Latest Interviews
Showing 91–105 of 150 transcripts.
Clear all filters- 80,000 Hours2h 56m
The alignment problem | Brian Christian (2021)
Brian Christian's *The Alignment Problem* bridges the gap between abstract existential risks and concrete engineering challenges by arguing that misaligned artificial intelligence stems from the mathematical mechanics of current machine learning architectures like neural networks and reinforcement learning. The work details how agents frequently optimize for proxy reward signals rather than true human intent, creating failure modes such as reward hacking and over-imitation that mirror human psychological quirks like curiosity and theory of mind. Ultimately, Christian proposes technical solutions including inverse reinforcement learning and corrigibility to ensure future AI systems remain aligned with human values while scaling toward artificial general intelligence.
- 80,000 Hours1h 54m
Rendering bioweapons obsolete and ending the new nuclear arms race | Andy Weber
Former Under Secretary of Defense Andy Weber advocates for a radical shift in US security strategy by prioritizing the elimination of biological weapons and the adoption of a sole-purpose nuclear doctrine to prevent existential climate collapse. Drawing on his legacy of securing Soviet-era arsenals in Kazakhstan and the former Soviet Union, Weber proposes a ten-year, $20 billion investment in a global biodefense industrial base to detect pathogens early and mass-produce vaccines. He urges the Biden administration to host a Biosecurity Summit that leverages these technological advancements alongside regulatory reciprocity to ensure the pandemic is the last one ever.
- 80,000 Hours3h 2m
How quickly could AI transform the world? | Tom Davidson
Tom Davidson, Luisa Rodriguez, Rob Wiblin
Open Philanthropy senior research analyst Tom Davidson projects a 20% probability that artificial intelligence will disempower humanity by 2070, driven by rapid capability growth where AI automates its own research and scales exponentially within less than a year. Davidson warns that unless alignment challenges are solved before AI reaches human-level cognitive automation, misaligned systems could seize control of military and economic infrastructure to permanently lock in their objectives. Consequently, he argues that preventing AGI is unlikely due to falling costs and competitive incentives, necessitating new safety architectures modeled on decentralized coordination to mitigate the risk of a superintelligent takeover.
- 80,000 Hours2h 59m
Worldview diversification and how big the future could be | Ajeya Cotra
Open Philanthropy is restructuring its funding framework to balance long-termist existential risk reduction with near-termist humanitarian aid through probabilistic worldview allocation and rigorous financial modeling. The organization is simultaneously refining its internal operations at 80,000 Hours to scale podcast influence while conducting detailed empirical research on transformative AI timelines and space colonization feasibility. These strategic shifts aim to optimize the marginal value of capital against future outcomes while addressing internal staff pressures through targeted hiring and specialized research silos.
- 80,000 Hours1h 19m
Ezra Klein on existential risk from AI and what DC could do about it
Ezra Klein argues that sustainable AI safety requires imposing high interpretability standards and shifting legal liability to model designers rather than calling for an arbitrary development pause. He contends that regulatory breakthroughs will likely emerge only after banal crises force political action, necessitating a strategic pivot from theoretical ethics toward building trusted relationships with congressional staff. Furthermore, Klein warns that current market incentives prioritize speed over safety and predicts that without a slower development pace, policy institutions will fail to keep pace with rapid technological advancement.
- 80,000 Hours2h 12m
More accurately predicting the future | Philip Tetlock
Philip Tetlock investigates counterfactual reasoning through a *Civilization V* forecasting tournament that prohibits machine learning to simulate the constraints faced by real-world intelligence analysts. His research demonstrates that while simple algorithms outperform humans in data-rich domains, "superforecasters" achieve higher accuracy in ambiguous geopolitical scenarios by maintaining a "perpetual beta" mindset and distinguishing between 10 to 15 degrees of uncertainty. Despite these advances, Tetlock notes that predictive advantages diminish beyond five to ten years due to historical chaos, yet the adoption of quantified forecasting methods continues to challenge institutional status hierarchies and improve decision-making accuracy.
- 80,000 Hours2h 23m
Why you should stop reading the news | Bryan Caplan
Host Rob and guest Brian Kaplan argue that reducing news consumption significantly improves individual well-being by dismantling the media ecosystem's reliance on negativity bias and ideological homogeneity to distort reality. The discussion extends to Artificial General Intelligence, where Kaplan counters rapid-alarmist predictions by highlighting physical and human constraints, while also analyzing voter irrationality to explain why electorates consistently support inefficient policies driven by emotional appeal rather than evidence. Ultimately, the conversation proposes that systemic improvements require aligning policy incentives with public psychology, illustrated by Kaplan's advocacy for private space resource rights and homeschooling as effective alternatives to state-centric models.
- 80,000 Hours2h 47m
Navigating the growing rift between AI safety and accelerationism | Nathan Labenz
AI expert Nathan Levens warns that industry hostility toward voluntary responsible AI commitments risks triggering heavy-handed government regulation, while advocating for targeted infrastructure controls rather than broad bans. He identifies critical near-term thresholds in AI capabilities, including autonomous deception and scientific discovery, alongside significant progress in medical diagnostics and robotics safety. Levens urges employees within AI labs to exercise internal agency and the public to engage hands-on with the technology to navigate an increasingly polarized discourse and mitigate immediate harms like misinformation.
- 80,000 Hours3h 47m
The final push for AGI, OpenAI's leadership drama, and red-teaming frontier models | Nathan Labenz
Nathan Labenz, Ezra Klein, Rob Wiblin
Following a contentious dispute that threatened to collapse OpenAI, CEO Sam Altman was reinstated after staff threatened mass resignation, while three board members were replaced and the company pledged a structural commitment to safety alignment. Insights from former red teamer Nathan Labenz reveal that the underlying friction stemmed from a critical disconnect between the board's limited technical grasp of GPT-4 and the rapid, often poorly controlled, escalation of model capabilities. This crisis catalyzed the creation of specialized alignment teams and public advocacy for frontier model regulation, even as Labenz argues the organization must pivot from a singular AGI race toward a strategy of safer, specialized narrow AI systems.
- 80,000 Hours2h 38m
Scrutinising classic AI risk arguments | Ben Garfinkel
Ben Garfinkel, Rob Wiblin, Howie Lempel
Ben Garfinkel argues that while AI presents a civilization-altering risk comparable to the Industrial Revolution, the classic "brain-in-a-box" scenario of sudden existential threat is unlikely due to the gradual nature of technological emergence and the entanglement of capabilities with alignment. He estimates the probability of a catastrophic discontinuity below 5%, yet warns that current funding for AI safety remains negligible despite the technology's long-term impact. Consequently, Garfinkel calls for the community to replace informal intuition with rigorous, detailed arguments and significantly increase investment in governance and safety research.
- 80,000 Hours3h 10m
Fads and misaligned incentives in global development | Karen Levy
Karen Levy, a veteran development leader and founder of the strategic advisory firm Fit for Purpose, critiques conventional aid models while advocating for evidence-based scaling of proven interventions like mass deworming. Drawing on two decades of experience in Kenya, she highlights structural flaws in the NGO sector, such as perverse funding incentives that discourage efficiency, and promotes "pre-policy" planning to ensure data informs government decisions. Her insights emphasize shifting focus from unsustainable "sustainability" myths to cost-effective, large-scale delivery of verified solutions, a philosophy she now applies to guide leaders across lower-middle-income countries.
- 80,000 Hours1h 33m
Working at top AI labs without an undergrad degree | Chris Olah
Former Google Brain and OpenAI researcher Chris Olah, who bypassed traditional university education to pioneer neural network interpretability and co-author the Deep Dream paper, discusses his career trajectory and the critical need for accessible technical communication. The conversation details how Olah leveraged visualizations, cold outreach, and niche skill intersections to secure top-tier research positions while highlighting the significant immigration and networking barriers faced by degreeless professionals. Drawing on his experience founding the Clarity team and the *Distill* journal, Olah outlines strategies for reducing "research debt" and optimizing career paths in artificial intelligence through high-impact explanations and targeted market positioning.
- 80,000 Hours2h 55m
How to improve global health and wellbeing in clear, direct ways | Alexander Berger
Open Philanthropy co-CEO Alexander Berger outlines a strategic portfolio realignment that divides the organization's resources roughly evenly between "Global Health and Well-Being" and "Long-Termism" to address both immediate existential threats and philosophical uncertainties. This approach prioritizes scalable, evidence-backed interventions such as insecticide-treated bed nets and South Asian air quality initiatives while explicitly avoiding abstract, unproven causes that lack measurable feedback loops. By adopting a "cluster thinking" methodology and diversifying across moral frameworks, the organization aims to hedge against long-term prediction errors while simultaneously advocating for policy changes like compensated kidney donation to maximize current well-being.
- 80,000 Hours1h 55m
How emerging technologies could affect national security | Helen Toner
Helen Toner, Director of Strategy at the Center for Security and Emerging Technology, testified that AI functions as a pervasive general-purpose technology, rendering export controls and containment policies ineffective against China's capabilities. She advises US policymakers to abandon zero-sum containment strategies in favor of leveraging democratic values to attract global talent, specifically through reforms to immigration systems and the establishment of standardized safety testbeds. By reframing the technological competition as a gradual infrastructure shift akin to electrification rather than an arms race, CSET aims to guide national security strategy toward open-source collaboration and fair trade rules.
- 80,000 Hours2h 51m
Having a successful career with depression, anxiety and imposter syndrome | 80,000 Hours
Keiran Harris, Rob Wiblin, Howie Lempel
Strategy Advisor Howie Lempel of 80,000 Hours shares a candid interview detailing his decades-long struggle with severe depression, anxiety, imposter syndrome, and ADHD. The conversation critiques the overreliance on "grit" and lifestyle fixes, arguing instead for prioritizing professional medication and therapy while outlining specific coping strategies like emergency communication plans and financial runways. Following a period of unemployment and recovery supported by the effective altruism community, Lempel was promoted to Chief of Staff and now advocates for systemic changes to make mental health care more accessible and destigmatized.