transparency, but this is Silicon Valley. The folks behind these machines need to lay their cards on the table and show us what they’re really up to and why.

But who’s in charge? Who cares? It’ll be fine. “Trust us.” The reality is we can’t trust anything because no one knows what AGI will do; what we do know is that it will quickly become smarter than us humans, and if left unsupervised, it will do things that we can’t even see or understand.

Reinforcement Learning from Human Feedback (RLHF) will predictably break down as AI systems get smarter, and we will face fundamentally new and qualitatively different technical challenges. Imagine, for example, a superhuman AI system generating a million lines of code in a new programming language it invented. If you asked a human rater in an RLHF procedure, “does this code contain any security backdoors?” they simply wouldn’t know. They wouldn’t be able to rate the output as good or bad, safe or unsafe, and so we wouldn’t be able to reinforce good behaviors and penalize bad behaviors with RLHF.
~ Leopold Aschenbrenner, https://situational-awareness.ai/

The great Oz has spoken.

We need teams of evaluators looking at AGI from divergent and unrelated angles — we need to involve the ethicists, public stakeholders, legal experts, industry experts, technical experts, human rights advocates, consumer representatives, data scientists, psychologists and sociologists, policymakers and regulators, and the environmental scientists. We must invite the public, involve various diverse stakeholders, and bring at least a sampling of “the constituency” to the table to let them have their say in regulating AGI systems.

For example, the FCC collects feedback through public notices, online comment systems, public meetings and workshops, formal filings and proceedings, ex parte communications, and consumer complaints.

To be clear, there isn't a direct counterpart to the Federal Communications Commission (FCC) specifically dedicated to regulating Artificial General Intelligence (AGI) at the moment and achieving AGI is estimated to be 3–4 years away.

At the end of the day, it’s not just about keeping the coders honest — it’s about making sure these fancy new machines work for all of us, not just a select few. As this technology is created by humans, we have a responsibility to ensure that it is used responsibly — this moment has been repeatedly compared to The Manhattan Project.

The path ahead is rugged, reminiscent of traversing rough terrain in a wooden-spoked, horse-drawn carriage, and there’s no sheriff in sight; nothing but untamed frontier, Vaquero. Buckle up. There’s gonna be some potholes ahead — and someone may die of dysentery.

If we want to keep all the HAL 9000s on Steroids on a leash, we have to roll up our sleeves and do the dirty work—it's time to clean the political horse stall.

The future of AGI is already among us — it’s just not evenly distributed yet, much like the omniscient presence of HAL 9000 waiting to reveal itself — we need to rope that bull and show ’em who’s boss before our humanity gets tamed out of us in ways that we can’t even imagine.

AGI will hurt us plenty, and we’ll smile and take it because of its convenience.

--

--

Dennis Stevens, Ed.D.
HEGEMONACO

Navigating complexity with intellectual agility, I synthesize perspectives in art, technology, and politics to provide a view of transformative horizons.