Fast index
AI people & papers index
Quick links by what each person or item is famous for, so the AI section stays easier to use as it grows.
Bio-data source card
George Church and DNA Data Storage
Source: YouTube Short on George Church and DNA storage · Core paper: Church, Gao & Kosuri, Science 2012.
Church’s team really did encode digital information into DNA and read it back. The Managing Expectations angle is the correction: DNA storage is real, but claims about bodies holding cosmic history are speculative metaphors, not conclusions from the paper.
Primary paper
Situational Awareness: The Decade Ahead
Author: Leopold Aschenbrenner · Published: June 2024 · Length: 165-page PDF.
The paper argues that frontier AI should be treated as a strategic transformation: compute, energy, model security, chip supply, capital, national competition, and governance all matter. The useful Managing Expectations angle is simple: separate what the source says from what it forecasts.
pages in the primary PDF
public release year
ways to read it: facts, forecasts, policy, open questions
Managing Expectations note
This is not posted as prophecy. It is posted as a serious strategic paper worth reading, testing, and arguing with. Forecasts are not facts; but leaders who ignore frontier AI entirely are managing expectations badly.
Physical AI source check
When Robots Outnumber People at Figure AI
Source: Brett Adcock / Figure AI · Claim: robots now outnumber humans at Figure.
The milestone is worth noticing because humanoid robotics is moving from demo videos into fleet learning. But the expectation to manage is just as important: a robot count is not the same as a worker count.
Capability forecasts
What the paper predicts about model progress, automation, and the decade ahead — presented as forecasts, not settled outcomes.
Compute and energy
Why chips, data centers, electricity, and capital become central to AI strategy.
Security and model weights
Why frontier model security, espionage risk, and infrastructure control matter.
Governance
How regulation, democratic accountability, corporate concentration, and national competition collide.
Godfather of AI profile
Yoshua Bengio
Role: Full Professor of Computer Science at Université de Montréal, Founder and Scientific Advisor of Mila, Co-President and Scientific Director of LawZero, Canada CIFAR AI Chair, and 2018 A.M. Turing Award recipient.
Bengio is one of the three best-known “godfathers of AI” because his work helped make modern deep learning possible. The Managing Expectations angle is his turn from building the field to warning that frontier AI needs stronger safety science, governance, and less-dangerous system designs.
Latest check, July 2, 2026: LawZero released a Bengio-led technical publication, Safety from Honesty in a Disinterested AI Predictor, alongside a Scientist AI blog sequence; Bengio also pointed readers to the UN Independent International Scientific Panel on AI preliminary report. Treat these as serious safety/governance proposals and warnings, not settled prophecy about any exact AI timeline.
Managing Expectations note
Bengio should not be reduced to a one-line “AI doom” quote. He matters because technical credibility, institution-building, and current safety warnings all meet in one person: deep learning, Mila, the Turing Award, International AI Safety Report work, and LawZero’s Scientist AI proposal.
Bengio’s main theories and arguments
Deep learning pioneer
Bengio shared the 2018 A.M. Turing Award with Geoffrey Hinton and Yann LeCun for foundational work behind modern deep learning.
Representation learning
His research emphasizes systems that learn useful internal representations rather than relying only on hand-coded features.
Causality and reasoning
Later Bengio research pushes beyond pattern recognition toward causal understanding, reasoning, and systematic generalization.
Responsible AI
He helped draft the Montréal Declaration for Responsible AI and publicly argues for safety, democratic governance, and risk mitigation.
Scientist AI
Through LawZero, Bengio proposes safer “Scientist AI”: systems designed to understand and explain the world without dangerous agentic drives.
Frontier-risk warning
In the Diary of a CEO interview and public work, he warns that fast capability growth creates serious job, governance, security, and catastrophic-risk questions.
Diary of a CEO interview
Dec. 18, 2025 — “Godfather of AI: We Have 2 Years Before Everything Changes!”
The main Diary of a CEO episode with Steven Bartlett was identified and logged. Metadata shows upload date 2025-12-18, channel The Diary Of A CEO, and duration about 100 minutes. A related clips-channel excerpt, “Godfather of AI WARNS: You Won’t Believe The Truth”, was uploaded 2025-12-19.
For this page, the interview is treated as a public warning and strategic forecast, not a guaranteed timeline. The local transcript has been saved under research/ai/transcripts/ for future deeper extraction.
Bengio public links, socials & professional contact
AI safety profile
Dr. Roman V. Yampolskiy
Role: Associate Professor at the University of Louisville’s Speed School of Engineering and founding/current Director of its Cyber Security Laboratory. His main public research area is AI safety and security.
Yampolskiy is one of the more severe warning voices in AI safety. His core value for Managing Expectations is that he asks the hard question: if an AI becomes more capable than its designers, can we actually explain it, predict it, contain it, align it, or shut it down?
Latest check: his official YouTube / Roman Forum feed is active in 2026, with June uploads on recursive self-improvement and superintelligence conflict. Treat these as current public risk arguments and interview/forum materials — not proof that the most severe timelines are settled.
Managing Expectations note
Do not present Yampolskiy’s view as settled consensus. Present it as a serious, security-heavy AI-risk argument: if advanced systems are unexplainable, unpredictable, or uncontrollable, normal business optimism is not enough.
Yampolskiy’s main theories
AI control problem
His strongest theme is that there may be no universal method to guarantee a smarter-than-human AI remains under human control once it has broad capability, autonomy, and access.
Unexplainable
Advanced models may become too complex for humans to fully interpret. If we cannot understand why a system acts, trust and verification become fragile.
Unpredictable
Highly adaptive systems can behave in ways their creators did not forecast, especially after deployment into open-ended social, economic, military, or internet-connected environments.
Uncontrollable
Yampolskiy is skeptical that boxing, confinement, rules, or shutdown plans can be guaranteed to work against a sufficiently capable system, because humans and infrastructure become attack surfaces.
Safety as security
He treats AI risk partly as a cybersecurity problem: access control, model weights, containment, verification, persuasion, hacking, and side channels matter.
Deployment restraint
His argument pushes toward slowing down or constraining risky deployment until society has stronger evidence that advanced systems can be made safe.
Public links & social media
Futurist / AI prediction profile
Ray Kurzweil
Role: inventor, computer scientist, author, futurist and one of the best-known public advocates of exponential AI progress, AGI by 2029, longevity escape velocity, and the singularity.
Kurzweil belongs in the AI section because his track record is unusually specific for a futurist. His own 2010 review says 127 of 147 predictions for 2009 were correct or essentially correct — the source of the famous 86% claim. The Managing Expectations angle: useful forecaster, not prophet.
Managing Expectations note
Kurzweil’s strongest record is around information-technology curves: computation, communications, AI capability and pattern recognition. Be more cautious with exact dates, human/AI merger timelines, longevity escape velocity, and social consequences.
Kurzweil’s main AI arguments
Law of accelerating returns
Information technologies improve exponentially because each generation of tools helps build the next generation.
AGI by 2029
Kurzweil continues to argue that human-level AI is plausible around 2029, with the Turing-test debate preserved publicly through Long Bets.
Biology becomes information
He expects AI, simulation and biotechnology to compress drug discovery and eventually push toward longevity escape velocity.
Human/AI merger
His singularity thesis expects AI to move from external tools toward seamless cognitive augmentation and eventually biological/computational integration.
New research component
AI Papers Library
Purpose: a living Managing Expectations library for AI leaders, frontier-lab papers, interpretability reports, safety arguments, governance comments and public interviews.
The first lane starts with the Anthropic group shown in the Bloomberg Originals profile — Dario and Daniela Amodei, Jack Clark, Chris Olah, Jared Kaplan, Sam McCandlish and Tom Brown — then expands to OpenAI, DeepMind, Google, Meta, Bengio, Hinton, Yampolskiy and critical AI scholars. Latest update: Anthropic’s June 2026 Economic Index report on Claude usage cadences and AI labor telemetry.
Library rule
Track the paper trail first: title, authors, date, source URL, claim type, evidence level and plain-English significance. Do not confuse a media profile, valuation headline or executive comment with proof.