Week 11

Question 1

Prompt Engineering

Answer 1

Prompt engineering - is optimizing textual input to effectively communicate with LLMs. Becoming a prompt engineer is equal to learning to write clearly and concisely

Context + Specific Information + Intent + Response Format

Question 2

Prompt Priming

Answer 2

Prompt Priming

Problem Formulation: Determining what to ask the AI
- Starts with the basics: identify, analyze, and define the problem before you go into the AI Loop

LLMs are creative by design; knowing about hallucinations and confabulations, we must:
- Stop and think. Is AI an appropriate tool to use for the problem you defined
- If yes, then frame the problem in a way so the LLM understands exactly what we want

Problem formulation is the thinking you do before you attempt the prompt in AI
- outlining the focus, scope, and boundaries of the problem to ensure effectiveness
- focus on the precision you want in the output, and define boundaries (context, a target audience, constraints)

Exploration: Using major AI tools
- exploration means finding the best generative AI tool for the problem (ex. ChatGPT, Google Bard)

Question 3

Prompt Priming and Beyond

Answer 3

Prompt Priming and Beyond
Critical Thinking: weeding out poor AI content
- Generative AI tools can produce inaccurate, biased, and at times poor quality content … that’s not on the AI that’s on your as the user. You are accountable
- Critical thinking is the solution to this limitation (human in the loop)

Reflection: AI is here to augment, not replace
- you may be a great writer, a great designer, or even a great accountant, and perceive AI as a threat - how do you harness AI to give you superpowers
- remember: co-pilot is all about human in the loop, learning and iterating
- AI is not going to disappear, avoiding with blinders is a path to being left behind

Question 4

steps of co-piloting

Answer 4

steps of co-piloting:
- problem formulation: determining what to ask the AI
- exploration - using major AI tools
- critical thinking - weeding out poor AI content
- reflection - AI is here to augment, not replace

Question 5

Next Wave: AI and Productivity

Answer 5

Next Wave: AI and Productivity
ex. Excel (computational wizardry), PowerPoint (can automate first drafts), Word (will act like an editor)

Question 6

Next Wave: AI and Domain Specific Models

Answer 6

Next Wave: AI and Domain Specific Models
Domain specific - remixes the concept of LLMs by using smaller models that are trained on a subset of the larger data set along with proprietary data

Question 7

Why are Domain Specific Models so important

Answer 7

Importance of domain specific models:
- complexity and unique language with certain industries warrant a different approach
- Bloomberg serving as the ‘experiment’ for internal corporate AI models (Morgan Stanley has also launched one for the world of Wealth management)

Question 8

Next Wave: AI and Health Care

Answer 8

Next Wave: AI and Health Care
- ex read lengthy reports and allow doctors more time with patients
- ex. pattern match + diagnose + treat at scale that simply dwarfs what our current system provides

Question 9

Next Wave: Product Development

Answer 9

Next Wave: Product Development
Could see a $5B company with 1 full time employee

The wild future of GenerativeAI - the use case is essentially any natural language task you can think of across the entire end to end value chain
ex. design, software engineering, back-end/database, marketing, operations and post-transaction support

Key takeaway: being able to prompt effectively in natural language is going to be the new baseline to get started building a business due to the power of Generative AI

Question 10

Is Generative AI moving too fast

Answer 10

Is Generative AI moving too fast
- regulations are dramatically lagging and struggle to catch up
- fraud & crime on steroids
- employers trying to cap risk
- academic world upheaval
- could you really pause AI research

Question 11

AI Legal Considerations

Answer 11

AI Legal Considerations

Some types of machine learning models are legally prohibited because of the data or inability to identify how the model works - leads to discrimination

Consider red-lining laws and jailbreak

Question 12

Red-lining laws

Answer 12

Red-lining laws - in lending industry are a prime example since race and geography are highly correlated

Question 13

Jail break

Answer 13

Jailbreak - a user interaction strategy. That greats AI to break its own rules; LLMs like ChatGPT have Terms of Use that explicitly prohibit the jailbreaking model. To avoid jailbreaking, effective controls must be designed from the onset

Question 14

Considerations for the implementation of AI - Technical

Answer 14

Considerations for the TECHNICAL implementation of AI
- Computer power and data are two critical inputs for AI
- Data quality, inconsistent data, or inability to combine necessary data sources into a single data set capable of ingestion into machine learning systems are technical hurdles to deploying AI within a company
- the ability to capture data becomes a key activity and the dada itself a key resource before you can ever hope to meaningful leverage AI

Question 15

Considerations for the implementation of AI - Organizational

Answer 15

Considerations for the implementation of AI - Organizational
AI can be thought of as a information system - data, hardware (chips), software (model), people (engineers), and processes (data collection/abstraction)

• Transforming an organization into one that leverages data is much as a technical challenge as it is a people and change management challenge
• Processes, structure and culture are often the neglected areas of AI competency - without these AI will never shift from just something you do, to something that drives a competitve advantage (recall: be better by being different)
• Established AI needs a ton of data - with that data comes the need for information security, cybersecurity, and overall system redundancies

Question 16

Considerations for the implementation of AI - Societal

Answer 16

Considerations for the implementation of AI - Societal
- data misuse such as those Governments that have vast databases of PII and are less concerned about privacy than perhaps Canada, US, and EU
- at employer level, AI used to examine worker comms and monitor behaviour
- first mover advantage matters, and the data you collect and feed into an AI model reinforces that advantage
- fake news and DeepFakes - AI can be used for good to help detect fraud, but can also be used to sow misinformation and massive reduce the barriers to deceptive actions at scale

Question 17

Considerations for the implementation of AI - Ethical

Answer 17

Considerations for the implementation of AI - Ethical
Recall: Neural network – blackbox nature of machine learning means it is difficult and complex to understand why certain decisions are being made. Lack of transparency as AI decisions become less intelligible to humans

AI is NOT neutral. AI based decisions can be inaccurate (hallucinations and confabulations), and embedded with bias (based on training data, or coded biases, intentional or unintentional)

Question 18

Disruption breeds ____

Answer 18

Disruption breeds innovation

Question 19

Disruptive technologies

Answer 19

Disruptive technologies - are technologies that create market shocks and catalyze growth (also referred to disruptive innovation)

Question 20

Characteristics of disruptive technology

Answer 20

Characteristics of disruptive technology
Two primary characteristics of disruptive technology:
(1) they come to market with a set of performance attributes that existing customers don’t value (generally underperform at the low end of the market)

(2) over time the performance attributes improve to the point where they invade established markets – when this happens, it is often too late for the incumbent to defend

Example: digital cameras have all but wiped-out traditional film

Question 21

Once the performance attributes improve to where a new entrant can invade established markets, the new technology is often much ____ and _____than existing technology

Answer 21

Once the performance attributes improve to where a new entrant can invade established markets, the new technology is often much cheaper and better than existing technology

better and cheaper = displace the market leader and now the new entrant has not just the low end of the market, but also the high end. This is disruption

Question 22

Sustaining Innovation

Answer 22

Sustaining innovation - occurs when a company creates a better performing product to sell for higher profits to its existing customers
- leveraged by companies already successful in the industries
- motivation is profit; better products, pursue even higher profit margins

In isolation, it sounds like a solid strategy yet most successful companies built on only sustaining innovation will fail

Question 23

Disruptive Innovation

Answer 23

Disruptive Innovation - occurs when a company with fewer resources (and often a lower performing product) move upmarket to challenge the incumbents core business
- low-end disruption
- new market disruption

Question 24

Low-end disruption

Answer 24

Low-end disruption - company uses a low cost model to enter at the bottom of an existing market and claim a segment, causing the incumbent to retreat upmarket to make higher profit margins. This is a form of asymmetric motivation

Question 25

New market disruption

Answer 25

New market disruption - company creates and claims a new segment in an existing market by catering to an underserved or neglected customer base, slowly improving in quality until incumbent business are obsolete

Question 26

Sustaining Innovation vs Disruptive Innovation

Answer 26

Sustaining Innovation
- perform better and are of high quality
- willing to pay higher prices for higher quality products
- high-profit, high margin

Disruptive Innovation
- good enough and usually ignored by top end of market
- overserved by current offerings (or not served at all) settle for “good enough”
- low cost, low profit

Question 27

Sustaining innovation gives us a model to explain that …

Answer 27

Sustaining innovation gives us a model to explain that all technologies are constantly evolving to better performance and lower cost

Radical or more incremental change, btu the common features are:
- increasing the performance of established products
- within the main dimensions that customers want
- ex. cars, planes, computers

Question 28

what do innovation strategies drive?

Answer 28

innovation strategies drive purposeful, protective product decisions to disrupt an industry, or avoid being disrupted by another organization

Question 29

disruptors strategy for creating low-end disruption

Answer 29

disruptors strategy for creating low-end disruption:
to disrupt form the low end and launch a new growth business within an existing market:
- aim at the same markets dominated by industry leaders
- disrupt the industry leaders business model by harnessing the power of asymmetric motivation
- focus on least-demanding tiers of a market
- can the new entrant answer “yes” to these questions – (1) are the prevailing products more than good enough (2) can you create a new business model

Question 30

disruptors strategy for creating a new market disruption

Answer 30

disruptors strategy for creating a new market disruption:
to create a new market as a base for disruption and compete against non-consumption
- look for potential customers who haven’t been able to “do it themselves” for lack of money or skills
- target innovation at the segment of “non-consumption” to start

Examples: DropBox, Charles Schwab, Robinhood

can the new entrant answer yes to these questions
(1) is the innovation aimed at customers who will welcome a simple/good enough product
(2) will the innovation help customers do more easily and effectively something that they are already trying to do

Question 31

why do incumbent firms fail

Answer 31

incumbent firms fail because:
- its not neglect or a lack of awareness, its because they are hyper-fixated on listening to existing customers (difficulty in strategy change: cash cow vs new businesses with uncertain returns)
- because disruption tends to play at the edge, its easy to rationalize current customers don’t want the lesser performing thing
- most disruptive technologies tend to perform poorly on traditional financial metrics upon launch (course versus justify the capital allocations to power something underperforming)
- startups amass expertise quickly. Big firms forced to play catch-up. few ever close the gap with new leaders

Question 32

what happens when a potential disruptor is spotted

Answer 32

potential disruptor is spotted:
- build a portfolio of options on emerging technologies
- give some freedom to emerging division and possibly even separate it altogether to allow innovation with higher degree of autonomy
- note on acquiring your way out of a missing market and being disrupted: most acquisition fail to ever deliver the economic benefits (synergies) modelled out

Question 33

Disruption happens from ______ or ________

Answer 33

Disruption happens from below or from above
Below - cheaper, typical disruption theory - good enough
Above - more relevant fictional, market creation, new segment creation

Question 34

Big Bang Disruption

Answer 34

Big Bang Disruption
- these types of innovations compete from day 1 and are often called unexpected disruption
- they are BETTER and CHEAPER than existing products and disrupt above and below simultaneously
- large-scale fast-paced innovation that can disrupt stable businesses very rapidly
- unplanned and unintentional
- do not follow conventional strategic paths or normal patterns of market adoption

Question 35

Classic Technology curve

Answer 35

Classic technology curve
A new technology is adopted slowly in stages by a series of customer types:
- starts with a small group of investors
- moves to early adaptors
- then moves to cross the chasm to early majority and then the late majority
- eventually picks up the laggards

Question 36

Classic technology curve versus big bang adoption

Answer 36

Classic technology curve versus big bang adoption:
- in the classic Moore model, innovations start small (innovators) then flow to early adaptors and then move to cross the ‘chasm’ to early majority and so on
- big bang model similarly has the niche start (trial users) but then rapidly jumps to the vast majority gaining widespread adoption almost instantly
- in a digital world (where bits dominate over atoms), disruptions can happen faster and on a larger scale than ever before
- in a world of bits, a product can go from launch to peak decline very quickly
- it does not mean disruptive companies decline as rapidly - this is referring to the products having shorter lifespans
- these companies build their DNA around constantly launching new products, and combining big bang with sustaining + disruptive disruption to continuously improve:
examples: Uber –> UberEats, AirBnb –> Long-term stays

Question 37

Why do we care about the aggressive drop

Answer 37

Why do we care about the aggressive drop:
software breeds speed. bits faster than atoms when it comes to innovation and thus lifestyle of products can be much shorter in the digital/software age

Question 38

3 characteristics of the big bang disruption

Answer 38

3 characteristics of the big bang disruption
(1) unencumbered development
(2) unconstrained growth
(3) undisciplined strategy

Question 39

characteristics of the big bang disruption –> unencumbered development

Answer 39

characteristics of the big bang disruption –> unencumbered development
- born of rapid fire, low cost experiments on fast-maturing ubiquitous technology platforms
- built out of readily available components (APIs, open-source, low-code/no-code) with lower cost

Question 40

characteristics of the big bang disruption –> unconstrained growth

Answer 40

characteristics of the big bang disruption –> unconstrained growth
- marketable to every segment simultaneously

Question 41

characteristics of the big bang disruption –> undisciplined strategy

Answer 41

characteristics of the big bang disruption –> undisciplined strategy
- better (10x) performance, lower price, greater customization

Question 42

Examples of big bang disruption

Answer 42

Examples of big bang disruption
- Google Maps
- Spotify
- Zoom

Question 43

Defining characteristics of the big bang disruption

Answer 43

Defining characteristics of the big bang disruption
Big bang disruption is not predictable and not linear. It does require technological advances that allow new entrants to immediately offer better and cheaper products than existing solutions

Question 44

Examples - low end disruption

Answer 44

• disrupting business model from low-end; aim at the same markets dominated by industry leaders
• least demanding tiers of the market
• product usually has a disruptive business model
• ex. chromebook, southwest

Question 45

Examples high end disruption

Answer 45

• create a new market for a base for disruption
• competing against non-consumption
• ex.

Question 46

how do you enhance your disruption radar

Answer 46

how do you enhance your disruption radar
- remove short-sighted, customer-focused, and bottom-line obsessed blinders
- have conversations with those on the experimental edge of advancements
- increase conversations across product groups and between managers and technologists to share learnings and promote collaborative innovation
- if employees are quitting to join a new company, with breakthrough/different tech in your industry, it might be worth paying attention to