anlak storytelling
All kinds
Document kinds
Loading document kinds...

Primary corpus lens for browsing and insights. Selection lives in the URL.

All formats
Orientation
Loading formats...

Landscape, portrait, square, or unknown document geometry.

Unlock
Oliver Wyman · consulting-deck

GPT-3 and the actuarial landscape

46 pages · 4 arc beats · 3 loops

Consulting-Deck Download JSON
Unlock

GPT-3 and the actuarial landscape

Oliver Wyman
arc beats above · slides in the middle · loops below · scroll → 3 LOOPS
SETUP TENSION ANALYSIS EVIDENCE RESOLUTION APPENDIX
HOVER FOR DETAILS · CLICK A SLIDE FOR FULLSCREEN · STEP 2
— No image bound to this slide. Wire up imgproxy to display the rendered slide JPEG.
—

Deck intelligence map

4
coverage by narrative range · generated from this deck JSON
Narrative range 46 total
Metadata
Components
Metrics
Tools
Frameworks
Beats
Loops
What Is 5 slides 100% 5/5 slides 100% 5/5 slides · 19 hits
— 0/5 slides
40% 2/5 slides
— 0/5 slides
100% 5/5 slides 20% 1/5 slides What Is 11 slides 100% 11/11 slides 100% 11/11 slides · 59 hits 45.5% 5/11 slides 54.5% 6/11 slides · 7 hits 9.1% 1/11 slides 100% 11/11 slides 45.5% 5/11 slides What Could Be 15 slides 100% 15/15 slides 100% 15/15 slides · 68 hits 6.7% 1/15 slides 33.3% 5/15 slides 40% 6/15 slides 100% 15/15 slides 60% 9/15 slides New Bliss 15 slides 100% 15/15 slides 100% 15/15 slides · 69 hits
— 0/15 slides
40% 6/15 slides 13.3% 2/15 slides 100% 15/15 slides 40% 6/15 slides

Slide inventory

46
every slide · same image gating as the playbook
01
Slide 1
front_matter
Open slide detailBeat · What Is
02
other
Open slide detailBeat · What Is
03
other
Open slide detailBeat · What Is
04
summarize
Open slide detailBeat · What Is
05
transition
Open slide detailBeat · What IsLoop · Golden Circle
06
present_framework
Open slide detailBeat · What IsLoop · Golden Circle
07
The slide highlights the convergence of data availability, hardware, software frameworks, and algorithmic innovation.establish_context
Open slide detailBeat · What IsLoop · Golden Circle
08
transition
Open slide detailBeat · What IsLoop · Golden Circle
09
The slide serves as an educational visual aid for a machine learning presentation.establish_context
Open slide detailBeat · What IsLoop · Golden Circle
10
The slide provides the standard OLS (Ordinary Least Squares) formulas for linear regression, though the title mentions 'Gradient Descent' which is an alternativpresent_framework
Open slide detailBeat · What IsLoop · Golden Circle
11
The slide uses a standard linear regression example to illustrate basic ML concepts.present_framework
Open slide detailBeat · What Is
12
The slide explains the derivation of beta coefficients for a simple linear regression model.present_framework
Open slide detailBeat · What Is
13
The slide sets up a transition from closed-form solutions to iterative optimization.transition
Open slide detailBeat · What Is
14
The slide contrasts the iterative nature of gradient descent (no closed-form formula for coefficients) with the model structure and cost function.present_framework
Open slide detailBeat · What Is
15
The slide illustrates the iterative nature of gradient descent by showing a poor initial fit (red line) for the data points.present_framework
Open slide detailBeat · What Is
16
The slide illustrates the iterative nature of gradient descent by showing a poor initial fit (red line) for the data points.present_framework
Open slide detailBeat · What Is
17
The slide uses a structured layout to define model components (structure, cost function, coefficients, resulting model) and shows the resulting regression line present_framework
Open slide detailBeat · What Could BeLoop · Mece Breakdown
18
compare_options
Open slide detailBeat · What Could BeLoop · Mece Breakdown
19
The slide uses a node-link diagram to illustrate the mathematical structure of three different machine learning models.present_framework
Open slide detailBeat · What Could BeLoop · Mece Breakdown
20
The slide illustrates the difference between a single sigmoid output node (binary/single-label) and multiple sigmoid output nodes (multi-label/multi-class) in apresent_framework
Open slide detailBeat · What Could BeLoop · Mece Breakdown
21
The slide illustrates a standard feed-forward neural network with an input layer, two hidden layers, and an output layer, specifically applied to the Iris dataspresent_framework
Open slide detailBeat · What Could BeLoop · Mece Breakdown
22
transition
Open slide detailBeat · What Could BeLoop · Mece Breakdown
23
Explains the technical process of how GPT models convert text to tokens and predict the next token as a classification task.present_framework
Open slide detailBeat · What Could BeLoop · Mece Breakdown
24
present_framework
Open slide detailBeat · What Could BeLoop · Mece Breakdown
25
Explains the mechanism of adding positional encoding vectors to word embedding vectors to preserve sequence information.present_framework
Open slide detailBeat · What Could BeLoop · Mece Breakdown
26
The slide uses a simple linguistic example to explain the concept of self-attention in AI models.present_framework
Open slide detailBeat · What Could Be
27
This slide illustrates the core linear algebra operations (matrix multiplication) used in the Transformer architecture's attention mechanism.present_framework
Open slide detailBeat · What Could Be
28
The slide explains the mathematical steps of the self-attention mechanism in transformer models.present_framework
Open slide detailBeat · What Could Be
29
The slide uses a direct screenshot of a ChatGPT interface to illustrate the capabilities of transformer architecture.illustrate_case
Open slide detailBeat · What Could Be
30
The diagram maps the standard Transformer architecture (Attention is All You Need) to a specific example involving actuarial text processing.present_framework
Open slide detailBeat · What Could Be
31
The slide illustrates the reduction in required demonstrations as models move from fine-tuning to zero-shot learning.present_framework
Open slide detailBeat · What Could Be
32
The slide uses a circular process diagram to represent a linear 4-step sequence.present_framework
Open slide detailBeat · New Bliss
33
transition
Open slide detailBeat · New Bliss
34
The slide shows a side-by-side comparison of a Jupyter notebook environment and the GitHub Copilot suggestion interface.illustrate_case
Open slide detailBeat · New BlissLoop · Quick Win Big Bet
35
The slide shows a two-column layout: left side details data loading steps, right side provides Python code for plotting histograms with average target overlays.illustrate_case
Open slide detailBeat · New BlissLoop · Quick Win Big Bet
36
The slide demonstrates a specific use case for LLMs in technical workflows.illustrate_case
Open slide detailBeat · New BlissLoop · Quick Win Big Bet
37
present_solution
Open slide detailBeat · New BlissLoop · Quick Win Big Bet
38
diagnose
Open slide detailBeat · New BlissLoop · Quick Win Big Bet
39
transition
Open slide detailBeat · New BlissLoop · Quick Win Big Bet
40
establish_context
Open slide detailBeat · New Bliss
41
present_framework
Open slide detailBeat · New Bliss
42
transition
Open slide detailBeat · New Bliss
43
summarize
Open slide detailBeat · New Bliss
44
state_next_steps
Open slide detailBeat · New Bliss
45
front_matter
Open slide detailBeat · New Bliss
46
front_matter
Open slide detailBeat · New Bliss