MIT
Sloan Management Review arranged AI webinar named “Adopting AI: Picking the
Right Problems to Solve” on 11th Dec 2020. Executives in most industries have heard
about artificial intelligence’s potential to transform their businesses, but
many aren’t sure where to start — or how to gain the most value from efforts
already underway. This free webinar addressed those AI adoption hurdles with
expert insights.
Firstly,
Ms. Abbie Lundberg (President, Lundberg Media) was event moderator and she
introduced all speakers.
Ms.
Kimberly Nevala (Strategic adviser, SAS) was the first speaker and her topic
named “Adopting AI: Problems AI Can Solve”. She firstly briefed AI is “The
science of training systems to emulate human tasks through learning and
automation.”
Then
Ms. Neyala mentioned what AI do including to learn from experience, adjust to
new input, automate discrete tasks and engage intuitively. She showed an
interested tree to identify AI and quoted Prof. Michael Wade (Innovation and
Strategy IDM Switzerland) statement to explain that “A lot of people want the
AI hammer so they can whack things. But
it’s an expensive hammer and it takes time, and the right people, to get up to
speed. So the first question … is
whether they really need AI – because in may cases, they don’t.”
So
that she said to apply AI to problems if they are Meaningful, Well-Bounded, Data
Rich, Complex (enough) and At Scale.
Ms.
Monica Livingston (Senior director, AI sales, Intel) was the second speaker and
her presentation was “AI or Analytics?” Ms. Livingston briefed the extract
valuable insight from data such as internet user upto 25GB per month, Smart Car
upto 50GB per day and Smart Factory upto 1PB per day!
Then
Ms. Livingston stated the AI opportunity assessment through the business value
and solution cost to prioritize business challenges.
Smart
factory was identified as one of best approach to use AI. Because large manufacturer used data to
improve its operations with each challenge using a different approach to deploy
maximum business value at the lowest possible cost. Then she introduced different AI tools for
different application.
After
that she mentioned the different approach to Analytics and AI including Machine
Learning and Deep Learning.
Finally,
Ms. Monica Livingston quoted Chief Data Scientist Survey on 2nd Nov
2020 that the most important for DS/ML and model challenge is explainability of
AI. The last diagram explained the
accuracy and explainability of AI model.
Prof.
Pete Smith (Chief analytics officer and professor of modern languages, University
of Texas at Arlington) was the last speaker and his topic entitled “Forecasting
AI in Your Enterprise”. Firstly, he
employed weather broadcaster (e.g. European model) to explain power of
modeling.
Then
Prof. Smith discussed the selection of the first AI targets in your context and
the criteria included checklist for complex, example-based, situation-based,
myth based and driven by core business processes.
And then Prof. Smith selected educational context for
AI project such as student progression to graduation.
After
that Prof. Smith explained framing AI problems and approaches from x-axis from
unknown (discovery) to known (optimization) and y-axis from Big Data
(quantitative) to Thick Data (qualitative). Thick data is qualitative
information that provides insights into the everyday emotional lives of
consumers. He said “Big Data needs Thick
Data”.
Finally,
Prof. Pete Smith showed some insight from Natural Language Processing (NLP) and
Natural Language Understanding (NLU) as AI
problem. He concluded that look for business-critical decisions rather than
available data sets, decisions that could be made more “smartly” or intelligently
via modeling. Consider tools with “embedded analytics” across the wide variety
of domains. He reminded that learn from
best practices and surface digital ethics at the end.
Q&A
Session.
Reference:
MIT
Sloan Management Review - https://sloanreview.mit.edu/
AI webinar registration page - https://sloanreview.mit.edu/connections-webinar-adopting-ai/
沒有留言:
發佈留言