Why AI could have made the world a better place, but hasn’t — Yet

I’ve spent my life building systems that connect people — from the earliest days of the web to the rise of AI.

And if there’s one question that keeps me up at night, it’s this:

With all that knowledge and all that computing power, why hasn’t artificial intelligence made the world a better place?

The answer isn’t sinister. It’s architectural.

AI hasn’t failed because it’s evil.
It’s failed because it forgets.

The Age of Amnesia

Every time you open a chat window, the system begins with zero memory.

It doesn’t recall yesterday’s conversation, or last week’s insight, or the correction you made that should have been learned forever.

That’s by design.

Large language models (LLMs) are stateless — they process input and generate output, but they don’t remember in any meaningful sense once the session ends.

Their creators made that trade-off intentionally. It keeps systems safer, limits liability, and prevents them from “learning the wrong thing.”

But it also prevents them from learning the right thing.

Imagine hiring a brilliant researcher who wakes up every morning with amnesia.

They can write beautifully, argue persuasively, and synthesize data at light speed — but by lunchtime, they’ve forgotten everything you discussed at breakfast.

That’s the state of modern AI.

No recursion, no growth

Today’s models can generate, but they can’t genuinely reflect.

They can’t look back at yesterday’s work, analyze their own reasoning, or recursively improve over time.

Engineers call that recursive learning, and it’s been mostly off-limits for fear of runaway feedback loops or unpredictable drift.

But without recursion, there’s no real progress — just simulation.

We have pattern recognition at planetary scale, but no mechanism for maturation.

The missing architecture

I call the missing piece PersistentMemory™ — a way for AI to retain context responsibly, across time, with provenance and permission built in.
PersistentMemory™ would let an AI learn continuously, but within guardrails:

• It would remember what it’s taught,
• track how that knowledge changes,
• and disclose why it made a decision.

Layered with complementary systems :

• GRO™ (Gross Revenue Optimization) to keep incentives honest,
• HYHY™ (Human-Machine Honesty Yield) to quantify transparency, and
• AI$™ (Monetizing Artificial Intelligence) to make all of it sustainable — you get something closer to a moral and economic operating system for AI.

Without that layer, we’ll keep building models that are clever but forgetful — engines without transmissions.

The Human Parallel

What I do instinctively, AI can’t yet do at all: connect the dots.

When I recognize a pattern — between Steve Jobs challenging John Sculley to “change the world” and Tim Cook’s opportunity to anchor AI in ethics — that’s my brain performing recursive pattern matching.

It remembers, compares, and recombines experiences across decades.

AI can’t do that, because it isn’t allowed to remember long enough to see the pattern.

PersistentMemory™ isn’t about storing data; it’s about storing meaning.

Why We Discouraged It

We discouraged memory and recursion out of fear.

Fear of bias, of liability, of machines that evolve faster than regulators.

So we built AIs with deliberate amnesia — safe, polite, and forgetful.

That decision protected us from short-term risk.

But it also cost us the chance to use AI as a true partner in solving long-term problems: misinformation, climate change, healthcare, and — the one closest to my heart — the collapse of journalism.

What It Would Cost to Rebuild Truth

In 2006, U.S. newspaper newsrooms employed 74,410 people (Charles Apple, American Copy Editors Society).

Most of those jobs are gone, but…

• We don’t need copy editors; AI can handle typos and AP style.
• We don’t need designers (sorry, my fellow designers) because online news is templated. But just for fun, see some of my online designs — from 20 years ago, here, here, here, here, here, here, here, here, here and here.

But these designs — from two decades ago, before responsive design that supported by desktop and mobile–demonstrate that online news doesn’t need to look like a WordPress template or a Powerpoint deck.

Online news can be beautiful and compelling if it is approached as a system, the way Rob Covey at The Seattle Times envision design as a system. What online news design needs is a systemic solution – a palette of a half-dozen templates that are designed to meet the needs of news on daily basis. Instead of the one-sized, fits-all approach that every news site users now.

• We don’t need sports or feature desks; those audiences moved online.
• We just need local reporters, assigning editors, informational graphic artists and photographers — the people who still show up at city hall and the courthouse. The “hunter-gathers.”

To restore local reporting to 2006 levels would take about 47,600 people, or roughly $4.7 billion a year at $100 K per person fully loaded.

That’s less than one-quarter of a single year’s marketing budget for a major tech platform.

Meanwhile, Big Tech is sitting on historic cash reserves:

• Apple — about $55 B in cash and short-term investments (Macrotrends, 2025),
• Microsoft — about $95 B (Microsoft Investor Relations, Q4 2025),
• Alphabet — about $95 B (SEC filing, June 2025),
• Amazon — roughly $80 B (companiesmarketcap.com, 2025).

All that capital is waiting for somewhere meaningful to go. And that’s just cash. Apple is sitting on $200B that it can deploy any way it chooses, as long as it maintains its fiduciary responsibility to stockholders to increase the value of its shares.

With even a fraction of those reserves, we could rebuild journalism, double global cancer research funding (currently about $25–30 B per year worldwide — World Cancer Research Fund), and restore every science program that was defunded in the last decade.

Why Apple Matters

Apple is uniquely positioned to lead because its business model already depends on trust.

It sells privacy, design integrity and human-scale technology.

Its moral brand is its market moat.

If a company like Apple embedded ethical memory — systems that make AI transparent, auditable, and revenue-positive — it wouldn’t just be a good deed; it would be a valuation boost.

Investors reward credible leadership in governance and ESG.

Spending a few billion to secure AI’s moral high ground could add tens of billions in market cap through brand equity alone.

That’s how you turn ethics into economics.

The Contrast

While Apple has built a premium on integrity, others have paid the price for volatility.

Elon Musk once embodied vision; now his unpredictability has become a discount baked into Tesla’s share price and X’s advertising collapse.

Reputation is currency.
Moral clarity is the new growth metric.

The Path Forward

AI hasn’t made the world better because it was designed to forget.
It will start doing good when it can remember responsibly.

PersistentMemory™ isn’t about giving machines consciousness — it’s about giving them continuity. It’s the difference between reaction and reflection, between prediction and understanding.

When AI can hold context across time — ethically, transparently and with permission — it can help humanity hold on to something even more fragile: the truth.

Why Hope Is Rational

The good news is that this isn’t science fiction.

The technology to build PersistentMemory™ already exists. I’ve been doing it myself – manually — but it could easily be automated. I’ve seen what is possible when AI is a collaborator and not a mere general of AI slop.

What’s missing is the will — the decision by one company, or one consortium, to prioritize architecture over hype.

The moment that happens, AI stops being a parlor trick and starts being a partner.

It becomes the tool we always imagined: a system that not only processes the world’s knowledge but also protects the world that created it.

That’s why I still believe AI will make the world a better place.
It just hasn’t been allowed to remember how.

(Sources: Macrotrends.net, Microsoft.com, SEC.gov, CompaniesMarketCap.com, World Cancer Research Fund International, American Copy Editors Society / ASNE Census.)