AI Hype Deconstructing the AGI Myth

The digital airwaves are thick with pronouncements from Silicon Valley’s high priests: artificial intelligence, they tell us, is on the precipice of achieving true consciousness, a “general intelligence” that will rival, perhaps even surpass, our own.

We are urged to believe that a future populated by sentient machines is not merely plausible but imminent, a matter of “sparks” about to ignite into a roaring bonfire of superintelligence.

Yet, as the din of this technological evangelism grows louder, a crucial question emerges: Is this a genuine scientific breakthrough, or a meticulously crafted narrative designed to serve a very different agenda?

According to Emily M. Bender, a professor of linguistics at the University of Washington, and Alex Hanna, director of research at the Distributed AI Research Institute, the latter is far closer to the truth.

In their forthcoming book, “THE AI CON: How to Fight Big Tech’s Hype and Create the Future We Want”, they peel back the layers of sensationalism to reveal a landscape where the pursuit of artificial general intelligence (AGI) is less about scientific discovery and more about corporate interests and a disturbing ideological undercurrent.

Their argument is sharp: the current AI hype, particularly around AGI, is “foolhardy at best, and deceptive and dangerous at worst.”

The core of the deception, Bender and Hanna contend, lies in the deliberate fuzziness of the terms themselves.

What, precisely, is artificial general intelligence? And what constitutes “superintelligence”??

Even leading AI firms like OpenAI have sidestepped a definitive answer, suggesting their own boards will ultimately decide when their algorithms achieve AGI.

This ambiguity is not accidental; it serves to maintain a constant state of anticipatory excitement, a perpetual “almost there” that fuels investment and public fascination.

Consider the recent history: In February 2022, OpenAI’s Chief Scientist Ilya Sutskever mused on Twitter that “today’s large neural networks are slightly conscious.”

Just months later, a Google engineer, Blake Lemoine, was convinced Google’s LaMDA language model was sentient, leading to his dismissal – not for his outlandish claims, but for leaking corporate information.

Subsequently, a Google VP Fellow, Blaise Agüera y Arcas, rather than refuting Lemoine, suggested LaMDA indeed “understands” concepts, blurring the lines between statistical pattern recognition and genuine cognition.

Then came Microsoft Research’s April 2023 paper, “Sparks of Artificial General Intelligence: Early Experiments with GPT-4,” claiming GPT-4 could solve “novel and difficult tasks” across diverse fields, thus showing the first “sparks” of AGI.

The language is carefully chosen, designed to evoke a sense of inevitable, self-propagating intelligence.

But Bender and Hanna argue these claims are not novel; they echo bombastic predictions made over 60 years ago about computers supplanting human intelligence.

The persistent narrative of sentient machines needing “robot rights” or Matrix-style superintelligence threatening humanity is, in their view, a highly effective marketing tactic.

It elevates mundane statistical models to the realm of the mystical, making them appear far more sophisticated and transformative than they are.

The payoff, according to estimates from consultancies like McKinsey, is potentially trillions annually, a figure that, while part of the hype machine, is certainly enough to keep venture capitalists pouring billions into generative AI firms.

Beyond the commercial imperative, the authors uncover a more sinister dimension to the AGI chase: its conceptual roots.

The very notion of “general intelligence” lacks an accepted, scientific definition.

When Microsoft’s “Sparks” paper provided a preliminary definition, it conspicuously omitted references to fields like psychology or cognitive neuroscience, disciplines that might actually offer insight into such a complex concept.

The definition cited in a prior version of their paper – “a very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience” – stems from a 1994 Wall Street Journal editorial signed by psychologists.

While seemingly benign, the authors contend that the project of identifying “general intelligence” is “inherently racist and ableist to its core.”

This makes the pursuit of artificial general intelligence not just misguided, but potentially dangerous, as it risks embedding and amplifying existing societal biases under the guise of technological advancement.

The investment in AGI is not just about following market trends; it’s driven by “ideological billionaires” – figures like Elon Musk and Marc Andreessen – who are not merely funding but actively proselytizing for AGI.

Bender and Hanna suggest this pursuit is intertwined with conservative politics, pro-birth policies, and a right-wing attack on multiculturalism and diversity, all cloaked in the shiny veneer of technological progress.

This revelation adds a troubling layer to the narrative, suggesting that the drive for AGI is not just about profit, but about shaping society according to specific, often regressive, worldviews.

The AI market itself is a tightly knit ecosystem.

Researchers and founders move seamlessly between tech giants like Microsoft, Google, and Meta, or launch startups that quickly secure millions in venture capital.

This creates a self-reinforcing loop where the same individuals and entities are both creating the technology and propagating the hype, ensuring a continuous flow of investment.

In 2022 alone, 24 Google researchers departed to join AI startups, while in 2023, a staggering $41.5 billion in venture deals flowed into generative AI firms.

This financial deluge, coupled with the sensational claims, creates a powerful feedback loop that drowns out dissenting voices and critical analyses.

The real implications of this pervasive hype extend far beyond the circulation of funds among investors.

Ordinary people are being told they face imminent job displacement, their livelihoods threatened by the very technology they are encouraged to believe is on the cusp of sentience.

This fear, paradoxically, serves to further entrench the power of a select few who are driving the AI narrative.

In essence, Bender and Hanna argue that the true “con” is not a robot uprising, but the systematic misdirection of public attention and capital.

The danger isn’t that AI will become superintelligent and destroy us; it’s that the exaggerated claims and unchecked investment obscure the real drawbacks and pitfalls of current AI technologies, while concentrating immense power and wealth in the hands of a small, ideologically driven elite.

It’s a sophisticated marketing campaign masquerading as a scientific frontier, and understanding this distinction is crucial to fighting Big Tech’s hype and building a future that truly serves humanity, not just its most powerful actors.