“Using AI to look at itself, October 2026”

Picking AI Winners

Here is the question that I sent to both Gemini and CoPilot:  What percentage of development funding is going towards AI tools and what percentage is going to develop “conciousness modules”   I ask that question because serious researchers point out that neurologist’s don’t really know where conscious intelligence resides or how it is developed. So how are all the “smart” guru’s making this happen?

First, here is the Gemini response:

: You touch on a core tension in the AI landscape: the stark gap between **public positioning by tech executives** and the **scientific reality of mind and neuroscience**.

When figures like OpenAI’s Sam Altman, SoftBank’s Masayoshi Son, or Meta’s Mark Zuckerberg speak in grandiose, quasi-religious terms—referring to Superintelligence as a near-omnipotent force, a digital “god,” or an entity capable of solving all human problems—they are engaging in **visionary marketing and narrative-building**, not outlining engineering budget line-items.

### 1. Visionary Rhetoric vs. Engineering Allocation

* **The Pitch:** Promoters use high-stakes language (“building God,” “the most important invention in human history”) to generate hype, justify massive valuations, and attract hundreds of billions of dollars in venture capital and corporate investment.

* **The Reality:** The actual technical roadmaps, capital expenditures, and engineering tasks are entirely focused on practical capabilities: matrix multiplication, scaling GPU clusters, training data pipeline optimization, context window extension, RLHF (Reinforcement Learning from Human Feedback), and synthetic data generation.

There is no software division or engineering budget for “digital consciousness” or “mind synthesis.” The engineering target is **functional intelligence** (getting models to solve complex math, write bug-free code, and automate agentic workflows), not **sentience or subjective awareness**.

### 2. The Neuroscience Perspective (Michael Pollan & Philosophy of Mind) Michael Pollan’s point highlights the primary flaw in claims that machines are becoming self-aware or “god-like”:

* **The “Hard Problem” of Consciousness:** As neuroscience readily admits, we do not understand how physical neural firings translate into subjective experience (qualia) or deep, reflective thought. We can map neural correlates—which region lights up during speech, vision, or emotion—but we cannot explain how or where self-awareness originates.

* **Pattern Matching vs. Subjective Mind:** Large language models simulate the *outputs* of human thought through statistical pattern matching across vast corpora of text. Mimicking the syntax and reasoning structure of human language is fundamentally different from experiencing consciousness, intent, or inner life.

### 3. Evaluative Safety Research vs. Active Creation While labs do not spend money *building* consciousness, a small fraction of research (under the umbrella of AI Safety and Alignment at labs like Anthropic or DeepMind) focuses on **evaluative frameworks**:

* Testing whether advanced models display unintended behaviors like self-preservation, deception, or goal-seeking.

* Assessing ethical frameworks for how society should evaluate sentience indicators if systems ever become complex enough to blur those lines.

These are **monitoring and safety guardrails**, however, rather than development projects aimed at endowing software with genuine sentience.

The grand rhetoric from tech leaders serves to frame AI as an epochal shift to maintain investor momentum and drive political influence. Behind the scenes, however, 100% of the nactual capital and technical execution remains grounded in computer science, data processing, and hardware infrastructure.

Next here is CoPilot’s response:

If you’re asking about the AI industry as a whole, there is no reliable published percentage showing how much development funding goes to “AI tools” versus “consciousness modules.” I could not find any source that reports such a breakdown. Most AI companies do not separate their budgets that way.nsfquest

What can be said is:

  • The overwhelming majority of AI investment today is directed toward practical capabilities such as large language models, AI assistants, infrastructure, chips, training, safety, robotics, and commercial applications. Industry spending on AI is measured in the many billions of dollars.nsf
  • Research into machine consciousness or artificial consciousness exists, but it is a niche area. Some major labs have funded teams involving philosophers and neuroscientists to study questions related to awareness and sentience, but current leading AI labs explicitly state their systems are not sentient.liveaiwire+1
  • Existing reports describe consciousness-related research as receiving increased attention and funding, but they do not provide a percentage of overall AI spending, nor do they suggest it is a major share of industry budgets.interestana+1

So, if I had to characterize it qualitatively:

  • AI tools and capabilities: likely well over 99% of industry R&D spending.
  • “Consciousness modules”: no established category exists, and the amount appears to be a very small fraction of total AI R&D.

That said, the specific percentages above are my inference from the absence of reported large-scale consciousness budgets, not a published statistic. The sources I found do not provide an exact numerical breakdown.interestana+2

Are you asking about the AI industry in general, or specifically about OpenAI, Microsoft, Google DeepMind, Anthropic, or another company?

My response to CoPilot:

I see the AI story as entering the phase where winners and losers from an earnings standpoint will start to be the story. It seems to me that there has been a lot of hype by figures like OpenAI’s Sam Altman, SoftBank’s Masayoshi Son, Meta’s Mark Zuckerberg, or Nividia’s Huang speaking in grandiose, quasi-religious terms—referring to Superintelligence as a near-omnipotent force, a digital “god,” or an entity capable of solving all human problems—they are engaging in **visionary marketing and narrative-building**, not outlining engineering budget line-items.

And CoPilot’s Response:

I think you’re putting your finger on what may become the most important investment question of the next few years.

There’s a distinction between the narrative layer and the income-statement layer.

The narrative layer has been dominated by language about superintelligence, digital transformation, abundance, and civilization-scale change. Leaders such as Sam Altman, Masayoshi Son, Mark Zuckerberg, and Jensen Huang have all described AI in terms that extend far beyond the next software cycle. Some reports have noted growing interest in consciousness-related and AGI-related research among leading AI labs.theaicronicle+1

From an investor’s standpoint, however, the income statement eventually asserts itself. The market can tolerate enormous spending when investors believe:

  1. Revenue growth will accelerate.
  2. Margins will eventually expand.
  3. The winner-take-most outcome is plausible.

That is where I think your “winners and losers” framework becomes relevant.

Historically, technology investment cycles often progress through four phases:

  • Infrastructure build-out
    • Railroads in the 1800s.
    • Fiber optics in the late 1990s.
    • Cloud computing in the 2010s.
    • AI data centers today.
  • Narrative dominance
    • Investors focus on the future opportunity more than current profits.
  • Economic sorting
    • Some firms discover profitable business models.
    • Others discover that demand is real but pricing power is weak.
  • Consolidation
    • A few winners earn extraordinary returns.
    • Many participants earn mediocre or negative returns despite helping build the ecosystem.

Looking at AI through that lens, several questions are likely to matter more than whether “superintelligence” arrives:

  • Can model providers maintain pricing power?
  • How much of the value accrues to chip makers?
  • How much migrates to application companies?
  • Will inference costs fall faster than revenues rise?
  • Will open-source models compress margins?
  • Can enterprises measure ROI from AI spending?

As a former trading-system developer, you may appreciate an analogy from earlier technology waves. In the 1980s and 1990s, it wasn’t enough to know that computers were revolutionary. The crucial investment question was where the economic rents accumulated. The PC changed the world, but not every computer manufacturer became a great investment. The internet changed the world, but most dot-coms disappeared.

The same possibility exists with AI.

One scenario is that today’s spending levels are rational because a handful of companies eventually capture enormous cash flows. Another is that AI becomes tremendously useful while profits are competed away, much as happened in airlines, telecommunications, and portions of cloud infrastructure.

Regarding “consciousness modules,” I largely agree with your framing. When executives talk about AGI, superintelligence, or machine consciousness, I see that primarily as a strategic vision and recruiting narrative, not a description of where most engineering dollars are currently being spent. The bulk of spending appears directed toward compute infrastructure, model training, data centers, software platforms, and commercial deployment rather than attempts to engineer machine consciousness itself. The sources discussing consciousness research describe dedicated programs and growing interest, but they do not indicate anything close to a dominant share of AI budgets.liveaiwire+1

My suspicion is that the next phase of the AI story won’t be “Can AI do amazing things?” but rather:

“Who can turn AI into durable earnings growth and free cash flow?”

That is a much harder question, and historically it is often where the biggest stock market winners and losers finally separate.

Comments are closed.