Now that we have AI...now what?

On April 18th, MIT Media Lab became #AI ground zero for about 1,000 academics, Fortune 100 execs, VCs, and entrepreneurs across compelling panel sessions, "lightning talks" and intense debate/discussions across any free space you could find in the 6 story building.  The list of speakers and presenters are here: https://www.imaginationinaction.co/mit-04-2024.  

It was a continuation of last year's Imagination in Action AI Summit - where I opined "Third time's the charm - AI is finally here" https://www.linkedin.com/pulse/third-times-charm-ai-finally-here-musings-from-summit-douglas-kim/ 

In the year that has passed, companies have gone from exploration to full on building and applying #GenerativeAI at massive scale. Some key takeaways for me:

1. The dystopian predictions by skeptics that "millions of workers will be displaced" of a year ago seem distant.  It's clear from the summit most use cases center around people augmentation, not replacement. 

2. Ted Shelton proclaimed: "Whatever speed enterprises think they’re moving with AI, they’re probably not moving fast enough" - and he is right.  The difference in AI maturity across enterprises who are in the same industry is widening - with laggards falling further behind.  

3. With tools and capabilities becoming easier to adopt and apply - prioritization has become a problem. There are 200 possible use cases in a bank - where does one start that ensures the highest ROI, while preserving compliance and regulatory requirements?

4. The off-the-shelf LLMs are commoditized, and I predict will essentially become part of the 'stack' like the operating system and the database once were.  LLMs are so commonplace now, that you can immediately tell when someone has been using it to write their work or released a generic chatbot - they all sound the same, and write the same, hallucinate the same. The magic is in the first party proprietary data that sits on top of it.  You can't expect ChatGPT, or Anthropic - no matter how many parameters it's trained on, to know your industry's specific issues, your company's specific policies and procedures, your company's specific regulatory and legal requirements, your customers' specific preferences and profile - and so the magic will be as always - harnessing in a compliant fashion, the proprietary data that makes your business, and your customers unique.  

5. There will be lots of opportunities for lots of great companies who can provide tooling to harness this data, and companies who have proprietary data to apply it to very specific problem areas - that enterprises can leverage

6. I also agree with Yann LeCun - Chief AI Scientist at Meta on his prediction that AI monopolies held by large companies will give way to open source - just as we saw what happened to the Internet tools.  And to put the proverbial money where his mouth is, Meta has opensourced LLaMa. 

Thank you John Werner - for conceiving and orchestrating another incredible summit - you are a national treasure. 

And special mention and gratitude to the following enterprise leaders who lent their expertise as speakers and panelists - Alex Barretto of Dell, Christopher Paquette of Allstate, Jeffrey Saviano of EY, Kiran Vuppu of Wells Fargo, Marcin Detyniecki of AXA, Michael Wise of Universal Pictures, Ra'ad Siraj of MassMutual, Ted Shelton of Bain, Vibhor Rastogi of Citi Ventures and Vipin Gopal of Walgreens Boots Alliance.