It’s AGI or bust for all of us

Here’s a fun tidbit of information: whether you believe in the existence of paranormal phenomena or not, about half of the people in the US disagree with you.

Isn’t that weird?

Per a Newsweek article published in June of 2025, about 48% of American adults believe in “psychic or spiritual healing,” while 39% “express a belief in ghosts.” 

I mention this here, in a newsletter about AGI ethics, because this month’s topic is “Public Opinion.” And, much like our opinions on the paranormal, people tend to have very strong opinions on AI that are, by and large, based on nothing more than their personal feelings.

• Even a cursory glance at conversations on Reddit reveal deep biases toward/against specific LLM models ground in anecdotal experience
• There’s a culture war being waged on X.com over the specific amount of nudity xAI’s “Grok” model should be allowed to generate
• The debate over “AI slop” has polarized the creative and tech communities into distinct philosophical camps
• It’s becoming fashionable for tech savvy consumers to have a “favorite” chatbot, something reminiscent of the “console war” in previous tech cycles

It’s gotta be the shoes

How did we get here, to this place, where everyone is looking at the same information but taking away wildly different observations?

My answer: we all want to believe in something magical, even if it doesn’t make sense. And big personalities such as Elon Musk and Sam Altman make us feel like it’s okay to believe. 

It’s cool to believe in AGI. It’s fun. It’s hip. It’s what middle-aged C-suite executives think the kids are doing.

In lieu of a clear “killer app” for chatbots, LLMs have reached a sort of listless permeation of the mainstream. They’ve captured the public’s attention in much the same way that Nike has become the largest fashion brand in the world by market cap.

People who’ve never so much as dribbled a basketball are willing to pay $200 to wear the same shoes as Lebron James or Michael Jordan. 

Likewise, a seemingly small percentage of chatbot users have a distinct, specific use case that no other tool supports as well. For the rest of us, however, LLMs are mostly a curiosity.

The difference between Nike’s shoes and an LLM, however, is that it only costs about $35 to make a shoe. It takes billions of dollars to train and operate LLMs such as ChatGPT and Claude. And, unlike OpenAI, Nike supports the development of more shoes by being a profitable company.

Eventually, firms such as OpenAI will have to produce AGI or some sort of killer app that justifies further development just to stay alive. Public sentiment doesn’t last forever.

What if that killer app never comes? 

What if, no matter how good chatbots get at what they can currently do, most of us slowly stop coming back to the blank text box on our screens to answer the same vapid question over and over? Afterall, the most valuable commodity in the world is attention. 

The Gemini UI, waiting for attention. Source: Screenshot.

Chatbots aren’t good enough right now. While developers and pie-eyed pundits seem to believe that “vibe coding” and email summarization will fuel another 10 years of consumer fascination with LLMs, but this doesn’t track with reality. 

As someone who makes a living writing about AI and quantum computing for mainstream audiences, I don’t believe that the average person will buy into the ‘augmented power user’ paradigm.

Anyone who doesn’t already have a Python coding environment installed on their laptop is unlikely to embrace vibe coding or automated fintech trading to the same degree that optimistic technology enthusiasts and journalists have.

From where I’m sitting, the only future for consumer-facing narrow AI is a post-agentic, distributed intelligence that facilitates human endeavor at the environmental level. 

In order to become useful, AI needs to fade into the background and go about the work of facilitating human endeavor. It needs to stop requiring our limited attention. And that means foundational AI infrastructure needs to become exponentially more powerful and capable. 

The current approach — cobbling together thousands of GPUs — doesn’t lend itself to this form of distributed intelligence. We can chain discrete units together until we run out of silicon and we’ll never create a machine that’s mathematically capable of reproducing the machinations driving human thought.

For that, I firmly believe we’d need to transition the bulk of global AI development to a quantum computing architecture. 

Opinions and ethics

Here’s where things start to veer off course. 

The general public seems to be entirely convinced that each new LLM release brings humanity closer to some sort of golden age where everyone gets a universal basic income. This leaves most people with little awareness of just how little progress has been made in LLMs since ChatGPT was launched.

Meanwhile, in the same short time frame, the quantum computing sector has experienced what I would deem the most rapid maturity cycle in the history of science and engineering since Oppenheimer et. al., developed the atomic bomb.

I’ve long asserted that AGI cannot be achieved without quantum computing because there’s no evidence to support the notion that “thought” or “cognition” or “reason” can be achieved via binary math processes. 

Related: Human intelligence cannot be predicted

But, the question of whether AGI can be achieved on classical architecture is tangential to the real problem: public opinion.

I want Anthropic and OpenAI to be right. I want AGI to be “just around the corner…” but I don’t think LLMs are the right technology to develop AGI nor are GPU clusters the proper hardware to build it on. This puts me and my opinion (educated as I believe it to be) in the minority.

If I’m wrong, we can all have a big laugh as OpenAI, Anthropic, and the gang eat the world one trillion dollars at a time. 

But, if I’m right, companies such as OpenAI, Anthropic, and Perplexity, who’ve gone all-in on GPUs (or signed exclusivity agreements with firms that have) are in serious trouble. And that could result in catastrophic harm to the global economy. 

The bottom line is that OpenAI and its ilk are not in the business of building AGI. They’re in the business of building LLMs using GPU clusters. If those LLMs don’t magically produce AGI on their own, there’s no “science” to indicate a path forward. 

When public opinion wanes, as it does in almost every consumer-facing technology zeitgeist, and LLMs are no longer propped up on company hype, big tech won’t be the ones who suffer.

• Sony didn’t crash when the 3D TV market died (but companies building 3D apps for TV markets did). 
• Meta didn’t change its name back to Facebook when consumers got bored with the metaverse (but the majority of metaverse companies have now folded or pivoted to crypto)
• Intel didn’t go broke when the dotcom bubble crashed (but so many startups did that a global recession and banking crisis happened in its wake)
• Amazon didn’t go out of business when robot vacuums and smart speakers went out of style (but iRobot went bankrupt and Sonos had to restructure its business after sales hit rock bottom)

And Microsoft, Nvidia, and Alphabet won’t go out of business when people stop caring about chatbots. But it’s AGI or bust for OpenAI, Anthropic, and the rest of the LLM gang and the countdown to extinction started the moment GPT-5 and other cutting edge LLMs failed to wow the mainstream. Once user interest bottoms out, investor interest will follow.

You don’t have to be a Wall Street analyst to understand why Apple chose Google’s AI to power the next version of Siri: those other companies might not exist long enough to make good on the contract. 

Ethical concerns related to this article: 

• To what extent should the progress of AGI research be made public, and what are the risks of both secrecy and full transparency?
• Is it ethical to wage an AGI race with win conditions, specified or not, (to bring such a being into the world) when AGI rights are not yet established?
• How does society prevent unintended side effects-social, cultural, political-when rapid AGI-driven progress outpaces existing institutional or regulatory capacity?

Tristan Greene

A veteran AI journalist, Tristan was the creator and managing editor of The Next Web’s “Neural” imprint. He is one of the first and most prolific reporters to cover artificial intelligence as a full-time beat.