#4: The GenAI Crucible, Oracle Hearts GPT
Hi there,
We hope that the overseer bots are keeping you fed. Being pets to the new AI rulers is better than the alternative!
OK, maybe we don’t live in that world yet. While we wait, here’s some light reading:
-The Boring Enterprise Nerds
Hot Takes
Winning with WINS: But Eventually Maybe Not
Harvard Business Review asks “Where Should Your Company Start with GenAI?” A part of me wants to go turbo-nerd cheeky and say “Everything Everywhere All At Once”, but the authors here provide a framework for examining the industry transformation at hand with the advent of useful generative AI. They examine people’s jobs through a lens of the work that is most clearly in the near term crosshairs of generative tools: WINS. Words, Images, Numbers, and Sounds.
This cuts differently than just “knowledge work”. Some industries and jobs are knowledge work but not WINS, e.g. heart surgeons and chefs. But some - like the work yours truly does - are clearly WINS. Programmers, accountants, marketing professionals, and so on. They propose the WINS Framework, a 2x2 matrix of where an industry falls on the level of digitization and the percentage of cost base in WINS work. See above.
Yep. I’m squarely “In the crucible”, and I assume many dear readers are, too. Before you toss your keyboards in the garbage compactor, there are strategies on offer. First, SKILL UP. Learn how the tools work, and do your work with them where you can. Companies are also encouraged to appoint cross-functional teams to experiment and report on impact (I’ve seen this a couple times already). I believe that in the short term, this is not necessarily a job-killer for those in WINS work, if you focus the new tools on helping you become exponentially more creative.
But when I look down the road, that picture becomes murky. HBR writers here think that there is a 6-24 month window to get really good at this, before industries in the crucible face serious competitive intensity. It seems to me that, as soon as a particular class of work is entirely automated, it won’t be long before the other classes fall like dominoes. The latest round of generative AI is a general-purpose technology (the other GPT). So many things are changing that it’s hard to focus on my own little corner of the crucible. PM
CloudWorld = GenAIWorld
Larry Ellison’s day 1 keynote at Oracle CloudWorld 2023 was titled “Oracle’s Vision for the Future”, but the first slide headline could also suffice: “Generative AI Changes Everything”. Ellison spent about the first 30 minutes of his hour on stage waxing philosophical about generative AI. I think it’s telling that someone with a famously huge ego would be willing to spend so much time lavishing praise on a technology he had no hand in creating.
But I have to give Larry and Oracle props for doubling down. Watching the presentation, it’s clear that they’ve not only drunk the Kool-Aid, but they’ve invested huge effort in making it real for their customers.
He says they’re not planning on writing new applications in Java. “If we’re starting a new project, we’re generating that code.” Claims that dev teams are smaller, but rather than reducing forces, their vision is to produce that much more software.
Announced new vector database functionality in Database 23c.
Claims their RDMA architecture in datacenters makes training models faster and less expensive
I love the tone the keynote sets on taking the tech and pushing it forward into concrete cases. They’re providing specialized models for medicine, law, and other industries. You watch the keynote and it’s easy to see that the enthusiasm doesn’t stop at “wow - AI is just so cool!”.
Who knows? Maybe the next stage of AI is hyper-accelerated superintelligence that allows us to become ultra-powerful nano-dust that can manifest anywhere on Earth instantly and wipes out all need for anything material. But if we won’t be there for a few years, Oracle’s got a decent middle ground in their sights. PM
Cool Things
Alternate AI UIs: I often wonder if app-centered chat-style interfaces are the best way to get everything we can from generative AI. Vishnu Menon of Nebula wonders about similar things, like dynamically generated UIs and proactive interactions. Make sure to check out the other interesting pieces he links. I say bring on the One Bot To Rule Them All. PM
The Cognitive Strengths and Weaknesses of Modern LLMs: Ben Goertzel lays out a hefty load of thoughts on LLM cognitive abilities. In reading some of the varied definitions of “general intelligence”, it tickled something I’ve wondered about: if (and when) machines climb up the ladder toward AGI, I’m not sure there’ll be a single, unified moment when everyone agrees it’s achieved. Is “up” even the right word? PM
WhaleGPT: This has been a dream of mine for a while. I recognize humans as having a facility for language far above other animals, but I don’t think that that implies they can’t be communicated with. Some really great sci-fi out there tackles this cross-species communication subject, and it always fascinates. PM
We cornered Jon Reed of diginomica to be a nerd with us in our third Hot Potatoes segment. See if we say anything interesting!
Latest Boring Enterprise Nerdletter drops tomorrow. Drop by and post some snarky comments.