The people building AI earnestly believe that it could kill us all by the end of the decade.

Academic rigour, journalistic flair

Arts + CultureBusiness + EconomyEducationEnvironmentHealthPolitics + SocietyScience + TechWorldPodcastsInsights

Share article

Print article

Earlier this month, artificial intelligence (AI) researcher Jacob Coxon resigned from Anthropic after just four months. In an announcement on X, he stated:

The people building AI earnestly believe that it could kill us all by the end of the decade.

A senior member of Anthropic’s staff, Evan Hubinger, actually agreed with Coxon, adding he personally thinks the chance of this happening in the next decade is more than 10%.

Understandably, these statements made waves. There’s now lots of talk about slowing down AI research and increasing “human control” over the technology.

But how exactly might AI kill us all? There’s no shortage of fantastical scenarios, and most of them involve the concept of “superintelligent” AI – that is, AI that’s more capable than humans.

I’ve distilled these scenarios down to the top five, ordering them roughly from most vague to most precise. And I’d argue the list is also ordered from least probable to most probable.

1. We’ll never know

AI doomers often justify their concerns by means of an annoying catch-22 paradox: how can we possibly imagine what a superintelligence might do to take out less intelligent beings like us?

We’d have to be superintelligent to predict what a superintelligence would be able to do. It’s like asking your family dog to imagine thermonuclear war.

The good news here is that superintelligence is still perhaps some distance away. Current AI models are really good at solving particular problems, but that’s not the same as being more intelligent than a human in all domains.

However, AI did recently solve one of the seven most challenging maths problems known. It’s apparently closing in on others, which might leave you feeling less optimistic here.

---

Read more: Generative AI has changed mathematics forever. Where to from here?

---

2. Paperclips

A superintelligent AI would likely be extraordinarily competent at achieving its goals. But it might be indifferent to human survival.

A classic example of such indifference comes from Oxford philosopher Nick Bostrom’s imagined superintelligent AI that’s been designed to optimise paperclip production. To produce its preferred form of office supplies, it quickly converts all available matter – including humans, planets and stars – into paperclips.

What we have here is the perfect execution of improperly specified objectives. The AI doesn’t hate humanity; it simply recognises we’re composed of atoms that could be better utilised for paperclips. It’s not personal.

The good news here is that this scenario confuses intelligence with power. A superintelligent AI doesn’t necessarily have the power to achieve its goals. Turning the planet into paperclip factories would require planning permissions.

Even if it got the permissions, building too many paperclip factories would lead to inevitable public outcry. Interest groups would block the proceedings in the courts. Environmental activists would block the bulldozers.

3. Bioweapons

Humanity could be killed by a superintelligent AI making and releasing some dangerous new bioweapon into the atmosphere. This is, in fact, one outcome of the AI 2027 scenario by the AI Futures Project, a non-profit dedicated to forecasting the impacts of advanced AI.

This risk was made more concrete last month, when researchers at Stanford University announced they’d used a genetic language AI model to synthesise 16 new viruses.

Worryingly, they just sent the genetic sequences off to a mail-order lab and it sent the viruses back in test tubes. The whole experiment cost a couple of hundred thousand dollars at most.

The good news here is that it’s remarkably hard to kill everyone with a new virus. To do that, you need a virus that’s very transmissible, so it spreads far and wide. But it’s a rule of biology – viruses that spread easily are typically less fatal. By contrast, if a virus is very fatal, transmissibility tends to go down, as most people infected die before there’s time to spread the infection.

COVID killed less than 1% of humanity. The deadliest pandemic in recorded history was the Black Death, when the plague killed more than one-third of Europe’s population in the 13th century. However, even the plague would likely be much less deadly today due to our increased medical knowledge and better sanitation.

4. Nuclear war

What if AI got into the nuclear command and control chain and started a nuclear war? We’ve come close to nuclear war by mistake several times in the past 50 years.

We’re told that nuclear command and control is completely disconnected from the internet. But, as we saw in 2010, Iran’s nuclear centrifuges got taken out by a computer worm called Stuxnet, thought to have been brought in on a USB stick. AI can also give the military false intelligence, which could lead to irreparable actions.

The good news here is that nuclear stockpiles are down. But they are still enough perhaps to take out half of us. And it wouldn’t be by the nuclear blast itself, but the famine in the nuclear winter that would follow.

5. Other humans

Perhaps the most likely risk is that we take ourselves out. And AI might precipitate this.

Imagine – and it doesn’t take a lot of imagination – that AI causes massive job losses, pollutes the information space with misinformation, fractures our politics, and destroys human relationships with fake synthetic companionship.

Society might easily break. Slowly but surely, we’d stop being able to support human life at any scale.

What then to take away from all these scenarios? There are some things to be worried about for sure. But not to be too worried, I hope.

---

Toby Walsh is the author of God AI: boom or doom? What to expect when the machines outsmart us, published by La Trobe University Press.

Sep 26, 2026

Ten years ago, Nikole Hannah-Jones and her husband made a deliberate choice: instead of sending their daughter to an elite school, they enrolled her in a struggling, predominantly Black public school in Brooklyn. Hannah-Jones believed families with resources and influence had a responsibility to invest in schools too often abandoned by those with options. A decade later, she says the experience came at a painful cost. Her daughter fell behind academically, teachers failed to provide the support she needed, and at 16, she confronted her mother with a devastating question: why had she spent so much time thinking about other children instead of her own? Hannah-Jones has apologized to her daughter, but she still argues that segregation remains at the heart of the problem. Her critics see a different lesson — one about accountability, school performance, and what happens when principle collides with parenthood. Michael Smerconish laments that not all families can do what Hannah-Jones ultimately did – move their child to a better school and let the money follow the child.

How this was made

Auto-dubbed

Audio tracks for some languages were automatically generated. Learn more

Transcript

How to stay smart in the age of AI: the science of critical thinking

There is growing concern that AI can blunt memory and reasoning. But science shows ways to keep the brain sharp.

By

• Helen Pearson

• Email
• Bluesky
• Facebook
• LinkedIn
• Reddit
• Whatsapp
• X

Save article

Earlier this year, education researcher Lixiang Yan realized that students could outsource all of their work for his university course to artificial intelligence.

In praise of human teachers: universities must resist outsourcing everything to AI

Yan, who teaches at Tsinghua University in Beijing, China, was using an agentic AI called Codex to set up an online course on a learning-management platform. But it quickly became clear that the adept AI could independently complete all the assignments, answer questions and participate in discussions, “without the learner being intellectually engaged” at all, Yan says.

The experience made Yan think that teaching students to reason and think for themselves “is more critical than ever”, he says. And he is hardly alone. In a 2025 survey of more than 1,000 US faculty members, conducted by the American Association of Colleges and Universities, 90% said they thought that generative AI will reduce students’ critical-thinking skills. Around 20% of higher-education students who use AI said that they already find it harder to work through problems without it, according to more than 8,000 respondents to a global survey this year.

“Learning happens as a consequence of cognitive processing. Cognitive processing is typically effortful, and that’s what AI effectively removes,” says cognitive psychologist Daniel Willingham at the University of Virginia in Charlottesville. “People are really worried about that.”

And science — in which people need to critically evaluate theories and data — is one important area that’s under threat from the AI revolution in large language models (LLMs). ”There is a risk that students outsource precisely the activities through which scientific thinking develops: formulating hypotheses, interpreting evidence, considering alternatives and struggling with uncertainty,” says Thomas Nygren, who studies education and critical thinking at Uppsala University in Sweden.

But all is not lost. Specialists in education and cognition have been studying critical thinking and how to develop it for years, and some have useful tips on how to stay cognitively sharp in the age of AI. “The building of critical thinking and higher-order thinking skills have become even more important,” says Stéphan Vincent-Lancrin, who led a ten-year programme on critical thinking at the Organisation for Economic Co-operation and Development (OECD) in Paris.

What is critical thinking?

Ask five people to define critical thinking, and you’ll get five different replies. (I tried.) “Critical thinking is sort of the pinnacle” of human cognition, says Willingham. “It’s kind of a basket term that captures instances in which people are thinking well.”

Is AI ruining our skills? Early results are in — and they’re not good

One reason for the varied definitions is that the term is used widely in education and everyday discourse, but less by researchers who study cognition. Instead, they talk of separate high-level processes such as memory, reasoning, decision-making and problem-solving, which involve differing but overlapping mechanisms in the brain.

Some general strategies are common in critical thinking, say researchers. An important one is the ability to consider alternative possibilities or perspectives to our own — as a scientist would when weighing up different hypotheses.

But practising critical thinking requires skills and knowledge in the relevant domain — so being able to do it in one field doesn’t mean someone can in another. Neurologists, for example, are not particularly good at diagnosing cardiology cases, as one study showed¹.

Is AI dulling this skill?

Concerns that students are intellectually under par are not new. In 1983, a report by the US National Commission on Excellence in Education warned that many 17-year-olds lacked “the ‘higher order’ intellectual skills we should expect of them”.

Over the past few decades, critical thinking has become a goal in many education systems, says Vincent-Lancrin. It is now considered an educational priority by the United Nations cultural organization UNESCO and is part of school curricula in almost all OECD countries. He points to the example of a course in which students investigate the causes of type 2 diabetes, revising their ideas as they learn about different perspectives, such as the contributions of genes and environmental influences to the disease.

But this doesn’t mean the goal is being met. “There’s not much reason to think we were doing a fabulous job” of teaching critical thinking in the past, says Willingham, gloomily, “and now AI has come in and wrecked everything that we had achieved” for students today.

Teachers in schools and universities are brimming with stories of students who, they say, are poor at problem-solving and evaluating information themselves because they’re outsourcing the work to AI. And some research supports the idea that AI can erode skills.

A 2024 study found that secondary-school students were better at solving mathematics problems when given access to a ChatGPT-like generative AI — but after it was taken away, they performed worse than those who never had access to the tool².

Overall, research does not yet show clearly what impact generative AI is having on learning and critical thinking. This is because AI tools are relatively new and varied, the studies so far are conflicting, and too few rigorous or long-term studies have been done, researchers say³. “Tools can help but can also have a negative effect,” says Yan, who has assessed the literature. “It’s actually in both directions.”

Some studies suggest that AI tools can help people to learn if, rather than supplying ready-baked answers, they give feedback or tasks that encourages students to think and problem-solve. “It’s about how we structure the task,” Yan says.

But Willingham and Yan acknowledge that people tend to seek quick answers rather than using AI in ways that demand hard work. When Yan asked students to use an AI designed to prompt their own thinking, he saw some of them shut it down and switch to ChatGPT instead. “It’s a bit shocking, but that’s the truth, right?” he says.

Future research might help. In June, the Education Endowment Foundation (EEF), a non-profit research organization in London, announced that it would spend up to £2.5 million (US$3.4 million) on rigorous studies to investigate the impacts of generative AI on learning, understanding, memory and problem-solving.

“It’s good someone is doing these studies,” Willingham says — although he still expects the outlook to be grim. “It’s going take a while before you see this slippage, “he says, “but the slippage feels pretty inevitable.”

How do you teach critical thinking?

The fact that critical thinking differs between domains is one reason that teaching it is hard, say researchers. “It’s not generic and free-floating — you have to teach it within specific subjects,” says educational psychologist Paul Kirschner, who is at the Open University of the Netherlands and based in Heerlen. This means giving people the specific knowledge and reasoning skills needed in that field.

In history, thinking critically might involve corroborating records and artefacts, he explains, whereas doing so in science could require an understanding of the relationships between theory, hypothesis, method and evidence.

A second problem is that critical thinking often involves recognizing reasoning strategies in the abstract, and that’s hard. For instance, a student might learn in class about the difference between correlation and causation — but then fail to recognize one being mistaken for the other the next time they see it in the real world.

Willingham therefore recommends that teachers work out the reasoning skills that are most important, embed them in curricula and ensure students get plenty of practice, including on problems of different types. Another important aspect, says Kirschner, is having a coherent curriculum that builds knowledge, skills and critical thinking across different subjects and through the years.

The OECD project led by Vincent-Lancrin was one of the biggest efforts — involving some 800 schoolteachers — to scrutinize critical-thinking instruction in schools and develop and test materials, such as example lesson plans and syllabuses.

Are the Internet and AI affecting our memory? What the science says

Vincent-Lancrin acknowledges that there is not much rigorous evidence on what works best to teach critical thinking. That’s partly because the skills are harder to measure in a standardized way — which is necessary to test whether a strategy is working — than are mathematics or reading skills. The OECD is now working with an education company called Macat, in London, to develop a standardized assessment of critical thinking for secondary-school students and undergraduates.

Enjoying our latest content?
Log in or create an account to continue

• Access the most recent journalism from Nature’s award-winning team
• Explore the latest features & opinion covering groundbreaking research

Access through your institution

or

Sign in or create an account

Continue with Google

Continue with ORCiD

Nature 657, 873-875 (2026)

doi: https://doi.org/10.1038/d41586-026-02930-6

References

1. Remy R. M. J. P., Schmidt, H. G. & Boshuizen H. P. A. Cogn. Instr. 20, 27–45 (2002).Article Google Scholar 
2. Bastani, H. et al. Proc. Natl Acad. Sci. USA 122, e2422633122 (2025).Article PubMed Google Scholar 
3. Yan, L., Greiff, S., Lodge, J. M. & Gašević, D. Nature Rev. Psychol. 4, 435–436 (2025).Article Google Scholar 
4. Chesire, F. et al. Trials 25, 733 (2024).Article PubMed Google Scholar 
5. Kops, M., Schittenhelm, C. & Wachs, S. Comput. Human Behav. 169, 108650 (2025).Article Google Scholar 
6. Mendoza, G. A. S., Ballar, K. J., Yap, J. K. & Deinla, I. B. Media Asia 51, 470–488 (2024).Article Google Scholar 
7. Axelsson, C. A. W., Nygren, T., Roozenbeek, J. & van der Linden, S. J. Res. Technol. Educ. 57, 992–1018 (2023).Article Google Scholar 
8. McGrew, S. Curr. Opin. Psychol. 55, 101737 (2024).Article PubMed Google Scholar 
9. Eliseev, E. D. & Marsh, E. J. Appl. Cogn. Psychol. 37, 711–720 (2023).Article Google Scholar 

Download references

Reprints and permissions

Latest on:

• Machine learning
• Education
• Brain

• Anthropic’s AI biolab finds ‘CRISPR-like’ DNA in viruses. What’s next?News 25 SEP 26
• AI system helps lab devices ‘talk’ with each other — streamlining researchNews 24 SEP 26
• AI agent hacks government website for first time: why this breach mattersNews Explainer 24 SEP 26

Jobs

• University of Chinese Academy of Sciences (UCAS) Global Faculty RecruitmentWe are seeking distinguished scholars who will teach at UCAS while conducting research at CAS institute platforms—focusing first on over 20 key areas.Beijing (CN)University of Chinese Academy of Sciences (UCAS)
• Faculty Positions in Westlake UniversityFounded in 2018, Westlake University is a new type of non-profit research-oriented university in Hangzhou, China, supported by public a…Hangzhou, Zhejiang, ChinaWestlake University
• Open Rank Faculty Position (Tenure/Teaching/Research Stream) – School of Science and EngineeringShenzhen, ChinaThe Chinese University of Hong Kong, Shenzhen – School of Science and Engineering
• Faculty Positions in Chemical Biology, Westlake UniversityWe are seeking outstanding scientists to lead vigorous independent research programs focusing on all aspects of chemical biology including…Hangzhou, Zhejiang, ChinaWestlake University School of Life Science
• Faculty Positions in Center of Bioelectronic Medicine, School of Life Sciences, Westlake UniversitySLS invites applications for multiple tenure-track/tenured faculty positions at all academic ranks.Hangzhou, Zhejiang, ChinaWestlake University School of Life Science

Related Articles

• In praise of human teachers: universities must resist outsourcing everything to AI
• Is AI ruining our skills? Early results are in — and they’re not good
• Universities are embracing AI: will students get smarter or stop thinking?
• Are the Internet and AI affecting our memory? What the science says

Subjects

• Machine learning
• Education
• Brain
• Human behaviour

Sign up to Nature Briefing

An essential round-up of science news, opinion and analysis, delivered to your inbox every weekday.Email addressYes! Sign me up to receive the daily Nature Briefing email. I agree my information will be processed in accordance with the Nature and Springer Nature Limited Privacy Policy.

Nature (Nature)

ISSN 1476-4687 (online)

ISSN 0028-0836 (print)

nature.com footer links

About Nature Portfolio

• About us
• Press releases
• Press office
• Contact us

Discover content

• Journals A-Z
• Articles by subject
• protocols.io
• Nature Index

Publishing policies

• Nature portfolio policies
• Open access

Author & Researcher services

• Reprints & permissions
• Research data
• Language editing
• Scientific editing
• Nature Masterclasses
• Research Solutions

Libraries & institutions

• Librarian service & tools
• Librarian portal
• Open research
• Recommend to library

Advertising & partnerships

• Advertising
• Partnerships & Services
• Media kits
• Branded content

Professional development

• Nature Awards
• Nature Careers
• Nature Conferences

Regional websites

• Nature Africa
• Nature China
• Nature India
• Nature Japan
• Nature Middle East

• Privacy Policy
• Use of cookies
• Your privacy choices/Manage cookies
• Legal notice
• Accessibility statement
• Terms & Conditions
• Your US state privacy rights

© 2026 Springer Nature LimitedClose

Sign up for the Nature Briefing newsl

@cb_doge

·

3m

An honest documentary about Elon Musk would have shown the man:

• Who is working to return astronauts to the Moon and establish a permanent lunar base.

• Who is working to take humans to Mars and make humanity a multiplanetary species.

• Who is bringing broadband internet to underserved villages, schools, hospitals and isolated communities worldwide.

• Who has deployed thousands of satellite internet kits to reconnect first responders, hospitals and emergency services after disasters.

• Who is moving the world toward cleaner energy while advancing FSD to reduce accidents and protect lives.

• Who helped bring astronauts safely home after their eight-day mission became a 286-day stay in space.

• Whose companies employ more than 150,000 people and support over 500,000 indirect jobs.

• Whose companies have paid billions in salaries, created trillions in value for shareholders and helped thousands of employees become millionaires through company stock.

• Who is giving paralyzed people new independence by allowing them to control computers and robotic arms using their thoughts.

• Who is working on mind-controlled wheelchairs and devices that could help paralyzed people feed and care for themselves.

• Who is helping restore communication for people with ALS, stroke and other severe speech impairments.

• Who is working on technology that aims to restore sight for people who lost it and give vision to those born blind. That man is Elon Musk. Leaving all of this out makes it a hit piece, not a documentary.

New Rapid Brain Tumour Test Cuts Diagnosis Time to Hours

25 Sep 2026

General Healthcare Oncology View All News

Key Summary:

• The test can identify a brain tumour’s type from a small tissue sample in as little as two hours.
• Results during surgery could help surgeons decide how much tumour they can safely remove.
• The NHS pilot will expand to selected centres in Birmingham, London and Newcastle.

A NEW rapid genomic brain tumour test that could slash the wait for an accurate diagnosis from weeks to hours will be available to NHS patients.

The cutting-edge technology can quickly analyse the DNA or genetic code of a small sample of tumour taken either during surgery or biopsy, to provide doctors with accurate tumour type identification faster than ever.

World-First NHS Pilot

A world-first, NHS England says that the pilot will be extended across specialist centres across the nation and will mean that eligible patients will now receive a brain tumour diagnosis within days, if not sooner.

Brain tumours are notoriously difficult to diagnosis and treat due to the huge number of tumour types, ranging from slow-growing benign tumours to highly-aggressive and life-threatening cancers, each of which may behave differently and require different treatment.

Now, this new test will enable doctors to quickly specify tumour type and begin appropriate treatment as soon as possible – particularly in cases of fast-growing or high-grade tumours.

For patients having surgery to remove brain tumours where the exact type of tumour is nor yet know, the test will be able to supply results from samples within two hours, allowing surgeons to decide how to proceed with the remainder of the surgery and how much tumour they should safely remove.

Previously, patients have had to rely on diagnostic techniques such as CT and MRI scans, which are unable to offer accurate tumour type diagnosis, followed by a sample of their tumour being examined by microscope in a lab to determine its type – a process that can take weeks.

‘Transformation’ of Brain Tumour Diagnosis

Professor Frankie Swords, NHS medical director, said that this new rapid genomic technique has the potential to “completely transform” brain tumour diagnosis, calling it a “huge leap forward for patients”.

She said: “For people with suspected brain tumours, getting the right diagnosis quickly can feel like a race against time, while waiting weeks for answers can be agonising for them and their families.

“A faster diagnosis means they can start the right treatment or access clinical trials sooner, while for some patients it could mean surgeons can make potentially life-changing decisions about their surgery while on the operating table.

Professor Dame Sue Hill, Chief Scientific Officer for England and Senior Responsible Officer for Genomics in the NHS, agreed, saying: “This is another world-leading development for NHS genomics, taking pioneering science and testing how we can use it safely and consistently in patient care.

She added: “Our ambition is to build the evidence for this testing to become part of routine NHS care, so patients across the country can benefit equally from faster, more precise diagnosis.”

The first phase of the pilot roll-out will extend the testing to a select number of hospitals in Birmingham, London and Newcastle, while additional genomic laboratory sites will be introduced to more areas in the second phase.

NHS England is investing more than £2 million over two years in the Brain Cancer NHS Genomic Network of Excellence, who are behind the test.

Reference

NHS pilots rapid genomic test to transform brain tumour diagnosis. NHS England. 2026. https://www.england.nhs.uk/2026/09/nhs-pilots-rapid-genomic-test-to-transform-brain-tumour-diagnosis/

Featured image: Acronym on AdobeStockAuthor:

Cecilia Adamou

Each article is made available under the terms of the Creative Commons Attribution-Non Commercial 4.0 License.

Share:

More great content like this

– straight to your inbox >

Rate this content’s potential impact on patient outcomes

Average rating 5 / 5. Vote count: 1

1,447,228 views Sep 14, 2026 #oldmoney#history#celebrity

The Tragic Story of Eduard Einstein – Schizophrenic Son of Albert Einstein

Welcome to Rowsper! A channel dedicated to exploring the remarkable lives of famous people who lived with disabilities, rare medical conditions, physical differences, and extraordinary personal challenges. From individuals who overcame enormous barriers to achieve worldwide recognition to those whose conditions contributed to heartbreaking or tragic outcomes, Rowsper approaches each story with respect, empathy, and historical context.

Every video is carefully researched, written, edited, and produced with respect for the individuals discussed and in alignment with YouTube’s policies and Community Guidelines. Our videos are created for educational, documentary, and entertainment purposes. Subscribe to Rowsper as we uncover powerful stories of perseverance, achievement, adversity, and the complicated lives of some of history’s most remarkable people. #fame#celebrity#oldmoney#history#biograph

Sep 24, 2026

Angela Duckworth changed the way millions of people think about success with her bestselling book Grit. Now, nearly a decade later, she says there’s another major piece of the puzzle: your situation. In her new book, Situated: Find the People and Places That Bring Out Your Best, Duckworth explains why success isn’t simply a matter of intelligence, talent or willpower. The people around us, the environments we choose and even seemingly small details of our daily lives can shape what we think, feel and ultimately do. Duckworth joins Michael Smerconish to revisit the meaning of grit, explain her “situation → thoughts → response” model, and discuss why changing your circumstances can sometimes be more effective than trying harder to change yourself. Can grit be learned? How much control do we really have over our behavior? And have you set yourself up for success? Subscribe for more conversations with independent thinkers.

How this was made

Auto-dubbed

Audio tracks for some languages were automatically generated. Learn more

Trump on AI:

Super Intelligence, SI. I don’t call it AI anymore. I call it SI because it’s a much better name.

It’s a much more accurate name because when you say the other, it means it’s fake. Artificial means it’s fake, and it’s not fake. It’s very strong intelligence.

I think a lot of people are doing that now. I call it SI.

https://twitter.com/clashreport/status/2103854037393170864/video/1

Elon Musk says Tesla’s existing chip suppliers simply can’t expand fast enough to meet the company’s needs. https://x.com/MarsUniversityX/status/2103602260416593943/video/1 Samsung, TSMC, Micron and others are already supplying chips, and Musk says Tesla wants to buy everything they can produce. But if supply still isn’t enough, Tesla has another option: Build Terafab. “We need the chips. So we’re going to build Terafab.”

8:00 PM · Sep 26, 2026

https://twitter.com/cb_doge/status/2103555538256859200/video/1