How METR measures Long Tasks and Experienced Open Source Dev Productivity - Joel Becker, METR
AI models are crushing benchmarks. SWE-bench scores are climbing, and METR's measured time horizons are rising rapidly. Yet when we deployed these same models in a field study with experienced developers, they didn't speed up work. What's going on? Are benchmarks misleading us about AI capabilities? Are we missing something about how AI performs in the real world? In this talk, we'll reconcile lab and field evidence on AI capabilities. Drawing from METR's time horizon measurements and developer productivity RCT, we'll explore why impressive benchmark performance doesn't always translate to rea…
Watch on YouTube ↗
(saves to browser)
Chapters (18)
The Compute-Time Horizon Argument
1:43
Potential Constraints on AI Scaling (Power & Dollars)
4:23
The Problem of Eclipsing Evaluation Time
6:52
Meta's "J-Curve" of Developer Productivity
9:12
Unreliability of Self-Reported Time Estimates
11:43
Personal Experiences with AI Tools (Cursor) & Learning Curves
14:10
METR Study Deep Dive: Scatter Plots & Variance
16:48
The Controversy of "Conservative" Usage Estimates
21:41
Unpublished Hackathon Results (AI Allowed vs. Disallowed)
25:28
Why AI Struggles with Data Science & Messy Enterprise Data
30:35
Example of AI Failure on Complex Deployment Metrics
38:29
Quantifying Speed-Up: The Methodological Challenges
46:30
Future Metrics: "Watched" vs. "Unwatched" Time Horizons
52:52
Moving Beyond Benchmarks: "In the Wild" Transcripts
56:12
The "Agent Village" & Fuzzy Goal Measurement
58:53
The "Neurodivergent AI" Hypothesis & Interface Mismatch
1:06:31
Software-Only Singularity vs. Hardware Constraints
1:13:53
AI Applications in Chip Fabrication & Yield Improvement
Playlist
Uploads from AI Engineer · AI Engineer · 0 of 60
← Previous
Next →
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
AI Engineer Summit 2023 — DAY 1 Livestream
AI Engineer
AI Engineer Summit 2023 — DAY 2 Livestream
AI Engineer
Principles for Prompt Engineering - Karina Nguyen (Claude Instant @ Anthropic)
AI Engineer
Announcing the AI Engineer Network: Benjamin Dunphy
AI Engineer
The 1,000x AI Engineer: Swyx
AI Engineer
Building AI For All: Amjad Masad & Michele Catasta
AI Engineer
The Future of Work: Toran Bruce Richards, Silen Naihin et al
AI Engineer
The Age of the Agent: Flo Crivello
AI Engineer
See, Hear, Speak, Draw: Logan Kilpatrick & Simón Fishman
AI Engineer
Building Context-Aware Reasoning Applications with LangChain and LangSmith: Harrison Chase
AI Engineer
Pydantic is all you need: Jason Liu
AI Engineer
Building Blocks for LLM Systems & Products: Eugene Yan
AI Engineer
The Intelligent Interface: Sam Whitmore & Jason Yuan of New Computer
AI Engineer
Climbing the Ladder of Abstraction: Amelia Wattenberger
AI Engineer
Supabase Vector: The Postgres Vector database: Paul Copplestone
AI Engineer
[Workshop] AI Engineering 101
AI Engineer
The Hidden Life of Embeddings: Linus Lee
AI Engineer
[Workshop] AI Engineering 201: Inference
AI Engineer
The AI Pivot: With Chris White of Prefect & Bryan Bischof of Hex
AI Engineer
The AI Evolution: Mario Rodriguez, GitHub
AI Engineer
Move Fast Break Nothing: Dedy Kredo
AI Engineer
AI Engineering 201: The Rest of the Owl
AI Engineer
Building Reactive AI Apps: Matt Welsh
AI Engineer
Pragmatic AI with TypeChat: Daniel Rosenwasser
AI Engineer
Domain adaptation and fine-tuning for domain-specific LLMs: Abi Aryan
AI Engineer
Retrieval Augmented Generation in the Wild: Anton Troynikov
AI Engineer
Building Production-Ready RAG Applications: Jerry Liu
AI Engineer
120k players in a week: Lessons from the first viral CLIP app: Joseph Nelson
AI Engineer
The Weekend AI Engineer: Hassan El Mghari
AI Engineer
Harnessing the Power of LLMs Locally: Mithun Hunsur
AI Engineer
Trust, but Verify: Shreya Rajpal
AI Engineer
Open Questions for AI Engineering: Simon Willison
AI Engineer
Storyteller: Building Multi-modal Apps with TS & ModelFusion - Lars Grammel, PhD
AI Engineer
GPT Web App Generator - 10,000 apps created in a month: Matija Sosic
AI Engineer
Using AI to Build an Infinite Game: Jeff Schomay
AI Engineer
How to Become an AI Engineer from a Fullstack Background - Reid Mayo
AI Engineer
The Code AI Maturity Model and What It Means For You: Ado Kukic
AI Engineer
AI Engineer World’s Fair 2024 - Open Models track
AI Engineer
AI Engineer World’s Fair 2024 — Keynotes & CodeGen Track
AI Engineer
AI Engineer World’s Fair 2024 - Open Models track
AI Engineer
AI Engineer World’s Fair 2024 — GPUs & Inference Track
AI Engineer
AI Engineer World’s Fair 2024 - Keynotes & Multimodality track
AI Engineer
From Text to Vision to Voice Exploring Multimodality with Open AI: Romain Huet
AI Engineer
The Making of Devin by Cognition AI: Scott Wu
AI Engineer
The Future of Knowledge Assistants: Jerry Liu
AI Engineer
Llamafile: bringing AI to the masses with fast CPU inference: Stephen Hood and Justine Tunney
AI Engineer
Open Challenges for AI Engineering: Simon Willison
AI Engineer
Lessons From A Year Building With LLMs
AI Engineer
From Software Developer to AI Engineer: Antje Barth
AI Engineer
Unlocking Developer Productivity across CPU and GPU with MAX: Chris Lattner
AI Engineer
Copilots Everywhere: Thomas Dohmke and Eugene Yan
AI Engineer
Fixing bugs in Gemma, Llama, & Phi 3: Daniel Han
AI Engineer
Low Level Technicals of LLMs: Daniel Han
AI Engineer
Emergence Launch: AI Agents and the future enterprise: Dr. Satya Nitta
AI Engineer
How Codeium Breaks Through the Ceiling for Retrieval: Kevin Hou
AI Engineer
What's new from Anthropic and what's next: Alex Albert
AI Engineer
Using agents to build an agent company: Joao Moura
AI Engineer
Decoding the Decoder LLM without de code: Ishan Anand
AI Engineer
Running AI Application in Minutes w/ AI Templates: Gabriela de Queiroz, Pamela Fox, Harald Kirschner
AI Engineer
Building with Anthropic Claude: Prompt Workshop with Zack Witten
AI Engineer
DeepCamp AI