AI’s Image Problem: Why the Public Distrusts the Future, and Who’s Fixing It

Artificial Intelligence, a technology with roots stretching back over half a century, has only recently burst into the mainstream consciousness, largely thanks to ChatGPT.

Despite its sudden ubiquity, AI is already facing significant public distrust and dislike, a stark contrast to the decades it took for the internet and social media to draw similar ire.

Feature	AI Public Perception	Public AI Tool Usage
Risk vs. Benefit	57% say risks outweigh benefits	Not applicable
Concerns	Creativity, relationships, news, elections	Not applicable
Data Centers	NIMBY resistance across US	Not applicable
Usage Growth (Annual)	Not applicable	38% increase in use for research
Overall Popularity	Less popular than ICE	Over half now use AI tools

The AI Backlash: Instant Reflex or Media Echo Chamber?

The techlash against AI feels like an instant reflex, a stark contrast to the two decades it took for similar sentiments to brew against the internet and social media.

A significant 57% of Americans believe AI’s risks outweigh its benefits, making it less popular than even ICE.

Concerns range from its impact on creativity and meaningful relationships to the integrity of news and elections.

This widespread apprehension has even led to a rare bipartisan consensus, with both Republicans and Democrats adopting a NIMBY stance against data centers nationwide.

Yet, amidst this growing public skepticism, the use of AI tools is soaring, with over half of the population now employing them for tasks like research—a 38% increase in just one year.

This dichotomy between professed dislike and increasing utility suggests a complex relationship with the technology.

“It is, however, difficult to discern true public opinion from performative public opinion polls, as polls reflect media’s self-fulfilled prophesies. In reality, even as views of AI sour, the public’s use of the tools soars.”

The Oligopoly Problem: Who’s to Blame for AI’s Bad Rap?

The author argues that the problem isn’t AI itself, but rather the “AI boys” who form the new oligopoly.

The recent OpenAI trial, pitting Musk against Altman, exemplified the public’s weariness of tech titans.

Figures like Thiel, Andreessen, Karp, Zuck, Ellison, and Bezos, along with “moral entrepreneurs” like Tristan Harris and Eliezer Yudkowsky, are seen as contributing to the negative perception.

Their dismissal of humans as “meat computers” certainly doesn’t help public relations.

However, AI is constantly amazing, even skeptical observers, with its capabilities and tricks, as highlighted by the author’s experiences co-hosting AI podcasts and editing a Bloomsbury book series on AI.

To truly understand AI’s inner workings, the author turns to master communicators like Jensen Huang of Nvidia and Yann LeCun of NYU and AMI Labs.

These leaders offer realistic perspectives, shying away from both exaggerated claims of superintelligence and dire warnings of human extinction.

Both Huang and LeCun have significant economic interests in AI’s success, yet they prioritize educating diverse audiences, from customers to policymakers, about the technology’s true state and potential.

Their ability to communicate complex ideas effectively is a crucial skill for fostering informed public discourse.

Jensen Huang: The Maestro of Scale and Accelerated Computing

Jensen Huang, Nvidia’s founder and CEO, is a master communicator, known for his captivating keynotes that blend product announcements with profound lessons on AI’s scale.

His presentations, often two hours long, showcase new chips and boards, explaining Nvidia’s achievements and market position.

In his March 2024 developers’ conference keynote, Huang unveiled the Blackwell GPU, a colossal chip housing 208 billion transistors.

He vividly illustrated how these chips, multiplied across data centers, represent a staggering six quadrillion transistors, powering the wonders of generative AI.

Huang’s key lesson revolves around the value of scale, emphasizing that generative AI’s breakthroughs are not just theoretical but also a product of immense computing power, or “compute” in industry argot.

He also declares the end of Moore’s Law, touting Nvidia’s accelerated computing which has boosted power 1,000x in eight years, a significant leap from the 100x increase every decade predicted by Moore’s Law.

Huang is also a strong advocate for robotics, envisioning a future where AI-powered digital twins predict every possible scenario for factory robots and autonomous cars.

His ability to articulate the value of what’s being built, rather than making outlandish predictions, sets him apart.

“AI is not a tool. AI is work. AI is work that can use tools.”

Yann LeCun: The Contrarian Visionary of World Models

Yann LeCun, an NYU professor and AI godfather, offers a different, yet equally compelling, communication style.

A researcher and educator first, LeCun is known for his clear-eyed academic approach, devoid of hyperbole or fear-mongering.

He earned the prestigious 2018 Turing Award for his foundational work on deep learning, alongside Yoshua Bengio and Geoffrey Hinton.

LeCun is a notable contrarian regarding Large Language Models (LLMs), arguing that while impressive, they are essentially a “dead-end” because they operate within the finite sphere of text representation.

Instead, he advocates for world models, which aim to teach machines experiential constraints much like a toddler or kitten learns about gravity and object permanence.

This approach addresses Moravec’s paradox, highlighting the difficulty for machines to acquire the perceptual and mobility skills of a one-year-old.

His proposed Joint-Embedding Predictive Architecture (JEPA) uses real-world photos and videos to train models to understand representations of the world and predict consequences of actions, rather than merely predicting pixels or mimicking human teachers.

JEPA promises models that can handle a wider range of tasks with less training, are more efficient by focusing on specific changes, and, crucially, are designed to be safer.

Unlike the “general machines” touted by some AI developers, JEPA will produce specialized agents with bounded tasks, each built to “anticipate the consequences of its own actions.”

LeCun firmly believes that “reality is way more complicated than language,” underscoring the need for AI that understands the physical world.

His support for open source, exemplified by his advocacy for Meta’s Llama models, reflects a scientific worldview rooted in collaboration and continuous discovery.

The Future Outlook: Informed Discourse and Responsible AI Development

AI is here to stay, and its future trajectory hinges on informed public discourse, not mere protest or regulation.

To have a voice in its direction, a deeper understanding of AI is paramount, necessitating critical listening to the actual builders of the technology.

The communication skills of leaders like Jensen Huang and Yann LeCun are invaluable in this regard, offering realistic and educational perspectives.

Their ability to demystify complex concepts empowers a broader audience to engage with the implications of AI.

By dissecting their communication strategies, we can foster more effective and honest discussions about technology, ensuring a future for AI that benefits all, rather than being controlled by a select few.