A New Year - All about AI

(Only subject today)

That Freight Train You See Coming Full Speed is - the AI Express

Everywhere you turn today, everything you read or watch on TV, AI is there. Just plowed through Thinking With Machines: The Brave New World of AI by Professor Vasant Dhar, Ph.D., Professor at New York University's Stern School of Business and the Center for Data Science and one of world’s foremost AI pioneers. Filled with technical (at least to me which ain't saying much) explanations of how AI came about and has evolved, TWM also reflects on the future, which is the most interesting part to me.

As we all know by now, AI, even at this adolescent level, has this insane ability to access vast amounts of data, millions of times faster than a human does, and then draw conclusions. Identify facts, find answers, make predictions. This capability requires that AI models be "trained", which is jargon for being fed vast quantities of information and taught how to evaluate it and formulate responses using algorithms. Don't ask me to explain this technically but I'm guessing that is essentially what is going on in these vast data centers popping up all over the globe. The ones that require gargantuan amounts of energy we don't always have.

Here are several points that hit home.

What can AI do already and will do a great deal more of it in the generations to come?

Change how work is done. How fast it can be done and how thoroughly. It will reduce the influence of emotions and prejudices on the results. It will MOST DEFINTELY replace many forms of human effort.

What AI is unlikely to do?

Make humans obsolete. At least anytime soon. For several reasons.

1. AI is not infallible. It makes mistakes. As its "knowledge base" expands, it relies increasingly on "un" curated data that is wrong or "fake". Remember - GIGO? Then there is hallucinations, which is an AI term for the fact that AI often makes up things. Like the sources it cites for the answers it provides.

2. It is subject to linguistic nuances, for example, "no" is anathema in the Chinese language.

3. Lots of the content on the internet is false or misleading and yet it is "ground truth" for most of AI's predictive training.

4. In domains where the majority of "facts" are objective, verifiable, AI can already perform at high levels. Where factors are highly variable and context is dynamic, AI predictive models are less reliable. It's one thing to differentiate black from white, hot from cold and get it right 99% of the time. Sentencing a convicted criminal or curing an uncommon illness are not so black and white. Correctness demands context.

5. Then there is the question of who decides how AI will be trained and sets its guardrails? What if Northern California/Silicon Valley is calling the shots? Armies of supposedly objective, highly knowledgeable humans are employed as "Enforcers". Their job is to make sure AI gets it right...in their view. Who chooses the Enforcers? Uh - the guys creating AI - the Googles and Open AIs of the world choose the Enforcers. And...presumably their Chinese counterparts. Anybody thinking fox "watching" the henhouse?

6. What about AI in the wrong hands? Luigi Mangione found instructions online to use a 3D printer to make the gun he used to murder healthcare CEO, Brian Thompson in midtown Manhattan. Why not use AI to plan the murder? Maybe, one day, even do the killing itself?

7. And, of course, AI is a long way from possessing the skill to inject emotion, feelings and common sense on any sort of rigorous level.

8. Finally there is the matter of Trust. The higher the consequences of getting it wrong the more difficult it is to trust. Obviously an autonomous car that drives off a cliff or a cancer diagnosis that is "dead" wrong have consequences few can tolerate. Anybody remember Knight Trading? Biggest market maker in 2012. Specialist in high frequency trading until a piece of mangled code caused them to buy high and sell low. The resultant $460 million loss wiped them out. Ouch.

Who will succeed most in the age of AI?

Those who know the most about the subject and ask the best questions. Those who know the most will be less reliant on AI and less subject to false answers. Those who ask the best questions will able to leverage AI in the most effective manner. Getting to the most useful answers efficiently and error-free. This skill will go a long way to making one AI-proof.

In other words, those who possess the knowledge, curiosity, logic rigor to tell when AI is wrong or at least might be wrong. Particularly in domains where consequences of error are life and death.

Who will be most adversely affected by the increasing role of AI?

Those most reliant on AI to make their contributions. Common sense, eh? The more you rely on AI, the less valuable you are. Think "parrot".

What is Future of AI?

Undeniable, Irrefutable. Irrepressible.

One need only look at the vast commitment of resources and the top priority AI is being accorded by world powers - both private and government - to be certain AI is here to stay. Yes, there will probably be bubbles and they will burst - as they did for technology in the early 2000s, but that will be a mere blip. A chance to buy on the dip.

Some experts speculate AI will someday become CEOs of major corporations. Able to evaluate business propositions faster, less emotionally, more rigorously than human CEOs. Who's your boss?

So What Really Cool Things May AI Do?

Healthcare: Our healthcare system in general has critical limiting factors beyond its cost and unaffordability. Factors that will exacerbate challenges as our population continues to age. One is overworked healthcare professionals, many of whom not only lack the time to consult thoroughly and patiently, but also must rush through recording findings. This common shortcoming retards learning and future diagnoses. Add in the increasing dependency on specialization and it's difficult to aggregate all that's going on in one place. Not a problem if the one's dealing with a torn ligament or a broken wrist. But definitely a problem for highly complex illnesses. Imagine if AI could streamline data collection and assimilation so the attending knew "everything" going on, past and present, at one time?

And then access a treasure of trove of cases with exactly the same profile to see how they were treated and how things turned out? A database of what worked and didn't.

Channel Appropriate Content: Not based just on what you might "buy next", but on what you might benefit from based on your actual needs. Your profile? Most content sent your way today is for exploitation purposes. What if it was sent with only the objective of your benefit not your wallet.

DOGE on Steroids: No need to elaborate on this one.

The limits of our imagination seem to be the only boundaries of AI. Google reports that AI already writes 25% of its code. AI models are consuming vast quantities of data every minute. YouTube posts 16,000 hours of video every 30 minutes. That's the number of hours an average human is awake in four years. "Every 30 minutes" and it's all going into the AI memory banks where it can be accessed, integrated and formulated in nanoseconds. Whew!

One Final Caution

The famous philosopher Immanuel Kant theorized most people prefer to remain in a perpetual "state of immaturity", allowing others to think for them. They do so out of laziness, fear and preference for security over thinking critically and independently. Politicians certainly subscribe to this theory. See where AI fits here?

Benevolent Dictatorship

Long been a fan of this model for governing. Admittedly I have no idea how it would work or more importantly how "benevolent" ones get chosen. But think about it. If AI reduced the need for delegation especially on really important decisions and the levers of control were in a really good guy's (or lady's) hands wouldn't everybody be better off? Just a thought. No more elections, no more power-grabbing politicians.

Oops. That's Communist China without the Benevolence. Okay - maybe not my best idea.

Now, back to today's news.

Quality "Learing" Center

Bad enough a Learning Center can't spell Learning. but this one received $4 million in taxpayer subsidies and was empty. Guess which state? Guess who's governor? First name rhymes with dim.

Another beaut from the Gopher State. Law allows a registered voter to vouch for up to 8 others' residency so they can sign up and vote same day - without ID.

I still maintain the 2020 election was stolen.

But fear not. Ilhan Omar's husband has been doing just fine. While Ilhan's fellow Somalis have been ripping off taxpayers on Omar and Walz's watch, hubby has been rockin' and rollin'. His wealth management firm, Rose Lake Capital has grown 20 times in under a year. He's almost as shrewd as the Old Crone.

Epistemic Death Wish

Leftists/Liberals/Progressives/Democratic Party (i.e., synonyms) believe in the wrong things - pure and simple. And have for decades. They masquerade as compassionate do-gooders looking out for the disenfranchised, the unfortunate. When, in reality, underneath their cloak of many colors, is pure greed. An unquenchable thirst for power - pretty much at any cost. Why? No secret. To accumulate personal wealth. To own three mansions, to amass 400% stock market gains, to increase valueless companies they own 20 times in a few short years.

After all, who, in good conscience, can be against helping the poor, the sick, the homeless, the persecuted? I'll tell you who. The millions of taxpayers whose hard earned money is shoveled down a rathole of fraud and abuse.

But the left doesn't seem to care and certainly has no intention of shifting direction. Buying votes through taxpayer-funded handouts is the only play in their gameplan. Call it the "don't bite the hands that feeds" strategy. Moderates who suggest otherwise are chewed up and spit out by the Ilhan Omars, Zohran Mamdani and AOCs of the world. There are no plays to address America's problems in the Democrats gameplan. No plays to stem illegals overrunning the country, to ensure the energy necessary to live our lives productively, or the military so we are protected. Nope - no plays at all. Just straight handoffs - I mean handouts.

Hopefully, as we did in 2016 and 2024 (and probably in 2020) a majority of American voters will keep in mind what's under the cloak and pull the right lever. Load the (ballot) box to stop the leftwing ground game. We'll find out this November and it's gonna be dicey.