AI Predictions…

Penned APR ‘23

Predictions

I’m a big fan of Scott Galloway & Tyler Cowen for many reasons. Making predictions is up there among the many benefits they’ve bestowed upon my development. It's a way to clarify your thought process, put skin in the game, and guard against the self-serving tendencies of hindsight bias.

So here's my "bet" on AI so that I can compare how wildly wrong I likely will be, but I'll overall be better for the thought process:

• I contend that AI will usurp the buzzword data in 3-5 years such that we will have "AI Analysts." Happened much, much faster!
• These nerds, of which I hope to be one, will know how to deftly use AI to shorten the flash to bang of quotidian tasks--e.g., building PowerPoint decks, generating draft remarks, etc.--and vitally, they will know how to converse with the technology. Chief amongst the skills that will determine an AI Analysts' efficacy = question asking. Again, mostly wrong. Rather than outsourcing, AI has become this/we’re all empowered to do it.
• This shift from data to AI (not mutually exclusive) will reprioritize skills such that premiums on emotional intelligence will prioritize over hard skills like coding (but the latter will certainly not vanish). Somewhat right (for now)
• In an unembellished sense, an AI Analyst will have an intimate relationship with their AI tech of choice. My genetic optimism biases me to believe that will engender greater empathy and social connection. Still, the opposite could be equally plausible: That is, the AI replaces their need for human connection, and/or the best AI analysts become experts in exploitation, refining their PSYOP ability with each AI conversation/query. Too early to tell
• More broadly, I am still reading to determine what, if any, insights I can offer between the difference between how bias shows up in data vs. AI. I am confident it will, but how? 
• Daniel Kahneman often spoke about the superficial/misleading ways in which data scientists presented data sets as "objective." The colloquial "garbage in, garbage out," is somehow universal but always other-directed (everyone is an above-average driver).
• This is still one of my greatest fears/an aspect with which we don’t wrestle enough. That is, the ‘bed of leaves’ issue where we don’t know the ground truth from what it’s training. At least tech like Glean has popularized sourcing/citing.
• The primary limitation I currently see with text-generated AI is its inability to yield novel insights. If you know the DIKW pyramid, AI currently feels stuck at between the I & K, taking information and applying it to your question but unable (for all kinds of reasons) to specify beyond the general. Some of that can be overcome with user back-n-forth. But more consequentially, there seems the reliance issue of AI scrapping all of the digitally available information and being able to answer your prompt but not go one step further to make connections that you possibly didn't ask for but for which people who have K & W are paid a premium. This is an idea I have ineffable feelings on that I'll try to make more coherent in the future.
• Somewhat true. And maybe those most succuseful will learn how to hack this.
• Aesthetically, every time I read an essay, it feels like AI was taught by a strict grammarian who judges on form rather than content--e.g., endless stiff prepositional transitions. DFW would not be a fan. But this will likely evolve quickly (asymmetrically).

Roundup of AI content in ‘23 (How Quaint)

Alright, here's the fun for this week:

How many pay-for-fee services will AI then render obsolete? Or is the economy of AI that the niche/high-class producers pay for specialization? If so, that's basic economics; nothing new to see here

*I cut out two prompts dealing w/ former safeguards by AI on financial and political advice.

As an extra, if you struggle to write workouts, have AI provide you with a first draft (or seven).

• This is still fantastic advice.