nhaliday + microsoft   12

What Peter Thiel thinks about AI risk - Less Wrong
TL;DR: he thinks its an issue but also feels AGI is very distant and hence less worried about it than Musk.

I recommend the rest of the lecture as well, it's a good summary of "Zero to One"  and a good QA afterwards.

For context, in case anyone doesn't realize: Thiel has been MIRI's top donor throughout its history.

other stuff:
nice interview question: "thing you know is true that not everyone agrees on?"
"learning from failure overrated"
cleantech a huge market, hard to compete
software makes for easy monopolies (zero marginal costs, network effects, etc.)
for most of history inventors did not benefit much (continuous competition)
ethical behavior is a luxury of monopoly
ratty  lesswrong  commentary  ai  ai-control  risk  futurism  technology  speedometer  audio  presentation  musk  thiel  barons  frontier  miri-cfar  charity  people  track-record  venture  startups  entrepreneurialism  contrarianism  competition  market-power  business  google  truth  management  leadership  socs-and-mops  dark-arts  skunkworks  hard-tech  energy-resources  wire-guided  learning  software  sv  tech  network-structure  scale  marginal  cost-benefit  innovation  industrial-revolution  economics  growth-econ  capitalism  comparison  nationalism-globalism  china  asia  trade  stagnation  things  dimensionality  exploratory  world  developing-world  thinking  definite-planning  optimism  pessimism  intricacy  politics  war  career  planning  supply-demand  labor  science  engineering  dirty-hands  biophysical-econ  migration  human-capital  policy  canada  anglo  winner-take-all  polarization  amazon  business-models  allodium  civilization  the-classics  microsoft  analogy  gibbon  conquest-empire  realness  cynicism-idealism  org:edu  open-closed  ethics  incentives  m 
february 2018 by nhaliday
Performance Trends in AI | Otium
Deep learning has revolutionized the world of artificial intelligence. But how much does it improve performance? How have computers gotten better at different tasks over time, since the rise of deep learning?

In games, what the data seems to show is that exponential growth in data and computation power yields exponential improvements in raw performance. In other words, you get out what you put in. Deep learning matters, but only because it provides a way to turn Moore’s Law into corresponding performance improvements, for a wide class of problems. It’s not even clear it’s a discontinuous advance in performance over non-deep-learning systems.

In image recognition, deep learning clearly is a discontinuous advance over other algorithms. But the returns to scale and the improvements over time seem to be flattening out as we approach or surpass human accuracy.

In speech recognition, deep learning is again a discontinuous advance. We are still far away from human accuracy, and in this regime, accuracy seems to be improving linearly over time.

In machine translation, neural nets seem to have made progress over conventional techniques, but it’s not yet clear if that’s a real phenomenon, or what the trends are.

In natural language processing, trends are positive, but deep learning doesn’t generally seem to do better than trendline.


The learned agent performs much better than the hard-coded agent, but moves more jerkily and “randomly” and doesn’t know the law of reflection. Similarly, the reports of AlphaGo producing “unusual” Go moves are consistent with an agent that can do pattern-recognition over a broader space than humans can, but which doesn’t find the “laws” or “regularities” that humans do.

Perhaps, contrary to the stereotype that contrasts “mechanical” with “outside-the-box” thinking, reinforcement learners can “think outside the box” but can’t find the box?

ratty  core-rats  summary  prediction  trends  analysis  spock  ai  deep-learning  state-of-art  🤖  deepgoog  games  nlp  computer-vision  nibble  reinforcement  model-class  faq  org:bleg  shift  chart  technology  language  audio  accuracy  speaking  foreign-lang  definite-planning  china  asia  microsoft  google  ideas  article  speedometer  whiggish-hegelian  yvain  ssc  smoothness  data  hsu  scitariat  genetics  iq  enhancement  genetic-load  neuro  neuro-nitgrit  brain-scan  time-series  multiplicative  iteration-recursion  additive  multi 
january 2017 by nhaliday

bundles : techie

related tags

2016-election  accuracy  additive  agriculture  ai  ai-control  alignment  allodium  alt-inst  amazon  analogy  analysis  analytical-holistic  anglo  anglosphere  announcement  antidemos  apollonian-dionysian  apple  aristos  arms  art  article  asia  atmosphere  audio  authoritarianism  barons  being-becoming  benevolence  big-peeps  biodet  bioinformatics  biophysical-econ  biotech  books  bots  brain-scan  brands  business  business-models  california  canada  cancer  canon  capital  capitalism  career  cartoons  charity  chart  china  civil-liberty  civilization  cjones-like  class  climate-change  coarse-fine  cold-war  collaboration  commentary  communication  comparison  compensation  competition  complement-substitute  composition-decomposition  computation  computer-vision  concrete  conquest-empire  contrarianism  cooperate-defect  core-rats  corporation  cost-benefit  counter-revolution  courage  course  creative  crime  crooked  cs  current-events  cycles  cynicism-idealism  dark-arts  darwinian  data  database  death  debt  decision-making  deep-learning  deep-materialism  deepgoog  definite-planning  degrees-of-freedom  democracy  detail-architecture  developing-world  dimensionality  dirty-hands  discussion  distribution  drugs  duplication  early-modern  economics  education  efficiency  egalitarianism-hierarchy  einstein  elite  ems  energy-resources  engineering  enhancement  entrepreneurialism  environment  envy  essence-existence  estimate  ethics  europe  evolution  examples  expert  expert-experience  explanans  exploratory  extra-introversion  facebook  failure  faq  fashun  FDA  feudal  fiction  finance  flexibility  focus  foreign-lang  frontier  futurism  gallic  games  genetic-load  genetics  genomics  geoengineering  geography  germanic  giants  gibbon  gnosis-logos  god-man-beast-victim  google  government  growth-econ  hard-tech  harvard  heterodox  hidden-motives  high-variance  higher-ed  history  hn  homo-hetero  honor  hsu  human-capital  human-ml  hypocrisy  ideas  impetus  incentives  individualism-collectivism  industrial-revolution  inequality  info-dynamics  innovation  insight  institutions  intel  interdisciplinary  interests  intricacy  investing  iq  iraq-syria  iteration-recursion  janus  japan  justice  knowledge  labor  language  latin-america  law  leadership  learning  lecture-notes  lens  lesswrong  let-me-see  leviathan  limits  literature  local-global  longevity  love-hate  machine-learning  macro  magnitude  management  marginal  market-power  markets  math  math.CA  maxim-gun  measurement  media  medicine  meta:science  meta:war  metabuch  metameta  microsoft  migration  miri-cfar  mobile  model-class  moments  monetary-fiscal  money  morality  mostly-modern  msr  multi  multiplicative  musk  myth  n-factor  narrative  nationalism-globalism  nature  network-structure  neuro  neuro-nitgrit  new-religion  news  nibble  nietzschean  nitty-gritty  nlp  noble-lie  northeast  nuclear  nutrition  nyc  occident  old-anglo  open-closed  optimism  order-disorder  org:anglo  org:biz  org:bleg  org:edu  org:lite  org:rec  organization  organizing  orient  outcome-risk  outliers  paradox  parallax  parasites-microbiome  patience  peace-violence  people  personality  pessimism  phalanges  pharma  philosophy  physics  planning  plots  polanyi-marx  polarization  policy  polisci  politics  power  power-law  pre-ww2  prediction  presentation  primitivism  princeton  pro-rata  probability  properties  proposal  psych-architecture  quantum  questions  quotes  random  randy-ayndy  ranking  ratty  realness  reason  recommendations  recruiting  redistribution  reflection  regulation  reinforcement  religion  rent-seeking  revolution  rhythm  risk  ritual  robotics  roots  saas  scale  science  scifi-fantasy  scitariat  search  securities  shakespeare  shift  signal-noise  signaling  singularity  sinosphere  skeleton  skunkworks  smoothness  social  social-choice  social-norms  socs-and-mops  software  space  speaking  speed  speedometer  spock  ssc  stagnation  stanford  startups  state-of-art  statesmen  stats  status  stereotypes  stochastic-processes  stock-flow  stories  strategy  structure  stylized-facts  success  summary  supply-demand  sv  synchrony  tactics  tails  tech  technology  telos-atelos  temperance  terrorism  the-classics  the-devil  the-founding  the-great-west-whale  the-trenches  the-watchers  the-west  the-world-is-just-atoms  theory-of-mind  theos  thick-thin  thiel  things  thinking  time  time-preference  time-series  tools  track-record  trade  tradeoffs  transportation  trends  tribalism  trump  trust  truth  twitter  uncertainty  unintended-consequences  urban-rural  us-them  usa  venture  virtu  visual-understanding  visualization  vitality  volo-avolo  war  wealth  welfare-state  west-hunter  whiggish-hegelian  white-paper  winner-take-all  wire-guided  wisdom  within-without  woah  world  world-war  X-not-about-Y  yak-shaving  yvain  zero-positive-sum  zooming  🤖 

Copy this bookmark: