Kicking off day 1 of #ODSCWest. I'll be tweeting through the next few days and pinging #GCDigital folks for interest or potential follow-up. #opensource #DataScience #openeducation @School_GC

First up: @asuonline will be hosting an educational analytics conference in March 2019 focusing on what we can learn from learner data and how to structure learning to better meet learner needs. #ODSCWest

#AI differential - being able to tailor products to specific users == better able to predict value to and from users. E.g. cost to deliver value to user vs. cost to company to deliver that value. Allows ruthless competitive focus - @DataRobot #ODSCWest

Given our data, huge number of #machinelearning models of interest to #GCDigital - security clearance, staffing, physical space optimisation, training & learning, regulatory & legal frameworks, risk assessments, energy consumption, service delivery and many more...

Interesting idea taking place this PM at #ODSCWest - reverse pitches. Investors pitch problems to startups. @sheamusmcgov

Keynote on future of #AI - focus on automation, meta-learning and unified framework for #machinelearning - #ODSCWest. Not there yet. New unique pipelines for every problem. Future == single pipeline for data.

Recommended reading: The Master Algorithm. amazon.ca/Master-Algorit…

#MachineLearning tools and procedures have been around for decades. Difference == computational power and storage. But we are still very early days.

To do #machinelearning today requires a pipeline of pre-processing, feature cleaning, feature engineering and model selection. Dozens of options and decisions in each step of the pipeline. Path forward == focus on meta-learning - learning about learning. #ODSCWest

Leads to a potential unified framework for #AI. Different paths between data and goals/outcomes. Makes me think of Lego blocs. Right now, we're still building the blocks (molecule by molecule), but folks are working on putting them together. Gets better every day. #ODSCWest

Predicted results == optimising over broader sets of pipeline. & then optimising optimisers. #ODSCWest

Practical implications == automation of pipeline == putting technology in hands of non-data scientists. Transform businesses and increase scope of problems #datascience can handle. #ODSCWest

Still huge amounts of thinking re: #bias in our #data and trickle-down impacts of predictions. (@TBS_Canada is doing cool work on an ethical framework for #AI in this space @supergovernance @CorriveauNoel ) #ODSCWest

Next up: Reza Zadeh @matroid - Turning #ML research into products for industry. Challenges: adequate accuracy, scaling #ML and user interfaces. #ODSCWest

Lol. "Self driving cars have been 5 years away for 13 years." Accuracy matters in #machinelearning & we underestimated what good enough needs to look like.

Took 40 years of accuracy in machine translation to get to a successful product (started with tapping submarine cables in cold war). Speech recognietion is just getting there now as well. Only took 70 years. Next up: computer vision (nouns not verbs)? #ODSCWest

Work to advance adequate accuracy can cost us a lot more than we think given the state of euphoria around #AI today. Subjective acceptance matters, but it's important to think about what we really need for success. Easy to trip. #ODSCWest

Next: Scaling #machinelearning == balance between model complexity, computing power. Lots of work happening in #AI chips to optimize on different tasks. Stack called #Software2point0 (This is all @Reza_Zadeh, btw) #ODSCWest

Takeaway: model your computational power requirements and plan accordingly to bring something to market. #ODSCWest

Next: User Interfaces (I'm reading in #UX here) often overlooked by #AI experts. #UI experts tend to assume that results are perfect. Need much stronger cross-walks to meet user expectations.& able to assume mistakes & get feedback from users and improve #UX of our #AI products.