Gen AI and Superhuman?

Thanks to John Sviokla for highlighting the Superhuman article about AI and AGI from Ethan Mollick (who has a brilliant book , must-read, Co-Intelligence)

A quick comment on AGI and then some points on the AI and superhuman article.

AGI—Artificial General Intelligence is really not when AI gets better at all tasks than humans. It is more like when AI is more intelligent and has volition, what we call 'will' or what looks like will to us. That is also the scary version. We already have movies about that.

It won't happen for a long, long time, if ever. Two big reasons:

• The human brain is crazy complex. Just modeling 85 billion neurons firing multiple times a second inside an n-dimensional set of networks, including 'compute' from glial cells and an always-on default network running at a symbolic and sub-symbolic level, is beyond our math, let alone our ability to intelligently model it.

• Most information isn't the right or wrong kind. Doubly so for the most pivotal info in business. Most information isn't what is the capital of France, who won the World Series in 1942, or which is four squared. It isn't information where there is one right answer. The majority of information is in a world where the info lives on a scale from super stupid to super genius, terrible to excellent, bumbling to brilliant. The mission of Gen AI, collaborative AI, and all AI is really to move us to the right on all those continuums that are a huge part of our endeavors and existence.

I wrote a paper published by the Sandhill Group in 2016, saying the future of the Internet is huge progress/leverage from humans living in an aggregated and augmented world. This is exactly what AI is really doing.

Being able to engage humans across groups and at scale, augmented by AI, in smart ways to quickly leverage cumulative IQ will become key. An early clever winner in that space is CrowdSmart.ai

Five thoughts on AI and Superhuman:

1. Your Formerly Smart Friends

I have four PhD friends who I call for domain-specific input. They are above-average smart and have read extensively. I have a fifth ‘friend’ who has read and absorbed 100,000 times more than all four of them put together. The fifth is GenAI. So, of course, I use the fifth friend. It is a widely different thing than a human brain but super functional because of the scale of information it has processed, not because it's a better 'brain'. It is a tool (like the brain), but they are very different tools.

2. Damned if Do, Worse If You Don't

As this progresses, it will eliminate a huge amount of what we have thought of as professional work, certainly within the next four years.

Don’t think of this as beating 95% of people on tests; think of it as being better than the average worker at a very broad amount of knowledge work. Think of it as 10x cheaper and 20x faster. Also, the overhead of GenAI versus that of humans is different.

Many people say it won't eliminate work. It will just make it faster, smarter, and better. That is nonsense.

If it makes my coders, entry-level lawyers/paralegals, copywriters, analysts, call center agents, insurance adjusters, researchers, etc., 40% more productive, and I have 100 of them, then maybe I really only need 60.

(Go find John Sviokla's HBR piece on WINS work for a brilliant exposition of the kind of work).

The economic and sociological impacts of this technology have not been fully considered. Efficiency gains of this scale positively impact FCF, margins, revenue, and ipso facto shareholders but negatively impact net labor demand and macroeconomics.

Suddenly, In 1912, Henry Ford's cars required 20 times less labor than their competitors. In 1912, there were 400 car manufacturers in the US. Twenty years later, there were 40, with 3 having 80% of the market.

Not adopting this won't help.

In global competition, you have to keep up. Throttling tech adoption would have much worse impacts. The UK, France, and certainly China are a lot more organized and focused on innovation ecosystem enablement, leveraging all their assets and scaling at a national level than we are here in the US. I won't count on the invisible hand of long-dead, different-era Adam Smith to solve this one optimally.

3. Automation Bad

Automation (and tech progress) has always had bad effects mixed in with the progress. Nail guns, for instance. Robot weapon systems have already killed people they weren't supposed to. Usually, that's a human leaving them turned on, and some unfortunate troops march into its kill box down range. On occasion, guns misfired due to bad engineering and software done by humans. Asbestos, lead paint, DDT, X-ray development, and chlorofluorocarbons were also examples.

4. We Are Not Advanced

It is easy to look back and say, wow look how advanced we are, especially if you remember the early 90s.

If a superior alien intelligence operations research/anthropologist landed in America and looked at our systems for education, transportation, healthcare, city management, food supply, and energy, their only conclusion is some not smart people spent decades screwing this up. They are all rat nests of inefficiency, lack of design thinking, under-tech leveraging, and costly blobs with diluted, under-optimized performance and outcomes. The map of connections between consumption and supply across those domains is mindblowing. These things have evolved in profit-centric, compartmentalized, silo-ed, over-legislated domains and have huge optimization and impact improvement upsides.

That is good news, is the current innovation wave in AI (GenAI, Federated AI, Hyperscale/Hyperspeed AI, Agentic AI) can harvest all of those profit pools and drive real change for the first time. Change is hard and requires funding. The inefficiencies in those markets are the funding fuels. I expect smart cities, healthcare, and education to take the lead on that.

Foundational, pre-built integration platforms layered with smart AI, Gen AI, and construction space for quick mini app-like solutions, all working in real-time from data and events, will be game changers,e.g. VANTIQ.

A platform to answer the questions you will think up once you see your real world in real-time AND to easily implement the solutions, controls, and processes to excel.

"The future is already here ... it is just not evenly distributed".

William Gibson.

He also said, " I don't know what I am looking for yet, but I know I have to find it."