nhaliday + research-program   52

Information Processing: Mathematical Theory of Deep Neural Networks (Princeton workshop)
"Recently, long-past-due theoretical results have begun to emerge. These results, and those that will follow in their wake, will begin to shed light on the properties of large, adaptive, distributed learning architectures, and stand to revolutionize how computer science and neuroscience understand these systems."
hsu  scitariat  commentary  links  research  research-program  workshop  events  princeton  sanjeev-arora  deep-learning  machine-learning  ai  generalization  explanans  off-convex  nibble  frontier  speedometer  state-of-art  big-surf  announcement 
january 2018 by nhaliday
Darwinian medicine - Randolph Nesse
The Dawn of Darwinian Medicine: https://sci-hub.tw/https://www.jstor.org/stable/2830330
TABLE 1 Examples of the use of the theory of natural selection to predict the existence of phenomena otherwise unsuspected
TABLE 2 A classification of phenomena associated with infectious disease
research-program  homepage  links  list  study  article  bio  medicine  disease  parasites-microbiome  epidemiology  evolution  darwinian  books  west-hunter  scitariat  🌞  red-queen  ideas  deep-materialism  biodet  EGT  heterodox  essay  equilibrium  incentives  survey  track-record  priors-posteriors  data  paying-rent  being-right  immune  multi  pdf  piracy  EEA  lens  nibble  🔬  maxim-gun 
november 2017 by nhaliday
Superintelligence Risk Project Update II

For example, I asked him what he thought of the idea that to we could get AGI with current techniques, primarily deep neural nets and reinforcement learning, without learning anything new about how intelligence works or how to implement it ("Prosaic AGI" [1]). He didn't think this was possible, and believes there are deep conceptual issues we still need to get a handle on. He's also less impressed with deep learning than he was before he started working in it: in his experience it's a much more brittle technology than he had been expecting. Specifically, when trying to replicate results, he's often found that they depend on a bunch of parameters being in just the right range, and without that the systems don't perform nearly as well.

The bottom line, to him, was that since we are still many breakthroughs away from getting to AGI, we can't productively work on reducing superintelligence risk now.

He told me that he worries that the AI risk community is not solving real problems: they're making deductions and inferences that are self-consistent but not being tested or verified in the world. Since we can't tell if that's progress, it probably isn't. I asked if he was referring to MIRI's work here, and he said their work was an example of the kind of approach he's skeptical about, though he wasn't trying to single them out. [2]

Earlier this week I had a conversation with an AI researcher [1] at one of the main industry labs as part of my project of assessing superintelligence risk. Here's what I got from them:

They see progress in ML as almost entirely constrained by hardware and data, to the point that if today's hardware and data had existed in the mid 1950s researchers would have gotten to approximately our current state within ten to twenty years. They gave the example of backprop: we saw how to train multi-layer neural nets decades before we had the computing power to actually train these nets to do useful things.

Similarly, people talk about AlphaGo as a big jump, where Go went from being "ten years away" to "done" within a couple years, but they said it wasn't like that. If Go work had stayed in academia, with academia-level budgets and resources, it probably would have taken nearly that long. What changed was a company seeing promising results, realizing what could be done, and putting way more engineers and hardware on the project than anyone had previously done. AlphaGo couldn't have happened earlier because the hardware wasn't there yet, and was only able to be brought forward by massive application of resources.

Summary: I'm not convinced that AI risk should be highly prioritized, but I'm also not convinced that it shouldn't. Highly qualified researchers in a position to have a good sense the field have massively different views on core questions like how capable ML systems are now, how capable they will be soon, and how we can influence their development. I do think these questions are possible to get a better handle on, but I think this would require much deeper ML knowledge than I have.
ratty  core-rats  ai  risk  ai-control  prediction  expert  machine-learning  deep-learning  speedometer  links  research  research-program  frontier  multi  interview  deepgoog  games  hardware  performance  roots  impetus  chart  big-picture  state-of-art  reinforcement  futurism  🤖  🖥  expert-experience  singularity  miri-cfar  empirical  evidence-based  speculation  volo-avolo  clever-rats  acmtariat  robust  ideas  crux  atoms  detail-architecture  software  gradient-descent 
july 2017 by nhaliday
natural language processing blog: Whence your reward function?
I think the most substantial issue is the fact that game playing is a simulated environment and the reward function is generally crafted to make humans find the games fun, which usually means frequent small rewards that point you in the right direction. This is exactly where RL works well, and something that I'm not sure is a reasonable assumption in the real world.
acmtariat  critique  reinforcement  deep-learning  machine-learning  research  research-program  org:bleg  nibble  wire-guided  cost-benefit 
december 2016 by nhaliday
Yann LeCun's answer to What are some recent and potentially upcoming breakthroughs in deep learning? - Quora
There are many interesting recent development in deep learning, probably too many for me to describe them all here. But there are a few ideas that caught my attention enough for me to get personally involved in research projects.

The most important one, in my opinion, is adversarial training (also called GAN for Generative Adversarial Networks). This is an idea that was originally proposed by Ian Goodfellow when he was a student with Yoshua Bengio at the University of Montreal (he since moved to Google Brain and recently to OpenAI).

This, and the variations that are now being proposed is the most interesting idea in the last 10 years in ML, in my opinion.
deep-learning  expert  reflection  synthesis  ama  q-n-a  research-program  research  adversarial  qra  acmtariat  big-picture  nibble  ideas  expert-experience 
july 2016 by nhaliday
Bottoming Out – arg min blog
Now, I’ve been hammering the point in my previous posts that saddle points are not what makes non-convex optimization difficult. Here, when specializing to deep learning, even local minima are not getting in my way. Deep neural nets are just very easy to minimize.
machine-learning  deep-learning  optimization  rhetoric  speculation  research  hmm  research-program  acmtariat  generalization  metabuch  local-global  off-convex  ben-recht  extrema  org:bleg  nibble  sparsity  curvature  ideas  aphorism  convexity-curvature  explanans  volo-avolo  hardness 
june 2016 by nhaliday

bundles : academemeta

related tags

abstraction  academia  accretion  acm  acmtariat  adversarial  ai  ai-control  akrasia  algebraic-complexity  algorithms  alignment  altruism  ama  announcement  anthropology  aphorism  apollonian-dionysian  applications  article  asia  atoms  attention  auto-learning  automation  average-case  backup  bayesian  behavioral-econ  being-right  ben-recht  berkeley  biases  big-picture  big-surf  bio  biodet  bits  blog  bonferroni  books  bostrom  broad-econ  caltech  chart  chicago  china  clarity  clever-rats  cog-psych  commentary  communication-complexity  comparison  complexity  computational-geometry  concept  concrete  conference  convexity-curvature  cool  cooperate-defect  coordination  core-rats  cost-benefit  cracker-econ  critique  crux  cs  cultural-dynamics  culture  curvature  darwinian  data  data-science  decision-making  deep-learning  deep-materialism  deepgoog  defense  definition  descriptive  detail-architecture  differential-privacy  dimensionality  discrimination  discussion  disease  distribution  economics  econotariat  EEA  effective-altruism  EGT  elegance  empirical  epidemiology  equilibrium  essay  ethnocentrism  events  evidence-based  evolution  examples  expert  expert-experience  explanans  explanation  exploratory  explore-exploit  exposition  extrema  fall-2015  finance  foreign-policy  frequentist  frontier  futurism  game-theory  games  gelman  generalization  gradient-descent  group-selection  hard-core  hardness  hardware  hari-seldon  hashing  henrich  heterodox  hi-order-bits  high-variance  hmm  hn  homepage  hsu  human-capital  human-ml  hypothesis-testing  ideas  immune  impetus  incentives  info-dynamics  info-econ  info-foraging  information-theory  institutions  intelligence  interdisciplinary  interests  interpretability  interview  intricacy  iteration-recursion  latent-variables  learning-theory  len:short  lens  liner-notes  links  list  local-global  logic  lower-bounds  machine-learning  markets  markov  math  maxim-gun  medicine  meta:science  metabuch  metameta  methodology  micro  microfoundations  mihai  military  miri-cfar  model-class  models  monte-carlo  motivation  mrtz  msr  multi  neurons  news  nibble  nips  no-go  off-convex  online-learning  open-closed  open-problems  openai  optimization  org:bleg  org:econlib  org:edu  org:inst  org:junk  org:mat  org:med  org:rec  organization  organizing  oss  overflow  oxbridge  papers  parasites-microbiome  paul-romer  paying-rent  pdf  people  performance  perturbation  piracy  polisci  pragmatic  prediction  preprint  presentation  princeton  priors-posteriors  probability  prof  psychology  puzzles  q-n-a  qra  quantum  quantum-info  questions  quixotic  rationality  ratty  reading  realness  red-queen  reduction  reference  reflection  regulation  reinforcement  replication  repo  research  research-program  retention  rhetoric  rigor  risk  robust  roots  safety  sanjeev-arora  sapiens  science  scitariat  sebastien-bubeck  security  seminar  sensitivity  similarity  singularity  social  social-norms  social-science  soft-question  software  sparsity  spectral  speculation  speedometer  spock  state-of-art  stats  stories  strategy  stream  study  sublinear  summary  survey  synthesis  talks  tcs  tech  technology  tensors  threat-modeling  track-record  tradeoffs  trends  tutorial  twitter  uncertainty  unintended-consequences  unit  unsupervised  values  video  virginia-DC  volo-avolo  west-hunter  wiki  wire-guided  workshop  yoga  🌞  🎩  👳  🔬  🖥  🤖 

Copy this bookmark: