Myself and Ansel attended this interesting conference that brings together a few different communities.  The single hall had areas dedicated to Data, Information Security, Development Operations, Communication Technologies and Customer Experience. 

It was interesting to see if there are common themes among this quite diverse set of communities. 

The following are a few of the sessions that we attended

Immediately we walked into the conference we were by a talk given by Capital One on elevating data literacy.  This felt very relevant to me given my recent time with NHS Supply Chain.  The gentleman spoke of lots of “Lunch and Learns”    There seemed a lot of polling to find skills that were being demanded.  This reminded me of Steve Miranda’s adage “This is not a democracy”.  The most requested skill was excel.  It made me wonder whether having many more people take extracts into excel would improve decision making or create a tower of babel. 

The next session that we were keen to attend was a panel discussion Governance to Guardrails.  There were speakers from ITV, Pandora.  While we thought it was going to focus on AI, the data theme seemed to come through most strongly.  Issues mentioned included: understanding data provenance, data governance.  There were some more AI themes that came through: emphasis on human involvement, emphasis on safe experiments giving visibility to the model.  They mentioned a community of practice which resonated with me.  Somewhere that you are comfortable to not appear silly with a question.  They called it an AI Guild which seemed to combine the very new with the very old.  They did talk about sharing prompts. The creating of prompts is the next artisan craft. 

They also talked about build versus buy in LLM models, which seemed to misunderstand the creation of tooling on top of very few LLM’s that will exist.  However, they made the point about ensuring that you understand the benefits that you are trying to achieve as well as the risks that you are absorbing as you take AI capability from suppliers. 

The standout moment for me in this panel discussion was in the Q&A section, when someone made the point (rather than asked the question) that the biggest issue in LLM based development is failure modes.  What do you consider an error when the objective is essentially to give your best guess?

I attended a session on crisis management.  It was very much a security incident response session.  It covered many aspects: business continuity, crisis communications and information supply chain risks. 

The gentleman and lady that presented highlighted mistakes that companies often make.

1.  Containment

   1. The belief that you can disengage the company systems from the public internet. 

   2. Teams then find that all of the tools that they need are on the other side of the fence. 

   3. A scattered enterprise can't cut itself off from itself

   4. Cutting yourself off from your customers because you are compromised yourself. 

   5. Contain and eradicate rather than bring back online. 

2. Communications

   1. Clients tend to be better at external Comms than Internal.

   2. Teams using what's app when internal systems are down

   3. Comms templates for cyber. 

   4. Channels for interfacing to outside worlds.

3. Not knowing the impact of the attack.

   1. Rolls Royce has 15,000 organizations in their supply chain.

   2. Business impact analysis

   3. Redundancy

I attended this session because I thought it might be on the bleeding edge of user experience design, but I was really more about the history of crafts moving into new crafts in Digital.  It was delivered by the University of Arts in London (UAL). 

Amazing presentation taking you through design of the AV systems for learning delivery, right through to the creation of virtual environments in cinematography.