Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstall“Doomer” and “booster” are useful shorthand, not official camps. In the 2026 debate over artificial-intelligence risk, doomers emphasize the possibility that increasingly capable systems could cause catastrophic harm; boosters (sometimes called accelerationists) prioritize rapid development and the benefits of deployment, often viewing extinction warnings as exaggerated. The people placed under either label do not share one platform, and a substantial middle position supports innovation while demanding limits on unacceptable risks.
The most important disagreement is not whether AI can be beneficial. It is how much uncertainty society should tolerate, what evidence is sufficient for intervention, and who should set the rules.
As an Amazon Associate I earn from qualifying purchases.
What do “doomer” and “booster” mean?
The terms grew out of online arguments rather than formal organisations. A doomer is generally someone who gives serious weight to severe future outcomes from advanced AI, including loss of control or catastrophic misuse. A booster argues for fast progress because AI could produce large gains in science, health, productivity or living standards, and may regard sweeping restrictions as more dangerous than the technology itself.
Neither label tells you a person’s position on every issue. Someone can warn about frontier-system risks while supporting commercial development, or welcome powerful AI while backing mandatory testing. The labels also hide differences in time horizon: current harms such as fraud and cyberattacks are observable, while many arguments about superintelligent systems concern capabilities that do not yet exist.
#1 Best Overall
“Booster” can also cover several distinct instincts. Some advocates want fewer rules because they believe competition and experimentation produce the fastest public benefits. Others support regulation aimed at clear harms but oppose rules designed around hypothetical extinction scenarios. Similarly, a person called a doomer may be focused on technical alignment, misuse by states or criminals, labour disruption, concentration of power, or all of these at once. Asking what risk, timeframe and remedy a speaker means is more useful than accepting the label.
Who is described as an AI doomer?
A September 15, 2026 Euronews explainer places several prominent researchers and technology leaders among the risk-warning voices. That description reflects the article’s framing, not membership in a formal movement. The named figures do not constitute a worldwide census of people concerned about AI safety, and their views on probability, regulation and timelines are not identical.
Dario Amodei
Amodei, the chief executive of Anthropic, is associated with warnings that advanced AI could create very serious risks. Euronews also reports criticism that his company develops frontier systems while he discusses their dangers. That tension is central to the debate: building a technology and warning that it needs safeguards are not logically incompatible, but they invite scrutiny about whether proposed controls are strong enough, whether deployment is moving too quickly and whether a company can evaluate its own incentives fairly.
Geoffrey Hinton
Hinton is identified as another prominent warning voice. His inclusion reflects concern about the long-term consequences of increasingly capable systems, rather than a single published policy programme shared by every person called a doomer. A warning about extreme outcomes also does not settle what should happen to ordinary applications that pose more familiar risks.
Yoshua Bengio
Bengio is likewise presented as stressing the need to take advanced-AI risks seriously. The label should not be read as proof that he agrees with every pause proposal, timeline estimate or regulatory design associated with other risk-focused researchers. Serious concern can coexist with support for targeted experiments, transparency rules and continued work on beneficial systems.
Max Tegmark
Tegmark organised a 2023 letter calling for a six-month pause in development of the most advanced systems. That episode is often cited as an example of the precautionary position, but a past pause appeal does not establish a permanent opposition to all AI research or deployment. It does show the kind of threshold question that divides the debate: should developers stop when systems appear to be approaching poorly understood capabilities, or wait for stronger evidence of harm?
Who is described as an AI booster?
The same Euronews article groups several people as faster-development or skeptical voices. Their views differ substantially, especially on regulation. Calling them one bloc can obscure disagreements about federal authority, peer review, open research, commercial licensing and the seriousness of present-day harms.
Marc Andreessen
Andreessen is described as opposing regulation and arguing for an expansive approach to technological development. That is a governance preference, not evidence that he denies every present-day AI harm. A person can believe that competition will generate large gains while still accepting narrow rules against fraud, unsafe products or misuse of personal data.
David Sacks
Sacks is also described as opposed to regulation. Treating that stance as identical to every accelerationist argument would overstate the agreement among people placed in this broad tendency. Opposition to a particular law may reflect concerns about bureaucracy, national competitiveness, speech or enforcement, rather than a claim that no safety practice is needed.
Jensen Huang
Huang is a useful example of why the two-team framing breaks down. Euronews says he dismisses extinction claims while supporting a unified federal framework and industry peer review. A person can reject existential-risk forecasts and still favour common rules and independent scrutiny. His position illustrates that the crucial question is often not “regulation or no regulation?” but which risks should be regulated, by whom and with what technical access.
Yann LeCun
LeCun is described as viewing existential-risk warnings as overblown. That position concerns the credibility and scale of a particular risk category; it does not automatically answer questions about bias, labour disruption, privacy, security or concentration of power. Skepticism about extinction can coexist with concern about how systems are trained, deployed and controlled by institutions.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesThe cautious middle: risk limits without a blanket stop
Sam Altman’s September 2026 remarks are presented as a mixed position. He rejected both blind optimism and doomerism, and said competition by itself is not a reason for rash decisions. His formulation accepts that AI’s benefits are large while arguing that those benefits do not justify accepting any level of technological risk.
At the United Nations Security Council in September 2026, Altman said: “The industry must not accept too much technological risk just because the benefits are too great and that they feel too important to slow down.” This is not a call for a universal moratorium. It is an argument for slowing or constraining activities when the risk is unacceptable, even if the expected benefits are substantial.
This middle position overlaps with several international policy proposals: continue useful research and deployment, but require evidence, testing and accountability before systems with consequential capabilities are released. It also treats the distribution of benefits as part of safety. A safeguard that makes advanced tools available only to wealthy countries or dominant firms may reduce one exposure while creating another form of inequality.
Why experts disagree about existential risk
Disagreement is partly about facts and partly about decisions under uncertainty. The UK government’s 2024 interim international report says researchers differ in their expectations about how quickly capabilities will advance, how effective risk-limiting measures will be, and how society will respond. Those assumptions can lead to very different conclusions even when people agree on the underlying technical possibilities.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Different kinds of evidence
The International AI Safety Report 2026, written by more than 100 independent experts and advised by nominees from more than 30 countries and intergovernmental organisations, distinguishes between risks with strong empirical support and risks that depend on future capabilities. AI-generated media and cybersecurity vulnerabilities have robust evidence from observed systems and incidents. Scenarios involving advanced autonomous behaviour, large-scale loss of control or other future capabilities more often rely on modelling, controlled laboratory studies and theoretical analysis.
That does not make forward-looking analysis worthless. Policymakers routinely prepare for low-frequency, high-impact hazards before they can be observed at full scale. It does mean that confidence should be stated honestly: a model or laboratory demonstration is not the same as a documented catastrophe in the field. Evidence can justify precaution without justifying certainty.
The evidence dilemma
The 2026 report describes an “evidence dilemma.” Acting too early can produce ineffective or harmful interventions, lock in poor standards or deny societies useful tools. Waiting for conclusive evidence can leave people exposed while capabilities and deployment spread. There is no purely scientific answer to that trade-off because the acceptable balance depends on values, institutional capacity and who bears the downside.
Rank #3
For example, a government might require a test before a system is connected to critical infrastructure even if the test cannot predict every failure. A different government might judge that a costly approval process would delay medical or educational tools without materially improving safety. Both positions can claim to be evidence-based while assigning different weights to false alarms, missed dangers, compliance costs and access to benefits.
Free tools Windows power users keep installed
One-click scans. No signup required.
Different national contexts
The same report warns that risks can appear differently across countries. Adoption rates, infrastructure, labour markets, cybersecurity capacity and regulatory institutions vary. A measure that is workable for a well-resourced government may be difficult for a lower-income country, and a rule that reduces one danger could widen unequal access to beneficial systems.
The 2024 interim report was prepared with experts nominated by 30 countries, the European Union and the United Nations. That breadth does not create agreement on every forecast, but it underscores why a policy designed in one market cannot automatically be treated as a universal solution. Shared standards may reduce regulatory evasion, while financial and technical support may be needed so that smaller governments can perform evaluations rather than merely accept vendors’ claims.
A better way to compare the positions
Instead of asking which side a person is on, examine five dimensions.
| Dimension | Questions to ask | What the disagreement changes |
|---|---|---|
| Risk horizon | Is the focus on present misuse and social harms, or on future advanced capabilities? | Immediate safeguards can be justified by existing incidents; frontier controls may depend on forecasts. |
| Evidence standard | Does the argument rely on observed events, empirical tests, models, controlled studies or theory? | Higher uncertainty usually means wider disagreement about confidence and timing. |
| Precaution threshold | How much evidence is needed before testing rules, deployment limits or a pause? | A low threshold favours early intervention; a high threshold favours continued experimentation. |
| Governance preference | Are voluntary commitments enough, or are independent evaluations, shared standards and binding oversight required? | The choice determines who can inspect systems and enforce remedies. |
| Distribution | Who receives AI’s benefits, and who carries labour, infrastructure, security or safety costs? | A policy can reduce technical risk yet still be unfair if access and burdens are distributed unevenly. |
Two speakers may therefore agree that a model should be tested but disagree about the evaluator’s independence, the information the evaluator receives, the threshold for blocking release and who pays for the process. Those implementation details often matter more than broad labels.
What evidence is strongest today?
The 2026 safety report’s evidence base runs through December 2025. It is a 2026 synthesis, not a live incident tracker, so later events may not be included. Within that boundary, it gives the clearest support to harms already demonstrated in deployed or tested systems.
- AI-generated media: synthetic text, images, audio and video can enable deception, impersonation and information manipulation. The report treats this area as having robust empirical evidence. Useful responses can include provenance information, platform controls, media literacy and remedies for victims, although each has limitations.
- Cybersecurity vulnerabilities: AI systems can introduce or amplify security weaknesses; observed vulnerabilities provide a firmer basis for action than speculation about capabilities that have not appeared. Testing should include realistic attack attempts, logging and procedures for fixing newly discovered weaknesses.
- Future advanced-system risks: claims about loss of control, extreme autonomy or catastrophic misuse remain more dependent on models, experiments under controlled conditions and theoretical reasoning. Such work can identify dangerous pathways, but it cannot by itself prove that a catastrophe is inevitable.
The report does not supply a settled probability of catastrophe. Public arguments about “P(doom)” should therefore be read as individual estimates or scenario judgments, not as an agreed scientific statistic. A forecast is more informative when it identifies assumptions, a time horizon, uncertainty ranges and evidence that would update it.
Can a powerful AI also be aligned?
“Alignment” is another term that can conceal several questions. It may refer to whether a system follows a user’s legitimate instructions, respects a developer’s constraints, avoids harmful actions, remains corrigible when corrected or reflects broader social values. A system can perform impressively on benchmarks while failing one of these tests. Conversely, a system can be limited in capability and still cause harm through biased outputs, insecure integrations or irresponsible deployment.
The practical implication is that “powerful” and “aligned” should not be treated as mutually exclusive labels or as proven properties. Evaluators need to define the behaviour being tested, test under adversarial conditions, document uncertainty and monitor systems after release. Claims about alignment should specify the operating environment and the consequences if safeguards fail.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What policies sit between acceleration and a total moratorium?
A joint diplomatic statement hosted by the Netherlands on September 22, 2026 outlines a practical middle ground. It calls for measures that can be adopted without declaring that all advanced AI research must stop.
- Transparent safety protocols: developers should explain how they identify, measure and mitigate significant risks. Documentation should cover the system’s intended uses, known limitations, testing methods and escalation contacts.
- Mandatory pre-deployment testing: systems should be evaluated before release, especially when they could affect critical services or enable serious misuse. Testing should be proportionate to capability and context rather than a one-size-fits-all paperwork exercise.
- Independent evaluation with sufficient access: outside assessors need enough technical and operational access to test meaningful failure modes, rather than merely reviewing marketing claims. Independence is weakened if evaluators cannot inspect relevant model versions, tools, logs or deployment settings.
- Coordinated standards: compatible requirements reduce incentives to move risky development to jurisdictions with weaker rules. Coordination can also lower duplicated compliance work for firms operating across borders.
- Shared reporting of serious incidents: governments and developers should learn from failures and near misses instead of treating each event as isolated. Reports need enough detail to support prevention while protecting security-sensitive information and affected individuals.
The statement also says risk measures should not widen the gap between countries in access to AI’s benefits. These are diplomatic proposals, not evidence that a global regulatory system already exists. Turning them into practice would require funding, technical expertise, clear liability and processes for resolving disagreements among jurisdictions.
What companies are proposing
OpenAI’s 2025 recommendations call on frontier laboratories to share safety research and consider common safety principles and control-evaluation standards. The page also describes both benefits and potential catastrophic risks from superintelligent systems. Those recommendations are the company’s published position and forecast, not an independent scientific consensus.
Company commitments can move faster than legislation and help establish technical practices. They also raise obvious accountability questions: voluntary promises may change with leadership or commercial pressure, and companies may not be able to judge their own systems impartially. Independent access, incident disclosure and enforceable requirements address those weaknesses more directly. A sensible framework can combine internal safety work with outside scrutiny rather than treating either as sufficient alone.
How to evaluate a claim about AI risk
When a headline predicts either salvation or catastrophe, use this checklist:
- Identify the capability: What can the system demonstrably do now, and what is merely projected?
- Separate harm types: Fraud, bias, job displacement, cyber abuse and loss of control have different evidence and remedies.
- Check the time horizon: Is the claim about a current incident, a near-term forecast or a hypothetical advanced system?
- Ask what would change the author’s mind: Falsifiable benchmarks and pre-declared thresholds are more informative than absolute rhetoric.
- Inspect the proposed remedy: Does the recommendation involve testing, access controls, liability, standards, a pause or something else?
- Consider distribution: Who gains, who pays and which countries or communities are excluded?
- Check incentives and access: Is the speaker building, selling, regulating or auditing the system? What evidence can outsiders inspect?
- Distinguish a proposal from an implemented rule: A diplomatic statement or company pledge signals intent, but it is not proof of enforcement.
This approach avoids two common errors: dismissing real present harms because extinction is uncertain, and treating an uncertain future scenario as proof that every current AI project should stop. It also helps readers compare a technical forecast with a policy argument instead of treating one as a substitute for the other.
What readers should conclude
The doomer-versus-booster fight is not a vote between optimism and pessimism. It is an argument over evidence, acceptable uncertainty and institutional power. Risk-warning voices such as Amodei, Hinton, Bengio and Tegmark highlight severe possibilities; Andreessen, Sacks, Huang and LeCun are presented as more skeptical of existential claims or more supportive of rapid development, but they do not agree on every rule. Huang’s support for a federal framework and peer review illustrates that skepticism about extinction is compatible with oversight.
The strongest defensible position in 2026 is neither “nothing can go wrong” nor “catastrophe is inevitable.” It is to distinguish demonstrated harms from uncertain future scenarios, test systems before consequential deployment, make independent evaluation possible, share serious-incident information and design safeguards that do not reserve AI’s benefits for a few countries or firms. As the UK interim report puts it, “But nothing about the future of AI (artificial intelligence) is inevitable.” Choices about development and governance will shape which risks materialise and who receives the benefits.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Further reading
Readers seeking an accessible book-length introduction can start with Brian Christian’s The Alignment Problem: Machine Learning and Human Values, published in the listed 2021 paperback edition. It explains how machine-learning systems can diverge from human intentions and values. It predates the 2026 participants and policy proposals discussed here, so it is background rather than a current guide to the debate.
Frequently Asked Questions
Who are the AI doomers?
“Doomer” is an informal label for people who give serious weight to severe risks from advanced AI, including loss of control or catastrophic misuse. The September 15, 2026 Euronews explainer places Dario Amodei, Geoffrey Hinton, Yoshua Bengio and Max Tegmark among risk-warning voices, but they are not members of a formal organisation and do not share one complete platform.
Who are the AI boosters?
“Booster,” sometimes “accelerationist,” describes people who prioritise rapid AI development and its potential benefits, or who regard existential-risk warnings as overstated. Euronews places Marc Andreessen, David Sacks, Jensen Huang and Yann LeCun in a faster-development or skeptical grouping, while also noting important differences among them.
Why do experts disagree about existential risk and P(doom)?
They use different assumptions about how quickly capabilities will advance, how effective safeguards will be and how society will respond. Evidence for current harms is stronger than evidence for scenarios involving future capabilities, which often rely on models, controlled studies and theory. “P(doom)” is therefore an individual forecast or judgment, not a settled scientific statistic.
Recommended Free Tools
Can advanced AI be both powerful and aligned?
Potentially, but “aligned” must be defined and tested. It can mean following legitimate instructions, avoiding harmful actions, remaining corrigible or reflecting specified values. A system can pass capability benchmarks and still fail in one of these areas, so alignment claims require clear tests, adversarial evaluation and monitoring after deployment.
Does Jensen Huang support AI regulation?
The Euronews explainer says Huang dismisses extinction claims while supporting a unified federal framework and industry peer review. That combination shows why skepticism about existential risk does not necessarily mean opposition to all oversight.
Are the Netherlands-hosted safety proposals already global rules?
No. The September 22, 2026 joint diplomatic statement calls for transparent protocols, mandatory pre-deployment testing, independent evaluation, coordinated standards and shared serious-incident reporting. It is a policy proposal, not evidence that a worldwide enforcement system is already operating.
What is the strongest evidence for AI harm today?
The International AI Safety Report 2026 identifies robust empirical evidence for harms involving AI-generated media and cybersecurity vulnerabilities. Claims about extreme autonomy, loss of control and catastrophic future misuse remain more dependent on modelling, controlled laboratory work and theoretical analysis.
Is a six-month AI pause the same as opposing all AI?
No. Max Tegmark’s 2023 letter called for a six-month pause in development of the most advanced systems. A pause proposal expresses a precaution threshold for a particular stage of development; it does not by itself establish permanent opposition to every AI application, research project or deployment.
What should an independent AI evaluation include?
The diplomatic proposal calls for evaluators to have sufficient access to test meaningful failure modes. In practice, that means access appropriate to the risk, relevant model versions, tools or deployment settings, testing records and enough information to reproduce important findings. A review limited to marketing material is not meaningful independence.
The Bottom Line
Bottom line: “Doomer” and “booster” are shorthand for different risk instincts, not fixed teams. Current evidence is strongest for harms such as synthetic-media abuse and cybersecurity vulnerabilities, while claims about catastrophic loss of control remain more uncertain. A credible middle course combines continued innovation with pre-deployment testing, independent evaluation, shared standards, incident reporting and safeguards that distribute AI’s benefits fairly.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
