The Alignment Newsletter is a weekly publication with recent content relevant to AI alignment. This podcast is an audio version, recorded by Robert Miles (http://robertskmiles.com) More information about the newsletter at: https://rohinshah.com/alignment-newsletter/
…
continue reading
1
Alignment Newsletter #173: Recent language model results from DeepMind
16:43
16:43
Mais Tarde
Mais Tarde
Listas
Like
Curtido
16:43
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg HIGHLIGHTS Scaling Language Models: Methods, Analysis & Insights from Training Gopher (Jack W. Rae et al) (summarized by Rohin): This pap…
…
continue reading
1
Alignment Newsletter #172: Sorry for the long hiatus!
5:52
5:52
Mais Tarde
Mais Tarde
Listas
Like
Curtido
5:52
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg Sorry for the long hiatus! I was really busy over the past few months and just didn't find time to write this newsletter. (Realistically,…
…
continue reading
1
Alignment Newsletter #171: Disagreements between alignment "optimists" and "pessimists"
14:21
14:21
Mais Tarde
Mais Tarde
Listas
Like
Curtido
14:21
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg HIGHLIGHTS Alignment difficulty (Richard Ngo and Eliezer Yudkowsky) (summarized by Rohin): Eliezer is known for being pessimistic about o…
…
continue reading
1
Alignment Newsletter #170: Analyzing the argument for risk from power-seeking AI
13:01
13:01
Mais Tarde
Mais Tarde
Listas
Like
Curtido
13:01
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg HIGHLIGHTS Draft report on existential risk from power-seeking AI (Joe Carlsmith) (summarized by Rohin): This report investigates the cla…
…
continue reading
1
Alignment Newsletter #169: Collaborating with humans without human data
15:08
15:08
Mais Tarde
Mais Tarde
Listas
Like
Curtido
15:08
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg HIGHLIGHTS Collaborating with Humans without Human Data (DJ Strouse et al) (summarized by Rohin): We’ve previously seen that if you want …
…
continue reading
1
Alignment Newsletter #168: Four technical topics for which Open Phil is soliciting grant proposals
16:21
16:21
Mais Tarde
Mais Tarde
Listas
Like
Curtido
16:21
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg HIGHLIGHTS Request for proposals for projects in AI alignment that work with deep learning systems (Nick Beckstead and Asya Bergal) (summ…
…
continue reading
1
Alignment Newsletter #167: Concrete ML safety problems and their relevance to x-risk
17:10
17:10
Mais Tarde
Mais Tarde
Listas
Like
Curtido
17:10
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg HIGHLIGHTS Unsolved Problems in ML Safety (Dan Hendrycks, Nicholas Carlini, John Schulman, and Jacob Steinhardt) (summarized by Dan Hendr…
…
continue reading
1
Alignment Newsletter #166: Is it crazy to claim we're in the most important century?
15:42
15:42
Mais Tarde
Mais Tarde
Listas
Like
Curtido
15:42
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg HIGHLIGHTS The "most important century" series (Holden Karnofsky) (summarized by Rohin): In some sense, it is really weird for us to clai…
…
continue reading
1
Alignment Newsletter #165: When large models are more likely to lie
16:05
16:05
Mais Tarde
Mais Tarde
Listas
Like
Curtido
16:05
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg HIGHLIGHTS TruthfulQA: Measuring How Models Mimic Human Falsehoods (Stephanie Lin et al) (summarized by Rohin): Given that large language…
…
continue reading
1
Alignment Newsletter #164: How well can language models write code?
18:40
18:40
Mais Tarde
Mais Tarde
Listas
Like
Curtido
18:40
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg HIGHLIGHTS Program Synthesis with Large Language Models (Jacob Austin, Augustus Odena et al) (summarized by Rohin): Can we use large lang…
…
continue reading
1
Alignment Newsletter #163: Using finite factored sets for causal and temporal inference
19:27
19:27
Mais Tarde
Mais Tarde
Listas
Like
Curtido
19:27
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg This newsletter is a combined summary + opinion for the Finite Factored Sets sequence by Scott Garrabrant. I (Rohin) have taken a lot mor…
…
continue reading
1
Alignment Newsletter #162: Foundation models: a paradigm shift within AI
15:46
15:46
Mais Tarde
Mais Tarde
Listas
Like
Curtido
15:46
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #161: Creating generalizable reward functions for multiple tasks by learning a model of functional similarity
17:38
17:38
Mais Tarde
Mais Tarde
Listas
Like
Curtido
17:38
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #160: Building AIs that learn and think like people
17:26
17:26
Mais Tarde
Mais Tarde
Listas
Like
Curtido
17:26
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #159: Building agents that know how to experiment, by training on procedurally generated games
27:00
27:00
Mais Tarde
Mais Tarde
Listas
Like
Curtido
27:00
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #158: Should we be optimistic about generalization?
15:39
15:39
Mais Tarde
Mais Tarde
Listas
Like
Curtido
15:39
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #157: Measuring misalignment in the technology underlying Copilot
14:17
14:17
Mais Tarde
Mais Tarde
Listas
Like
Curtido
14:17
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #156: The scaling hypothesis: a plan for building AGI
14:17
14:17
Mais Tarde
Mais Tarde
Listas
Like
Curtido
14:17
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #155: A Minecraft benchmark for algorithms that learn without reward functions
12:43
12:43
Mais Tarde
Mais Tarde
Listas
Like
Curtido
12:43
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #154: What economic growth theory has to say about transformative AI
16:05
16:05
Mais Tarde
Mais Tarde
Listas
Like
Curtido
16:05
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #153: Experiments that demonstrate failures of objective robustness
15:37
15:37
Mais Tarde
Mais Tarde
Listas
Like
Curtido
15:37
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #152: How we’ve overestimated few-shot learning capabilities
14:59
14:59
Mais Tarde
Mais Tarde
Listas
Like
Curtido
14:59
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #151: How sparsity in the final layer makes a neural net debuggable
11:13
11:13
Mais Tarde
Mais Tarde
Listas
Like
Curtido
11:13
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #150: The subtypes of Cooperative AI research
12:34
12:34
Mais Tarde
Mais Tarde
Listas
Like
Curtido
12:34
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #149: The newsletter's editorial policy
14:14
14:14
Mais Tarde
Mais Tarde
Listas
Like
Curtido
14:14
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #148: Analyzing generalization across more axes than just accuracy or loss
21:57
21:57
Mais Tarde
Mais Tarde
Listas
Like
Curtido
21:57
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #147: An overview of the interpretability landscape
13:28
13:28
Mais Tarde
Mais Tarde
Listas
Like
Curtido
13:28
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #146: Plausible stories of how we might fail to avert an existential catastrophe
15:10
15:10
Mais Tarde
Mais Tarde
Listas
Like
Curtido
15:10
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #145: Our three year anniversary!
13:39
13:39
Mais Tarde
Mais Tarde
Listas
Like
Curtido
13:39
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #144: How language models can also be finetuned for non-language tasks
12:45
12:45
Mais Tarde
Mais Tarde
Listas
Like
Curtido
12:45
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #143: How to make embedded agents that reason probabilistically about their environments
14:45
14:45
Mais Tarde
Mais Tarde
Listas
Like
Curtido
14:45
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #142: The quest to understand a network well enough to reimplement it by hand
15:55
15:55
Mais Tarde
Mais Tarde
Listas
Like
Curtido
15:55
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #141: The case for practicing alignment work on GPT-3 and other large models
16:00
16:00
Mais Tarde
Mais Tarde
Listas
Like
Curtido
16:00
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #140: Theoretical models that predict scaling laws
19:21
19:21
Mais Tarde
Mais Tarde
Listas
Like
Curtido
19:21
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #139: How the simplicity of reality explains the success of neural nets
22:14
22:14
Mais Tarde
Mais Tarde
Listas
Like
Curtido
22:14
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #138: Why AI governance should find problems rather than just solving them
16:41
16:41
Mais Tarde
Mais Tarde
Listas
Like
Curtido
16:41
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #137: Quantifying the benefits of pretraining on downstream task performance
15:47
15:47
Mais Tarde
Mais Tarde
Listas
Like
Curtido
15:47
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #136: How well will GPT-N perform on downstream tasks?
17:20
17:20
Mais Tarde
Mais Tarde
Listas
Like
Curtido
17:20
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #135: Five properties of goal-directed systems
15:48
15:48
Mais Tarde
Mais Tarde
Listas
Like
Curtido
15:48
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #134: Underspecification as a cause of fragility to distribution shift
13:17
13:17
Mais Tarde
Mais Tarde
Listas
Like
Curtido
13:17
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #133: Building machines that can cooperate (with humans, institutions, or other machines)
17:12
17:12
Mais Tarde
Mais Tarde
Listas
Like
Curtido
17:12
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #132: Complex and subtly incorrect arguments as an obstacle to debate
17:44
17:44
Mais Tarde
Mais Tarde
Listas
Like
Curtido
17:44
Recorded by Robert Miles: http://robertskmiles.com More information about the newsletter here: https://rohinshah.com/alignment-newsletter/ YouTube Channel: https://www.youtube.com/channel/UCfGGFXwKpr-TJ5HfxEFaFCg
…
continue reading
1
Alignment Newsletter #131: Formalizing the argument of ignored attributes in a utility function
17:06
17:06
Mais Tarde
Mais Tarde
Listas
Like
Curtido
17:06
Recorded by Robert Miles More information about the newsletter here
…
continue reading
1
Alignment Newsletter #130: A new AI x-risk podcast, and reviews of the field
12:08
12:08
Mais Tarde
Mais Tarde
Listas
Like
Curtido
12:08
Recorded by Robert Miles More information about the newsletter here
…
continue reading
1
Alignment Newsletter #129: Explaining double descent by measuring bias and variance
13:11
13:11
Mais Tarde
Mais Tarde
Listas
Like
Curtido
13:11
Recorded by Robert Miles More information about the newsletter here
…
continue reading
1
Alignment Newsletter #128: Prioritizing research on AI existential safety based on its application to governance demands
18:30
18:30
Mais Tarde
Mais Tarde
Listas
Like
Curtido
18:30
Recorded by Robert Miles More information about the newsletter here
…
continue reading
1
Alignment Newsletter #127: Rethinking agency: Cartesian frames as a formalization of ways to carve up the world into an agent and its environment
22:56
22:56
Mais Tarde
Mais Tarde
Listas
Like
Curtido
22:56
Recorded by Robert Miles More information about the newsletter here
…
continue reading
1
Alignment Newsletter #126: Avoiding wireheading by decoupling action feedback from action effects
16:59
16:59
Mais Tarde
Mais Tarde
Listas
Like
Curtido
16:59
Recorded by Robert Miles More information about the newsletter here
…
continue reading
1
Alignment Newsletter #125: Neural network scaling laws across multiple modalities
14:41
14:41
Mais Tarde
Mais Tarde
Listas
Like
Curtido
14:41
Recorded by Robert Miles More information about the newsletter here
…
continue reading
1
Alignment Newsletter #124: Provably safe exploration through shielding
18:14
18:14
Mais Tarde
Mais Tarde
Listas
Like
Curtido
18:14
Recorded by Robert Miles More information about the newsletter here
…
continue reading