Phase Shifts in AI and Life
There’s phases. And then there’s phase shifts.
Last week was week 3 of life post-MAS (Links to week 1 and week 2). Way fewer conversations than the prior 2 weeks. I kept some time to prepare for two classes I’m teaching in week 4.
The familiar. A former head of data at a sovereign wealth fund, now in social work and teaching. My fellow panellists (one whom I kept crossing paths with, see here) and the audience at an AIMA panel at Morgan Stanley. An old friend who has been at the apex of global corporations, now deploying capital at a PE firm. Another old friend who went independent two years ago to help organizations build kinder workplace cultures. A former boss, now sitting on multiple boards. An old friend from my creative industries policy days, now CFO at a tech startup. A news editor at a global professional platform who I met a while back.
New connections. The head of strategy at a non-profit researching AI safety. The head of data of an AI company that a decades-old friend kindly invited me to meet. The APAC head of a global trading body from whom I learned about existing disclosures on algorithms in trading message protocols. A book publisher who started life as a teacher. Someone in the bread and butter of insurance getting into AI. Someone who is both doing wealth management but also in deep tech.
The takeaways went somewhere different. It started with a question I didn’t expect. About my pessimism when it comes to AI.
Which led me to think about phase shifts. Something much more fundamental than pivots.
Phase Shifts in AI
“You’re surprisingly pessimistic about AI, given your background.”
An old friend who has run way bigger operations than any I have ever worked in made this remark as I shared my views on AI. My views were grounded in some of the conversations I had (see the messy middle of AI, my post on week 2).
But I’m actually not pessimistic. So I tried to explain how surprising shifts happen quite frequently in AI.
Most folks talk about AI winters. But I’m actually more optimistic. I’ve seen too many phase shifts in AI not to be.
Neural networks existed for decades. Condemned as unworkable. Then backpropagation, compute, and data aligned, and suddenly they worked. CNNs existed for years. Then AlexNet won ImageNet in 2012. And computer vision became a solved problem. Sequence models existed, different versions of RNNs. But we got nowhere. Then transformers arrived in 2017, and suddenly language scaled to limits we never imagined. GPT existed as a useful toy for demos. Then ChatGPT launched in late 2022, and suddenly AI was accessible to everyone. Diffusion models existed. Then Midjourney and Stable Diffusion, and suddenly anyone could generate images.
We’ve been seeing this over and over again. And I’ve been asking people I meet if such shifts could also happen in other areas of AI - time series, networks, tables. For example, I notice more and more research on foundational models for time series, networks, and even specific domains like fraud detection. And I wonder if such data could be the solution as we reach the limits of text and image data. Phase shifts in these areas could totally change the way we operate in finance and business.
The pattern: the technology often exists for years. Then something tips. And suddenly it works in a way it didn’t before. Not incremental improvement. A change of state. A phase shift (I know I am taking some liberties with this phrase, but I feel it fits).
Phase Shifts in Life
That explanation to a friend made me realise that I saw phase shifts of a different kind this week. Not in technology. In people.
A former colleague’s daughter journeyed from banking to art. We talked about her journey for a while. Not a simple pivot. A transformation. It resonated with my own journey in art.
An old friend I knew since national service who ran massive operations now evaluates companies from the other side of the table. He has journeyed from public service to consultancy to retail and food MNCs and probably covered much of the globe in this journey. He shared what drove his decisions at each shift.
Another old friend and colleague has done consulting, social work, and left two years ago to run his own practice. He shared how he rationalized it.
Another has done bars, supermarkets and startups. Probably went through more than I could understand from just chatting with him.
A former head of data retired early, now teaching and doing social work. I told him that I could not do what he does - helping the sick and lost. I did not have the emotional fortitude.
The head of strategy at a non-profit researching AI safety has done social enterprises at a scale I could not imagine and in places I have never even traveled to.
The learned experience of the teacher turned book publisher was eye-opening. I could tell that publishing was a purpose, not a job or enterprise for him.
As I listened to them, I realized - some of these were not my lanes. Sometimes seeing someone else’s phase shift clarifies what yours is not.
For some of them, I don’t think these were simple pivots. Pivots are changes of direction. You can pivot back. These are changes of state.
And they have made me wonder. Am I in a pivot, or shifting to a new phase?
Three weeks out of MAS. I’m not too different. AI risk management. Finance. AI research. Art. But I feel slightly different. Less bounded.
Have you ever felt this shift?
#PhaseShifts #AIRiskManagement #Conversations #Reflections #Pivots