AI That Doesn't Try Too Hard - Maximizers and Satisficers

Robert Miles AI Safety · Intermediate ·🧠 Large Language Models ·6y ago
Powerful AI systems can be dangerous in part because they pursue their goals as strongly as they can. Perhaps it would be safer to have systems that don't aim for perfection, and stop at 'good enough'. How could we build something like that? Generating Fake YouTube comments with GPT-2: https://youtu.be/M6EXmoP5jX8 Computerphile Videos: Unicorn AI: https://youtu.be/89A4jGvaaKk More GPT-2, the 'writer' of Unicorn AI: https://youtu.be/p-6F4rhRYLQ AI Language Models & Transformers: https://youtu.be/rURRYI66E54 GPT-2: Why Didn't They Release It?: https://youtu.be/AJxLtdur5fc The Deadly Truth of General AI?: https://youtu.be/tcdVC4e6EV4 With thanks to my excellent Patreon supporters: https://www.patreon.com/robertskmiles Scott Worley Jordan Medina Simon Strandgaard JJ Hepboin Lupuleasa Ionuț Pedro A Ortega Said Polat Chris Canal Nicholas Kees Dupuis Jake Ehrlich Mark Hechim Kellen lask Francisco Tolmasky Michael Andregg Alexandru Dobre David Reid Robert Daniel Pickard Peter Rolf Chad Jones Truthdoc James Richárd Nagyfi Jason Hise Phil Moyer Shevis Johnson Alec Johnson Clemens Arbesser Ludwig Schubert Bryce Daifuku Allen Faure Eric James Jonatan R Ingvi Gautsson Michael Greve Julius Brash Tom O'Connor Erik de Bruijn Robin Green Laura Olds Jon Halliday Paul Hobbs Jeroen De Dauw Tim Neilson Eric Scammell Igor Keller Ben Glanton Robert Sokolowski anul kumar sinha Jérôme Frossard Sean Gibat Cooper Lawton Tyler Herrmann Tomas Sayder Ian Munro Jérôme Beaulieu Taras Bobrovytsky Anne Buit Tom Murphy Vaskó Richárd Sebastian Birjoveanu Gladamas Sylvain Chevalier DGJono Dmitri Afanasjev Brian Sandberg Marcel Ward Andrew Weir Ben Archer Scott McCarthy Kabs Miłosz Wierzbicki Tendayi Mawushe Jannik Olbrich Anne Kohlbrenner Jussi Männistö Mr Fantastic Wr4thon Martin Ottosen Archy de Berker Marc Pauly Joshua Pratt Andy Kobre Brian Gillespie Martin Wind Peggy Youell Poker Chen Kees Darko Sperac Truls Paul Moffat Anders Öhrt Marco Tiraboschi Michael Kuhinica Fraser Cain Robin Scharf Or
Watch on YouTube ↗ (saves to browser)
Sign in to unlock AI tutor explanation · ⚡30

Playlist

Uploads from Robert Miles AI Safety · Robert Miles AI Safety · 25 of 47

1 Predicting AI: RIP Prof. Hubert Dreyfus
Predicting AI: RIP Prof. Hubert Dreyfus
Robert Miles AI Safety
2 Respectability
Respectability
Robert Miles AI Safety
3 Are AI Risks like Nuclear Risks?
Are AI Risks like Nuclear Risks?
Robert Miles AI Safety
4 Avoiding Negative Side Effects: Concrete Problems in AI Safety part 1
Avoiding Negative Side Effects: Concrete Problems in AI Safety part 1
Robert Miles AI Safety
5 Avoiding Positive Side Effects: Concrete Problems in AI Safety part 1.5
Avoiding Positive Side Effects: Concrete Problems in AI Safety part 1.5
Robert Miles AI Safety
6 Empowerment: Concrete Problems in AI Safety part 2
Empowerment: Concrete Problems in AI Safety part 2
Robert Miles AI Safety
7 Why Not Just: Raise AI Like Kids?
Why Not Just: Raise AI Like Kids?
Robert Miles AI Safety
8 Reward Hacking: Concrete Problems in AI Safety Part 3
Reward Hacking: Concrete Problems in AI Safety Part 3
Robert Miles AI Safety
9 The other "Killer Robot Arms Race" Elon Musk should worry about
The other "Killer Robot Arms Race" Elon Musk should worry about
Robert Miles AI Safety
10 Reward Hacking Reloaded: Concrete Problems in AI Safety Part 3.5
Reward Hacking Reloaded: Concrete Problems in AI Safety Part 3.5
Robert Miles AI Safety
11 What Can We Do About Reward Hacking?: Concrete Problems in AI Safety Part 4
What Can We Do About Reward Hacking?: Concrete Problems in AI Safety Part 4
Robert Miles AI Safety
12 What can AGI do? I/O and Speed
What can AGI do? I/O and Speed
Robert Miles AI Safety
13 AI learns to Create  ̵K̵Z̵F̵ ̵V̵i̵d̵e̵o̵s̵ Cat Pictures: Papers in Two Minutes #1
AI learns to Create ̵K̵Z̵F̵ ̵V̵i̵d̵e̵o̵s̵ Cat Pictures: Papers in Two Minutes #1
Robert Miles AI Safety
14 AI Safety at EAGlobal2017 Conference
AI Safety at EAGlobal2017 Conference
Robert Miles AI Safety
15 Scalable Supervision: Concrete Problems in AI Safety Part 5
Scalable Supervision: Concrete Problems in AI Safety Part 5
Robert Miles AI Safety
16 Superintelligence Mod for Civilization V
Superintelligence Mod for Civilization V
Robert Miles AI Safety
17 Why Would AI Want to do Bad Things? Instrumental Convergence
Why Would AI Want to do Bad Things? Instrumental Convergence
Robert Miles AI Safety
18 Experts' Predictions about the Future of AI
Experts' Predictions about the Future of AI
Robert Miles AI Safety
19 AI Safety Gridworlds
AI Safety Gridworlds
Robert Miles AI Safety
20 Friend or Foe? AI Safety Gridworlds extra bit
Friend or Foe? AI Safety Gridworlds extra bit
Robert Miles AI Safety
21 Safe Exploration: Concrete Problems in AI Safety Part 6
Safe Exploration: Concrete Problems in AI Safety Part 6
Robert Miles AI Safety
22 Why Not Just: Think of AGI Like a Corporation?
Why Not Just: Think of AGI Like a Corporation?
Robert Miles AI Safety
23 How to Keep Improving When You're Better Than Any Teacher - Iterated Distillation and Amplification
How to Keep Improving When You're Better Than Any Teacher - Iterated Distillation and Amplification
Robert Miles AI Safety
24 Is AI Safety a Pascal's Mugging?
Is AI Safety a Pascal's Mugging?
Robert Miles AI Safety
AI That Doesn't Try Too Hard - Maximizers and Satisficers
AI That Doesn't Try Too Hard - Maximizers and Satisficers
Robert Miles AI Safety
26 Training AI Without Writing A Reward Function, with Reward Modelling
Training AI Without Writing A Reward Function, with Reward Modelling
Robert Miles AI Safety
27 9 Examples of Specification Gaming
9 Examples of Specification Gaming
Robert Miles AI Safety
28 10 Reasons to Ignore AI Safety
10 Reasons to Ignore AI Safety
Robert Miles AI Safety
29 Sharing the Benefits of AI: The Windfall Clause
Sharing the Benefits of AI: The Windfall Clause
Robert Miles AI Safety
30 Quantilizers: AI That Doesn't Try Too Hard
Quantilizers: AI That Doesn't Try Too Hard
Robert Miles AI Safety
31 The OTHER AI Alignment Problem: Mesa-Optimizers and Inner Alignment
The OTHER AI Alignment Problem: Mesa-Optimizers and Inner Alignment
Robert Miles AI Safety
32 Deceptive Misaligned Mesa-Optimisers? It's More Likely Than You Think...
Deceptive Misaligned Mesa-Optimisers? It's More Likely Than You Think...
Robert Miles AI Safety
33 Intro to AI Safety, Remastered
Intro to AI Safety, Remastered
Robert Miles AI Safety
34 We Were Right! Real Inner Misalignment
We Were Right! Real Inner Misalignment
Robert Miles AI Safety
35 Apply to AI Safety Camp! #shorts
Apply to AI Safety Camp! #shorts
Robert Miles AI Safety
36 Win $50k for Solving a Single AI Problem? #Shorts
Win $50k for Solving a Single AI Problem? #Shorts
Robert Miles AI Safety
37 Free ML Bootcamp for Alignment #shorts
Free ML Bootcamp for Alignment #shorts
Robert Miles AI Safety
38 Apply Now for a Paid Residency on Interpretability #short
Apply Now for a Paid Residency on Interpretability #short
Robert Miles AI Safety
39 Why Does AI Lie, and What Can We Do About It?
Why Does AI Lie, and What Can We Do About It?
Robert Miles AI Safety
40 Apply to Study AI Safety Now! #shorts
Apply to Study AI Safety Now! #shorts
Robert Miles AI Safety
41 AI Ruined My Year
AI Ruined My Year
Robert Miles AI Safety
42 Learn AI Safety at MATS #shorts
Learn AI Safety at MATS #shorts
Robert Miles AI Safety
43 Using Dangerous AI, But Safely?
Using Dangerous AI, But Safely?
Robert Miles AI Safety
44 AI Safety Career Advice! (And So Can You!)
AI Safety Career Advice! (And So Can You!)
Robert Miles AI Safety
45 Robot Dog! Unitree Go2 review #shorts #robot #dog
Robot Dog! Unitree Go2 review #shorts #robot #dog
Robert Miles AI Safety
46 Tech is Good, AI Will Be Different
Tech is Good, AI Will Be Different
Robert Miles AI Safety
47 Apply for the Affine Superintelligence Alignment Seminar #shorts
Apply for the Affine Superintelligence Alignment Seminar #shorts
Robert Miles AI Safety

Related AI Lessons

Moonshot AI and the Rise of Beijing’s Open-Source Frontier: What a $20B Valuation Signals for…
Moonshot AI's $20B valuation signals a shift in the AI landscape, with Beijing emerging as a hub for open-source innovation
Medium · LLM
How AI Models are making UI obsolete
Learn how AI models are revolutionizing UI with agents that automate manual clicking processes, making traditional UI obsolete
Medium · LLM
“LLMs Do Not Remember Anything”: They only process the context we give them.
LLMs don't have memory, they process context given to them, and bigger models won't solve context accumulation problems
Dev.to AI
Why My Coding Assistant Started Replying in Korean When I Typed Chinese
Explore how coding assistants can unexpectedly switch languages due to embedding space overlaps, and learn to analyze such phenomena using vector databases and language models.
Towards Data Science
Up next
5 Levels of AI Agents - From Simple LLM Calls to Multi-Agent Systems
Dave Ebbelaar (LLM Eng)
Watch →