AIIA Tour, Day 4

Firstly, a sincere apology to the folk at WorkDay. In the best real estate in New York competition, I claimed that they were on the 35th floor, when in fact they inhabit level 49. However, although I robbed them of 14 floors, they still didn’t win the comp😉.

For week 2 we’ve travelled across the US to Seattle and today we kicked off with a visit to AWS where Brad Fisher reminded us that the first AWS region in Australia only opened in 2012, which really wasn’t all that long ago considering the cloud revolution that’s happened/happening in Australian public sector. We had a refreshing change of format with Darren Canavor (from the security leadership team at Amazon, who more patents than Bezos, but no social media presence) being interviewed on a range of topics and covering a lot more than he was asked questions about. It was a fascinating session which ranged from intelligent authentication/authorisation systems to robotic arms that learnt do anything (including flipping water bottles with algorithms and A/B tests), to AI in shopping (I want a rug that looks like this other thing I own). His thoughts on security and Gen-AI were also fascinating and the idea of using side-channel techniques to understand when a model was misbehaving was thought-provoking. Lastly, I really loved the idea of ensuring you look for kindness as an attribute in your team members so that they can accept and respond well to harsh criticism. We then had Tom Soderstrom, Executive Enterprise Strategist, give us a wide-ranging session on tech futures as well as some great storied from NASA and JPL. I liked lesson 6 from past tech revolutions “value from new ideas explodes” with the advice that you should “light your first street with the new tech… now”. I’m curious about his prediction that we’ll all come back to the office 5 days a week so we can improve creativity, but I agreed with his closing comments including “think big, start small”, “Love the problem, not the solution” and “Experiment, experiment, experiment”.

Session 2 was a trip to SalesForce. I firmly believe that SalesForce needs a company name change, given what they do now has gone far beyond Sales CRM systems. Paul Tatum gave us an overview of SalesForce who started as a cloud company 24 years ago (I’m not sure I’d heard of “cloud” 24 years ago). Gisele Kapterion then spoke about the importance of trust and government and that if a system is modern, simple and reliable, citizens are more likely to trust it. She then shared some data on citizen willingness to share data with government (no surprises that the US citizens have a large number of people unwilling to share ANYTHING with government compared to other countries) and the levels of personalisation that government customers can incorporate in their systems. She left us with a thought “if ChatGPT can answer my question, why can’t you?” which really lifts the bar of expectations that agencies are expected to meet. Leah Koshiyama, Director of Product Management, then took us through general, and SalesForce specific aspects of Gen-AI risks and principles. Interestingly, among the obvious ones we’ve heard this week she also covered having high quality, representative, right-sized data models to ensure a smaller GPU footprint and associated carbon impact to run the model. We also learnt that there's a 65% gap between the number of organisations that understand the need for having Gen-AI principles and the number actually implementing them. We then had Paul, followed by Andrew Randall, diving into what SalesForce does for public sector and Defence respectively with a myriad of industry clouds and solutions built specifically for customer needs.

One more day to go tomorrow, and Adrian Motherway's promised us a session with Microsoft’s new recruit, Sam Altman, which should prove to be a blast 😜. Stay tuned.