Immediately before launching this new site I was posting regular assorted links to Facebook. I’ve collected those links below.
May 31st links
- Scott Aaronson’s reply to Giulio Tononi’s reply to Scott Aaronson on the integrated information theory of consciousness.
- Equity market guru accuracy ratings, based on 6500+ forecasts from 68 gurus.
- Machines vs. lawyers.
- Robin Goldstein pranks Wine Spectator with a fake restaurant.
- Steven Pinker’s next book is available for pre-order.
May 29th links
- FLI’s inaugural talks and panel: “The future of technology, benefits and risks” (video).
- How much do Y Combinator founders earn?
- Megaprojects (> $1B budget) almost never finish on time, on budget, and with the promised benefits.
- Google has a working prototype of a fully self-driving car with no steering wheel (video).
- SIGGRAPH 2014 technical papers preview (video).
May 27th links
- Sandberg, “Smarter policymaking through improved collective cognition.”
- Johann Schumann on high-assurance systems.
- “Communicating values to autonomous agents.”
- Robotic arm catches everything tossed in its direction (video).
May 24th links
- Randall Munroe used some creative methods to figure out that Google has ~15 exabytes of storage capacity, probably more than anyone else, including the NSA.
- Szilard’s chain reaction patent: visionary or crankish?
- AI: the next 25 years. Comments from a couple dozen AI scientists.
- GiveWell: Narrowing down U.S. policy areas ripe for philanthropic investment.
- In case you missed it earlier: The Yudkowsky Ambition Scale.
May 23rd links
- Aaronson, “Why I Am Not an Integrated Information Theorist.”
- Sandberg, “Can governments counter Moore’s law of mad science?“
- Brain stimulation makes man a Johnny Cash fan.
- Funny+depressing: What if air travel worked like health care?
May 21st links
- TIL the Bulletin of the Atomic Scientists also publishes papers on “stopping killer robots.”
- The evolution of CIRCA, an AI architecture designed to plan and act in real-time with some rigorous performance guarantees via e.g. formal verification and automatic abstraction methods.
- Arthur C. Clarke reflects (in 1962) on the challenges of technological prediction. Includes an interesting list of expected and unexpected inventions.
- 100+ pieces of recommended longform journalism from 2013.
- Nielsen, “Reinventing explanation.”
May 18th links
- Very long list of computer-related risks to public wellbeing, compiled over several decades by Peter Neumann of SRI.
- Dewey, “What could we do about intelligence explosion?” (A different view than e.g. Yudkowsky.)
- Garett Jones and Robin Hanson on long-run consequences of economic growth.
- Soskis, “The importance of criticizing philanthropy.
- Baidu has hired deep learning superstar Andrew Ng as head of research.
May 16th links
- ONR awards $7.5M for machine ethics research at Tufts, RPI, Brown, Yale, and Georgetown.
- Inside the world of Computer Go.
- New paper suggests that variation in a single gene can be worth 6 IQ points. But I bet the result won’t hold up.
- Last summer, not long after Zell Kravinsky had given almost his entire $45M real-estate fortune to charity… [he] donated a kidney [to a stranger].”
- 80,000 Hours analyzes, in rather-excruciating detail, 107 cases of significant career plan change they’ve plausibly caused.
May 14th links
- Scott Aaronson: The 10 most annoying questions in quantum computing.
- Christof Koch and Stuart Russell on machine superintelligence.
- Christiano, “Machine intelligence and capital accumulation.”
- Achilles and the tortoise discuss trend extrapolation: a dialogue by Anders Sandberg.
- GiveWell investigates philanthropic opportunities related to autonomous vehicles.
May 12th links
- Exponential and non-exponential trends in information technology.
- A profile of MIT’s new $25M research center aimed at AGI.
- 9000+ “spurious” correlations, w/ graphs. E.g. fishing-boat drownings in each year correlates (.95) with the Kentucky marriage rate in each year. & http://is.gd/ZeQb12
- Four of the 20th century’s greatest minds — von Neumann, Teller, Wigner, and Haar — came from a single high school while it was run by Laszlo Ratz. Was this a fluke, or the result of a copyable method?
- Quick thoughts on what computer science researchers can do to improve AGI outcomes in expectation.
May 10th links
- “Problems of self-reference in self-improving space-time embedded intelligence.”
- The annual “Humies” awards are given to evolutionary algorithms which show roughly human-level performance on various narrow tasks.
- Michael Fisher on verifying autonomous systems.
- Due to popular demand, ‘Smarter Than Us’ is now available in paperback.
- College majors, 1970-2011. Art & English majors remain dumbfoundingly popular.
May 7th links
- “Every country will have armed drones within ten years” and “The looming robotics gap.” http://is.gd/WbuDUo
- Shulman, “What do [fake scientific fields] tell us about fraud risk?”
- Hanson, “Fixing academia via prediction markets.”
- “Yes, IQ Really Matters” (overview).
May 5th links
- An interactive world map of synthetic biology labs.
- A retrospective analysis of 1000+ technological predictions (data sets included).
- “Formalism of requirements for safety-critical software: Where does the benefit come from?”
- Google’s self-driving car software has been improved (video included).
- $/QALY is widely used by the NHS to determine whether new treatments are a good value, but the Obamacare-created “Patient-Centered Outcomes Research Institute” is specifically forbidden from funding studies which include measurement of a treatment’s cost-effectiveness.
- Future generations research (a bibliography).
May 3rd links
- Philosopher wins $7.5M DoD grant to apply HoTT to program verification. Apparently this is what counts as “humanities” research at CMU.
- Utopian science.
- Eckersley & Sandberg, “Is brain emulation dangerous?”
May 1st links
- “The errors, insights and lessons of famous AI predictions – and what they mean for the future.”
- Peter Buffett and Will MacAskill debate effective altruism.
- A mesoscale connection strength survey of the mouse brain.
- The US Air Force’s ‘Blue Horizons’ program studies political and technological scenarios 20 years into the future, and has produced dozens of publications since 2007. The ‘Blue Horizons III’ report is very Kurzweilian, and was apparently taken down (but I link it below).
- “Careful engineering and cost analysis shows we can build pleasant, self-sufficient dwelling places in space within the next two decades…” – a 1974 paper by Gerard O’Neill, who inspired the L5 Society, which advocated “space steads” decades before the SeaSteading Institute began showing how hard it is simply to build *sea* steads in international waters *on Earth.*
April 29th links
- A new book I’m reading about how to rebuild technological civilization after a global catastrophe.
- A clarification: 5 kinds of nanotechnology.
- A very nice essay on the challenges of doing and interpreting science.
- Aaronson, “The Quest for Randomness.”
- Freeman Dyson: “I’m very proud of not having a Ph.D. I think the Ph.D. system is an abomination.”
- The untold story of Larry Page’s incredible comeback.
April 27th links
- “Psychological strategies for winning a geopolitical forecasting tournament” recommends probability training, model combination, and judgment aggregation.
- Does the brain calculate value? (a review)
- Extensive timeline of science fiction inventions that are now (more or less) reality, starting from 1634:
- Dewey, “Crucial Phenomena.”
- Omohundro, “Autonomous technology and the greater human good.”
- Tomasik, “International Cooperation vs. AI Arms Race.”
April 25th links
- Bostrom’s excellent ‘Superintelligence’ book has cover art and is available for pre-order on Amazon.
- Reminder: Habitable exoplanets are bad news for humanity.
- Keep in mind these 10 differences between brains and (today’s) computers whenever you hear the phrase “computational power of the human brain”:
- I interview Suzana Herculano-Houzel on cognitive ability and brain size.
- GiveWell interviews Phil Tetlock about forecasting tournaments.
- Yampolskiy, “Utility function security in artificially intelligent agents.”
April 23rd links
- Unsurprisingly, one can (in some cases) steal private data from a user’s brain via a side-channel attack on their commercially available brain-computer interface.
- New honors thesis compares CDT, EDT, TDT, and UDT (4 competing decision procedures).
- GiveWell interviews Stuart Russell about long-term AI safety challenges.
- GiveWell interviews Robin Hanson about conditional prediction markets, which could be more useful than the usual unconditional markets for predicting messy, distant outcomes relating to e.g. emerging technologies.
- Nick Beckstead interviews David Christian & Joel Mokyr about whether the industrial revolution was inevitable, i.e. how likely human civilization is to recover from a GCR.
- Nick Beckstead interviews Tyler Cowen about long-run issues in effective altruism.