AI Strategy

“All models are wrong, but some are useful” George Box, https://www.forbes.com/sites/bernardmarr/2019/03/19/how-to-develop-an-artificial-intelligence-strategy-9-things-every-business-must-include/#40f7af128360

Helps drive Competitiveness, hyper-personalisation, operational efficiency, and customer experience

Examples of projects: More personalised user experience (recommendation system More efficient processes (automation Customer churn Offer a discount or send a gentle nudge to close the sale for the customer Personalised consultation

Key elements of the AI strategy: And were talking about the pragmatic powered by Machine Learning (narrow focused) not pure AI (the one mimicking human intelligence)

Set the right expectations! High but pragmatic! ML algorithms analyse data to model that make predictions, take decisions, or identify context.

Choose multiple high-ROI use cases

There are as many use cases as there are business processes and customer experiences

Choose multiple AI projects like a VC chooses startups

Steps in order to get the AI strategy working:

• Get the data in order

Data sources available:

• Customer transactions
• Points of sale
• Inventory
• Product
• Supplier transactions
• Marketing
• Human Resources

How the model will manifest itself in the business process Engage with the business, design, and app dev early, data eng and the end users

Let people know that projects are in progress, and on the way Involve stakeholders in the process

Practice the responsible AI AI solutions are not perfect, like us, they are probabilistic Explain AI - how it will work, explain the models, and the process; Responsible AI mitigates risks Bias in the model, outliers, explain, document, Ethical consideration

Stay in charge! Mitigate the risks, govern edge cases, knowledge engineering - rules for the model (don’t make a loan >1mln$ or crucial/sensitive decisions) Human at the beginning, human nat the end

7 key elements of an Enterprise AI strategy Set proper expectations for AI Choose multiple ROI use-cases Get your hdata house in order Make use of state of the art tools Make AI a team sport Practice reponsible AI Stay in charge

6th May H2O Talk AI implementation Allows to shift from the reactive post transaction to Pre decision High level of automation, efficiency, and effectiveness Transformed Decision Making

ROI: Investment in DS Team Infrastructure

What are the challenges Time Talent Trust! - embed into ethos, culture, and processes Building the trust is a process:

• Educate (invite experts

Implementing The Vision

The Success Treating AI as a Strategic Change Management Initiative Buliding a Cross-fuctional team Selecting the right use cases and then expand - Achievable, high value, focussed and near term, data ready! Balanced Management of Expectations Planning and Managing your investment in AI - measure, purpose, cost, iterate

What’s next

From the Brian Cox meeting What is AI - different definitions Cognitive science, engineering, computer science

The Turing test??!!!!

No precise definition Machine learning specific tasks Very targeted approach What is intelligence? Ugh machine abilities progress within last few years How can we make artificial intelligence to work for least privileged if human intelligence can’t? How can me make AI work to improve quality of life and suppress inequalities?

General applications of AI for public and Most recent AI interesting and not so obvious apps!!! How it progressed over time(coding by examples) Lawyers example, healthcare, Amazon, Cambridge analytica good and bad examples

Main personal, AI driven goal: To find a proof for an interselestial life or this life itself(given the probable quantum computing powers)

Social skills, empathy, Explanatory AI Should we focus on this as much as we are now? Do we actually understand doctors explanation of the treatment? Why would we want to have an explanation form AI? It’s more about building trust between ppl and AI My bet and prediction: we won’t understand what was the underlying thirty behind the AI decision until we understand what lays behind human decision process and this won’t happen until we have quantum computers and another Einstein mind in psychology and or cognitive science. How can we understand AI when we don’t understand human brain?

Application in healthcare are predictably going to be improving massively. Discovery of different problems and delivery of various solutions Depersonalisation experiences Or personalised experiences? Wouldn’t it create an (?) bubble? Possibly the best idea is kind of mixed model of solution Hum? Technology wearable headband working memory increases by 20% Regulations? Framework? Guidelines? Principles? Ethic bible? Do we regulate other sciences like math, physics or biology? We do regulate industries cars, insurances, aircrafts And so on.

It will dramatically positively impact the world All of panelists’ agreed on that Will we always have control over the AI? People are phenomenal complicated intelligent computers If we had all the super intelligent people we wouldn’t never voted brexit Very good explanation how people developed different technologies. They weren’t asking the question why and how we start to fly, they started gluing feathers to their arms and tried to fly first before hundreds or maybe thousands years later the aerodynamics answered that question. Spam killing AI system which figure out that the best solution is to kill people creating spam! Lol The last techmology in the iPhone 10 - the cat in there is worth 15mln$

0 level: expertise level: individual contributor, team leader, manager, or CTO? 1 level: industry expertise ie automobile, retail, manufacturing etc 2 level: domain knowledge: sales, marketing, other 3 level: data science skills I want to develop, ie data wrangling, advanced analytics, visualisations, presentation and managing stakeholder expectations 4 level: ML expertise like NLP, CV etc