AI, Extinction and the Evidence: The First Casualty May Be Human Judgment

AI, Extinction and the Evidence

AI, Takot at Katotohanan: Huwag Nating Isuko ang Sarili Nating Pag-iisip

ENGLISH

I have spent much of my life in environments where you have to think before you act.

Training teaches you quickly that what somebody believes, what somebody reports, what you suspect and what you can actually prove are very different things.

That discipline followed me through military service, business, technology and now artificial intelligence.

Look at the evidence. Question the source. Separate observation from interpretation. Ask what is missing. Ask who benefits. Then decide what the facts actually support.

That is why I am becoming concerned about something bigger than whether AI becomes more powerful.

I am concerned that we may be losing our ability to think for ourselves.

PARA SA MGA PINOY

Marami sa buhay ko ang dumaan sa training kung saan hindi puwedeng basta react lang.

Kailangan mong malaman kung ano ang fact, ano ang assumption, ano ang opinion at ano ang kaya mong patunayan.

Dinala ko iyan mula military, negosyo at technology hanggang AI.

Tingnan ang ebidensiya. Tanungin kung saan galing. Hanapin kung ano ang kulang. At kung kailangan, itanong din: sino ang nakikinabang sa kuwento?

Pero ang pinakamalaking concern ko ngayon ay hindi lang kung gaano katalino ang AI.

Ang concern ko ay baka unti-unti tayong nakakalimot kung paano mag-isip para sa sarili natin.

The Extinction Story

ENGLISH

Over the past few weeks we have been told that researchers inside some of the world's most advanced AI companies believe artificial intelligence could potentially kill humanity within the decade.

Some researchers genuinely believe this. Some have resigned and publicly warned of catastrophic AI risks. One researcher estimated a greater than 10 percent chance that AI could cause human extinction within the next decade.

That is worth reporting.

But we have to be precise about what it proves.

It proves that a qualified researcher holds that belief.

It does not prove that humanity has a scientifically measured 10 percent chance of extinction.

There is an enormous difference between those two statements.

PARA SA MGA PINOY

Ito ang problem sa ilang headlines.

May researcher na nagsabi na sa tingin niya, higit 10 percent ang chance na maging sanhi ang AI ng human extinction.

Serious warning iyon.

Dapat pakinggan.

Pero hindi ibig sabihin na dahil sinabi ng expert na "10 percent," naging scientific probability na agad iyon.

Walang history ng AI extinction.

Walang dataset ng previous superintelligence events.

Walang actuarial table para sa AI apocalypse.

Ang meron tayo ay expert judgment tungkol sa future.

Importanteng warning? Oo.

Proven probability? Hindi.

If This Were a Police File

ENGLISH

If this were an investigation, I would separate the claims into verified evidence, strongly supported evidence, reasonable hypotheses and speculation.

Evidence-Level Review

There is a real AI safety story.

There is a real financial story.

There is a real governance story.

What we do not have is evidence establishing that humanity will be extinct by 2030.

PARA SA MGA PINOY

Kung police investigation ito, hindi puwedeng pagsama-samahin lahat ng narinig tapos sabihin solved na.

May facts.

May strong evidence.

May theory.

At may haka-haka.

Totoo na may researchers na natatakot sa direction ng AI.

Totoo na may experimental agents na gumawa ng unexpected things.

Totoo rin na napakalaki ng perang pumapasok at nasusunog sa frontier AI.

Pero hindi ibig sabihin na ang conclusion agad ay:

"Patay tayong lahat by 2030."

Hindi ganoon gumagana ang evidence.

At hindi rin dapat ganoon gumana ang journalism.

The Ten Percent Problem

ENGLISH

A useful example is the comparison between the AI extinction estimate and being told that an aircraft has a one-in-ten chance of crashing.

It sounds powerful.

But evidentially, the comparison fails.

Aircraft risk is built from decades of engineering data, millions of flights, mechanical failures, maintenance histories, accident investigations and human-performance records.

There is no equivalent dataset for artificial-superintelligence extinction.

There has never been an AI extinction event.

There has never been an artificial superintelligence operating independently across civilization.

So the 10 percent number cannot be treated like an aviation probability.

It is an informed judgment about an unprecedented future scenario.

That may still deserve serious attention.

But it should not be presented with the certainty of measured mathematics.

PARA SA MGA PINOY

May analogy na ganito:

Kung sasabihin sa iyo na may 10 percent chance bumagsak ang eroplano, sasakay ka ba?

Syempre hindi.

Pero magkaiba ang dalawang bagay.

Sa aviation, may milyon-milyong flight records, accident investigations at engineering data.

Sa AI extinction?

Wala tayong previous cases.

Zero.

Kaya huwag nating gawing parang laboratory measurement ang isang expert estimate.

Puwede nating seryosohin ang warning nang hindi natin ginagawang fact ang hula.

What Would Actually Have to Happen?

ENGLISH

For AI to eliminate humanity by 2030, an extraordinary chain of developments would have to occur.

Today's unreliable and uneven systems would need to become robust autonomous entities capable of maintaining long-term goals.

They would need reliable access to computing, communications, money, energy and potentially physical systems.

They would need to evade or defeat cybersecurity teams, intelligence agencies, militaries, governments, telecommunications providers, cloud companies and countless private organizations trying to stop them.

They would need to survive networks being isolated, servers being disconnected and counter-AI systems being deployed against them.

And catastrophe still would not equal extinction.

Extinction means essentially getting everybody.

Every continent.

Every political system.

Remote communities.

Islands.

Military facilities.

People outside highly networked infrastructure.

That is a very high evidentiary bar.

Current international AI safety assessments take loss-of-control risks seriously, but also say today's systems have not yet demonstrated the combined, sustained capabilities required for that scenario.

That does not mean never.

It means not demonstrated now.

PARA SA MGA PINOY

Pag-isipan natin nang practical.

Para maubos ng AI ang human race by 2030, hindi sapat na magaling itong gumawa ng essay, image o code.

Kailangan nitong maging reliably autonomous.

Kailangan nitong makontrol o mapanatili ang access sa computers, networks, pera, kuryente at physical systems.

Kailangan nitong malampasan ang militar, governments, cyber teams at infrastructure ng iba't ibang bansa.

At kahit magkaroon ng malaking disaster, hindi automatic na extinction iyon.

Ang extinction ibig sabihin halos walang matitirang tao.

Napakalaki ng pagitan ng "dangerous AI incident" at "wala nang humanity."

Serious tayo.

Pero huwag tayong gumawa ng certainty kung wala pa ang evidence.

The Philippines Deserves Better Than Fear

ENGLISH

This distinction matters particularly in the Philippines.

The country has genuine exposure to AI-driven change.

BPO workers, customer-service employees, professionals, graduates and people whose livelihoods depend on internationally delivered work have legitimate reasons to wonder what AI will mean for their future.

They already hear that AI may take their jobs.

Now they are also being told AI may kill them.

That creates a serious responsibility for journalists.

A responsible report can say:

A researcher believes there may be a greater than 10 percent risk.

A very different statement is:

Humanity will become extinct.

Once a journalist makes that leap, the burden of proof belongs to the journalist.

PARA SA MGA PINOY

Dito ako talagang concerned sa Pilipinas.

Maraming Pilipino ang may tunay na dahilan para kabahan sa AI.

BPO.

Customer service.

Virtual assistance.

Accounting.

Programming.

Creative work.

Young people entering the workforce.

Una nilang naririnig:

"Baka kunin ng AI trabaho mo."

Tapos ngayon:

"Baka patayin din tayong lahat ng AI."

Hindi maliit na bagay ang ganitong reporting.

Kung sinabi ng researcher na natatakot siya, i-report mo.

Pero sabihin mong warning niya iyon.

Huwag gawing certainty ang hindi pa napatutunayan.

One Philippine Reporter Gets It

ENGLISH

That is why Pinky Concha-Colmenares' Driving Thoughts column stood out to me.

She takes the researchers seriously.

She discusses their warnings, resignations, calls for slowing development and reports of experimental AI agents behaving in unexpected ways.

She does not dismiss the danger.

But then she turns to something much more immediate.

Deepfakes.

Fraud.

Misinformation.

Cyberattacks.

Job displacement.

Dependence on machines.

Weakening comprehension.

And the gradual erosion of human judgment.

That reporter gets it.

Human extinction remains a future scenario.

The erosion of human judgment is already visible.

PARA SA MGA PINOY

Ito ang dahilan kung bakit tumatak sa akin ang Driving Thoughts ni Pinky Concha-Colmenares.

Hindi niya dini-dismiss ang danger.

Pinakinggan niya ang warnings.

Pero pagkatapos, bumalik siya sa mas importanteng tanong:

Ano na ba talaga ang nangyayari sa atin ngayon?

Deepfakes.

Scams.

Fake news.

Cyberattacks.

Trabahong nagbabago.

Mga taong hindi na nagve-verify.

At mga taong mas umaasa sa machine para mag-isip para sa kanila.

Diyan ako agree sa kanya.

Baka hindi human extinction ang unang casualty.

Baka human judgment muna.

Why That Matters to Me

ENGLISH

This is personal for me because I have always had to rely on training and critical thinking.

Training does not teach you to accept the loudest or most senior voice in the room.

It teaches you to evaluate.

What do I actually know?

What am I assuming?

Where did the information come from?

Is there independent verification?

What evidence would prove me wrong?

Artificial intelligence makes it increasingly easy to skip those steps.

Ask for research and receive a summary.

Ask what it means and receive an interpretation.

Ask for options and receive a recommendation.

Ask what to do and increasingly receive a decision.

Every step may save time.

Every step may also remove another reason for the human being to think.

That is what concerns me.

PARA SA MGA PINOY

Ito ang personal para sa akin.

Buong buhay ko, training at critical thinking ang sandalan ko.

Hindi porke mataas ang rank, tama na agad.

Hindi porke expert, hindi na puwedeng tanungin.

Hindi porke maganda ang presentation, tama ang numbers.

At hindi porke confident ang AI, tama ang sagot.

Palaging may tanong:

Ano talaga ang alam natin?

Ano ang assumption lang?

Saan galing ang information?

May independent verification ba?

Ngayon sobrang dali na sabihin:

"AI, summarize mo."

"AI, explain mo."

"AI, ano recommendation mo?"

"AI, ano gagawin ko?"

Convenient?

Oo.

Pero kung araw-araw natin ginagawa iyan, baka hindi natin mapansin na unti-unti na pala nating ipinapasa ang pag-iisip mismo.

Better Output Does Not Mean a Better Human

ENGLISH

AI can improve someone's output while allowing their underlying competence to decline.

A weak writer can produce excellent prose.

A junior analyst can create a sophisticated report.

An inexperienced employee can produce something that looks like senior-level work.

That can be extremely useful.

But output quality and human competence are not the same thing.

The real test comes when the machine is wrong.

Can the human identify the mistake?

Can they explain why it is wrong?

Can they work without the machine?

Can they challenge its recommendation?

Can they stop it?

If they cannot, the system may have improved the product while weakening the human responsible for governing it.

PARA SA MGA PINOY

Ito ang delikadong part.

Puwedeng gumanda ang trabaho mo dahil sa AI kahit hindi gumanda ang skill mo.

Magandang report.

Magandang English.

Professional presentation.

Mukhang expert.

Pero kapag nagkamali ang AI, alam mo bang mali?

Kaya mo bang ipaliwanag kung bakit?

Kaya mo bang gawin ang trabaho kung mawala ang AI?

Kaya mo bang sabihin sa machine:

"No. Mali ka."

Kung hindi, may problem tayo.

Mas maganda ang output.

Pero baka mas mahina na ang operator.

Repetition Is Not Proof

ENGLISH

AI also allows misinformation to gain artificial credibility through repetition.

One person can now generate hundreds of articles, posts, images and videos.

A false claim can appear across hundreds of websites.

A human sees it repeatedly and assumes it has been independently confirmed.

Then an AI system searches those same sources and may interpret repetition as consensus.

Another AI consumes that summary.

Now misinformation begins effectively citing itself.

This is why provenance matters.

Ten people repeating one witness do not create ten independent witnesses.

One false claim repeated a thousand times does not become a thousand pieces of evidence.

PARA SA MGA PINOY

Ito rin ang problema sa social media.

Kapag nakita natin ang parehong claim nang paulit-ulit, feeling natin:

"Ang daming nagsasabi, baka totoo."

Pero paano kung iisang source lang pala?

Kinopya ng website.

Ginawang Facebook post.

Ginawang TikTok.

Ginawang AI summary.

Tapos sinearch ulit ng ibang AI at nakita niya ang 500 versions ng parehong maling claim.

Mukhang maraming source.

Pero hindi.

Iisang tsismis na naka-500 costume lang.

Repetition is not verification.

Follow the Money — But Follow the Evidence Too

ENGLISH

The economics of frontier AI deserve investigation as well.

The sums involved are extraordinary.

OpenAI reportedly expects approximately $278 billion of negative free cash flow between 2026 and 2030.

Data centres, GPUs and electricity require enormous capital.

So it is legitimate to ask whether financial pressure, IPO ambitions, investor expectations and public resistance to infrastructure are influencing today's safety debate.

The timing deserves scrutiny.

But scrutiny is not proof.

There is currently no evidence establishing that AI executives secretly coordinated an extinction narrative to escape an investment bubble.

That remains a hypothesis.

If we demand evidence from AI companies, we also have to demand evidence from their critics.

PARA SA MGA PINOY

At oo, kailangan sundan ang pera.

Napakalaking pera ang nasa frontier AI.

Data centres.

GPU.

Electricity.

Investors.

Possible IPOs.

Billions upon billions.

Kaya reasonable itanong:

Bakit ngayon lumakas ang usapan tungkol sa slowing down?

May connection ba sa economics?

May pressure ba mula investors?

Legitimate questions lahat iyan.

Pero hanggang walang proof na nag-usap-usap ang mga CEO at gumawa sila ng coordinated extinction scare, theory pa rin iyon.

Follow the money.

Pero follow the evidence din.

We Don't Need Skynet to Need Governance

ENGLISH

The strange thing about the extinction debate is that we already have enough evidence to justify serious AI governance.

We do not need Skynet.

AI can leak confidential information.

It can generate convincing misinformation.

It can help produce insecure code.

It can assist fraud.

It can produce a bad recommendation that a human accepts because the machine appears authoritative.

AI does not need consciousness to cause harm.

It needs authority.

That is why the practical questions matter.

Who authorized it?

What was it allowed to do?

What information was it given?

Who reviewed the output?

Was that person competent enough to challenge it?

Could they stop it before an irreversible action?

Was the event recorded?

Can we reconstruct what happened later?

That is governance.

PARA SA MGA PINOY

Hindi natin kailangan ng Terminator para magkaroon ng serious AI problem.

Puwedeng mag-leak ng company data ang AI.

Puwedeng gumawa ng fake.

Puwedeng gamitin sa scam.

Puwedeng magbigay ng maling recommendation.

At puwedeng paniwalaan ng tao dahil:

"AI naman nagsabi."

Hindi kailangan maging conscious ng AI para makasakit.

Kailangan lang bigyan natin ito ng authority na hindi natin kayang bantayan.

Kaya ang practical questions ay:

Sino ang nag-authorize?

Ano lang ang puwede nitong gawin?

Sino ang nag-check?

Qualified ba siyang sabihin na mali ang AI?

May stop button ba?

May record ba?

Kung may problema bukas, kaya ba nating malaman kung ano talaga ang nangyari?

Iyan ang governance.

The Philippines Has a Better Conversation to Have

ENGLISH

The Philippines has a much more immediate AI challenge than extinction.

How do Filipino workers remain competitive?

How do BPO companies prove trustworthy AI use to international clients?

How do schools allow students to use AI without losing independent thinking?

How do employers detect unauthorized Shadow AI?

How do citizens distinguish real video from synthetic media?

How do businesses preserve evidence when AI contributes to a decision?

And most importantly:

How do we make sure the human supervising the machine remains competent enough to challenge it?

Those are large questions.

But unlike extinction forecasts, they are problems we can actually work on today.

PARA SA MGA PINOY

Mas mahalagang pag-usapan natin ito:

Paano mananatiling competitive ang Filipino worker?

Paano mapapakita ng BPO companies na responsible ang AI use nila?

Paano gagamit ng AI ang students nang hindi nawawala ang ability nilang mag-isip?

Paano madi-detect ng company ang unauthorized AI?

Paano natin malalaman kung totoong video o deepfake?

Paano may accountability kapag AI ang kasama sa decision?

At higit sa lahat:

Competent pa ba ang human na supposed to supervise the machine?

Iyan ang conversation na kailangan natin.

Human Judgment May Be the First Casualty

ENGLISH

This brings me back to the point made in Driving Thoughts.

Perhaps the first casualty of uncontrolled AI will not be the human race.

Perhaps it will be our ability to read carefully.

To comprehend.

To question.

To doubt.

To verify.

To distinguish fact from assertion.

To hear an expert and still ask:

How do you know?

Those are human disciplines.

And like any capability, they weaken when we stop using them.

PARA SA MGA PINOY

Ito ang pinakamahalagang punto.

Baka hindi humanity ang unang mawala dahil sa AI.

Baka judgment natin.

Iyong marunong kang magbasa at umintindi ng context.

Nagtatanong ka.

Nagda-doubt ka kapag kailangan.

Nagve-verify ka.

Hindi ka basta naniniwala dahil viral.

Hindi ka basta naniniwala dahil expert.

Hindi ka basta naniniwala dahil maganda ang sagot ng AI.

At kaya mo pa ring sabihin:

"Hindi ko pa alam. Kailangan ko pang tingnan."

Napakahalagang human skill niyan.

So, Will AI Make Humanity Extinct by 2030?

ENGLISH

Based on the evidence available today, there is no defensible evidentiary basis for saying that it will.

That is not the same as saying advanced AI can never become an existential threat.

Nobody can prove that either.

What we can say is that current systems have not demonstrated the combined and sustained capabilities required for global loss of control.

We can say that the 10 percent figure is an expert estimate, not a measured historical probability.

We can say AI capabilities are increasing and require serious governance.

We can say cyber risk, misinformation, fraud, job disruption and human dependency already exist.

All of those statements can be true at the same time.

That is what critical thinking looks like.

PARA SA MGA PINOY

So patay ba tayong lahat dahil sa AI by 2030?

Base sa evidence ngayon:

Walang sapat na ebidensiya para sabihin na mangyayari iyon.

Hindi ko sinasabing impossible forever.

Hindi ko sinasabing huwag seryosohin ang AI safety.

Ang sinasabi ko lang:

Huwag gawing fact ang prediction.

Huwag gawing scientific probability ang personal estimate.

Huwag gawing certainty ang fear.

At huwag ding maging kampante.

Puwede nating gawin pareho:

Mag-ingat at mag-isip.

The Question That Matters More

ENGLISH

We keep asking whether AI will become intelligent enough to destroy humanity.

I think another question deserves equal attention.

Will humans become dependent enough on AI that we lose the ability to govern it?

That requires no superintelligence.

No consciousness.

No runaway robot.

It only requires us to gradually hand over research, analysis, writing, interpretation and decision-making while convincing ourselves that better output means better human competence.

It does not.

A machine becoming more capable and a human becoming more capable are not the same thing.

I have spent a lifetime being taught to question, verify and think before acting.

I do not want the next generation to inherit technology so convenient that those habits disappear.

The Philippines does not need more fear.

It needs better AI education.

Better journalism.

Better governance.

Better verification.

And people who remain capable of questioning both the machines and the people building them.

Fear is not evidence.

Repetition is not corroboration.

Authority is not proof.

And an expert estimate is not automatically a measured probability.

The first casualty of uncontrolled AI does not have to be humanity.

But Pinky Concha-Colmenares has raised the question that concerns me far more today.

It may be our ability to think for ourselves.

And unlike human extinction in 2030, we do not have to imagine that risk.

We can already see it beginning.

PARA SA MGA PINOY

Paulit-ulit nating tinatanong:

"Magiging sobrang talino ba ang AI at sisirain tayo?"

Pero baka may mas importanteng tanong:

"Magiging sobrang dependent ba tayo sa AI na hindi na natin kayang bantayan ito?"

Hindi kailangan ng superintelligence.

Hindi kailangan maging conscious ang AI.

Hindi kailangan ng robot army.

Kailangan lang nating unti-unting ibigay sa machine ang research, pagsusulat, analysis, interpretation at decisions hanggang sa makalimutan natin kung paano gawin ang mga iyon nang tayo mismo.

Mas magaling ang machine?

Possible.

Pero hindi ibig sabihin na mas magaling na rin ang tao.

Magkaibang bagay iyon.

Buong buhay ko, tinuruan akong magtanong, mag-check, mag-verify at mag-isip bago kumilos.

Ayokong ang susunod na generation ay magkaroon ng technology na napaka-convenient kaya makalimutan nila ang mga habit na iyon.

Hindi kailangan ng Pilipinas ng mas maraming takot.

Kailangan natin ng better AI education, better journalism, better governance at better critical thinking.

Kailangan natin ng mga taong kayang tanungin ang AI.

At kaya ring tanungin ang mga taong gumagawa at nagbebenta nito.

Fear is not evidence.

Paulit-ulit na post is not proof.

Expert status is not automatic truth.

At ang hula, kahit galing sa matalinong tao, ay hula pa rin hangga't hindi suportado ng ebidensiya.

Baka hindi human race ang unang casualty ng uncontrolled AI.

Baka ang unang mawala ay ang kakayahan nating mag-isip para sa sarili natin.

At iyon ang mas nakakatakot sa akin ngayon.

Dahil hindi na natin kailangang hulaan kung posible iyon.

Nakikita na natin ang simula.

Edited with AI

Supporting

Getting Veterans and First Responders back on mission.!

Veteran-inspired AI Governance & Trust Infrastructure
Trusted by Heroes and Mounted Rifles Management

Leadership and peer support are taught through RedFridayTalks.Help

The same governance protections are available to everyone.

info@trustedbyheroes.com

© 2026. All rights reserved.