How I improved my knowledge of transparency in AI in one evening by Georgina Hodgkinson

[A guest post from one of our members, Georgina Hodgkinson].

I attended a Meetup at Skills Matter hosted by BrianLinuxing from Linuxing in London on 19th June. I travel from Milton Keynes to attend these events when I can. They always have a packed agenda, speakers are impressive and the chance to come home with some swag.

The title of the meetup grabbed my attention, the Wonders of IBM, Watson, artificial intelligence, Swift coding and databases. AI is a huge topic and gaining another perspective, more nuggets to add to the toolkit is always useful.

I then look through the speakers – Louise Richardson, Business Development Consultant, former channel manager of IBM, David Spurway, CTO for IBM Power Systems, Zoe Osorio and Sean Greaves, Client Technical Specialists at IBM, and Matt Kilner, IBM development leader for Swift, IBM Hybrid Cloud.

This is no amateur night.

Linuxing in London has an impressive network. It is the largest Linux group in London. Lead by the friendly and welcoming Brian Byrne, he sets the tone for the group. He is proactive for women coders to attend his events and as such I see a balance of gender that is rarely seen. The network is friendly, eager to learn and share, all in the relaxed atmosphere of Skills Matter and getting to know this group while waiting in line for a pizza even vegetarian or gluten free or other food that has been arranged by sponsors. [Kindly provided by IBM].

1. Saving the planet with Linux and the IBM Z Mainframe – Louise Richardson at Cognition Foundry

The evening started with a story about a vision. A successful businessman on holiday swimming and touches yet another bit a plastic in the ocean. Something should be done. A business of Plastic Bank is founded by David Katz and Shaun Frankson.

I love Shaun Frankson’s quote from this video “code will be the script to save the planet”

IMG_5790.JPG

Cognition Foundry is looking for visionaries, dreamers who think big. They democratise IT and make it accessible for start-ups. [Editor’s note: Louise and Cognition Foundry donated three wonderful Echo Dot devices into the free raffle. Her talk can been seen here.]

 

 

2. Using Watson to talk to traditional systems without breaking the bank – David Spurway

Wow, David Spurway, CTO for IBM Power Systems. This man can cover valuable content at pace, his energy and passion is apparent. If you are someone that loves to see an overview of start to finish to have a bird’s eye view of a technology and then go back to revisit the details you will love David Spurway.

He is structured, he has an agenda that a mere mortal would not attempt, but he reassures you that the content is covered in his blog for later catch-up and there is always SkillsCasts that Skills Matter provide to recap.

His talk is personal, entertaining, accessible, taking the audience with him on his journey to leave no one behind and everyone gets something out of his talk. He combines human examples with memorable photos. He refers to his colleagues that have helped him produce this talk, he invites the audience to enhance and add extra experience. He is very inclusive.

IMG_20190619_185807.jpg

He talked about IBM Think. He is down to earth and refers to going with an idea of what to do or he would “drown in all the different things you can learn about”.

Combing hardware Iron man with AI to solving passwords – 50% of service desk issues.

Hands on Lab: How to build a Chabot using IBM Watson.

Digital Humans  creating avatars to feedback emotion when using a Chabot. More interactive but not trying to trick and fool you into thinking it was a real person. Cora already being used in RBS.

Referenced to Ross Cruickshank’s workshop.

His need to automate his own FAQ as to when computer systems come out of support or are at end of life inspired him to use these technologies and apply to his own need. He then talked through from beginning to end his thinking, designing, applying and plans to improve.

3. Everything you wanted to know about Swift but were afraid to ask

Matt Kilner is a senior developer at Swift@IBM, working to bring Swift to Linux and the Cloud. It was a a great introduction for me to Swift. A whistle stop tour of the history of Swift and Apple, December 2015 it was open sourced, IBM working with Apple and community to bring Swift onto Linux, 5.1 is now in beta version.

IMG_5810.JPG

Clarity and reasons as to why you should use Swift. It aims to make programming easier. The principles are safe, expressive and fast. Comparison of languages is comparable to Nodejs, Java and Typescript but has a lower footprint.

The language was explained and was comparable to other languages.

 

4. Designing Trust and Transparency with Artificial Intelligence (AI) – Zoe Osorio and Sean Greaves

This was the one that I was looking forward to. I attend Artificial Intelligence meetups, self-studying AI with Udemy, and other experts. The vibe of applying AI in a practical large scale is met with fear and paving the way for compliance and governance. As deep learning is becoming more and more opaque how this is going to be addressed was something I was very interested in learning.

Zoe and Sean from IBM did not disappoint with their talk. They explained from machine learning perspective about bias and fairness. The diagram they used was a great tool at explaining this. They used great accessible case studies so that I could latch onto the data science meanings without having to know how. So much was covered in this session about how models are created and having a knowledge of this should be something everybody in contact with AI should have an awareness of.

Machine learning models are only as good as the data you give them. For example, historically drug trials weren’t done on women however the data means that drug prediction and drug discovery learning models need to be aware of the data scarcity.

IMG_5824.JPG

Cultural issues of AI and not the technical side of AI is now a major part of implementing AI. Data is from the past with its prejudice, bias and unevenly distributed data. Technology is revealing what has always been there but makes the bias more visible so that these ethical problems need to be debated and known when using these datasets.

This is why AI to augment human ability as opposed to replacing it, is the way things are naturally evolving.

Tools:

AI Fairness 360
Watson OpenScale
H2O Driverless AI
lime