In a world of Large ACTION Models, what does the future of computing hold?
So last night I was reading a bedtime story to my daughter - we come up with a prompt together and set the task for Google Gemini. We had some great success with "Tell me a bedtime story of Warrior Princess Alysha. She meets an old lady on her way to the market. Alysha has empathy for the old lady and was rewarded with three magical beans. Alysha has an adventure in the clouds after climbing the resulting magical beanstalk."
But one time we had a typo and Gemini was prompted, "Tell me the bedtime story of Process Alysha who meets a dashing prince from another kingdom".
Our protagonist quickly gained analytical and engineering skills and, through applying her STEM skills brought progress and joy to her loyal subjects through rural electrification projects, high-efficiency railway infrastructure investments and improvements to regional governance models.
It got me thinking - in this world of increasingly sophisticated Foundational Models (FMs) - you probably know them as "Large Language Models" like Chat-GPT 4 (Turbo!) and Google's Gemini (and Anthropic's Claude!). What will the future hold for us consultants? What would an LLM-embedded in our daily operating systems look like?
Well for starters, the paradigm of mouse and keyboard will slowly evolve away to an augmented M-A-K UI : Mouse-and-Agent-and-Keyboard. After watching us interact with computers for long enough, a LAM (Large Action Model) embedded in the OS's UI would build a library of Actions (not just Languge). What do I mean by that?
"Get the latest news and summarise the headlines for me." It'll learn to open your favourite news sites, ingest the data and aggregate it.
"Book train tickets for my weekend trip" and it'll open the price comparison sites, your diary, get the right routes and book everything.
领英推荐
"Suggest ideas for meal prep next week, summarise suggested ingredients that meet my nutritional needs, and prepare a shopping list with my preferred suppliers" and it'll run the workflow.
Because to the outside world - my interactions with the Internet are just that : clicks, forms, emails, etc. If that were all intermediated by a Foundation Model (FM) what would the future look like?
In a world of FMs, LLMs and LAMs - what's a professional to do? Well, now more than ever qualities like empathy, creativity, judgement, and critical thinking will become even more important. Where FM's can handle the gruntwork of doing the heavy lifting, it now falls to us to direct the FM's, and even architect workflows around FM's (organising FM's to coordinate the work of other FM's!).
Just my musings on a Sunday evening - feel free to comment on speculate further in the comments section below!
Program & Project Leader | Strategic Planning, Data Analytics, Digital & AI Transformation | I Help Businesses Boost ROI with Data-Driven Strategies
10 个月Izam Ryan, great stuff. It's really thought-provoking. I'm adding my two cents here: for LAMs to be subtle and widely adopted, ironically, they need to be disruptive at the same time. That can be achieved only if these agents gather a lot of contextual information about you—all of your interactions with digital surfaces and beyond. Will you give them permission to do so to become the AI-augmented man of tomorrow? Embedded artificial intelligence means changing natural intelligence a lot :)