Attribution: Prompt by JPxG, model by Boris Dayma, upscaler by Xintao Wang, Liangbin Xie et al. (Apache License 2.0 or BSD), via Wikimedia Commons
By the Fabulous Dadbots: Dave S., Mark M., Mark O., Dennis C., Paul C., and Geoff Carter
Something appealing about that ‘60s imperative—Question Authority—in a proletariat way (to me)…
I have a boatload of things on my mind. First topic is Artificial Intelligence (AI) and ChatGPT. White hot topic. I heard about it 3 weeks ago and just about every day since. Whilst on vacation, cocksure son told me it will render Google search engines obsolete in 3 years: “you’ll never do a google search again”. NPR reinforced this tonight. If you haven’t heard about it there was a nice intro to it on NPR Marketplace. My takes:
- Like anything shiny and new, you’ll hear the whizz bang stuff.
- My shorthand: When you do a google search, you get a bunch of offshoot questions to refine the search in the direction you want. ChatGPT seems to leapfrog that—especially if one asks a more specific question. Several of the whizz bang examples you’ll see put a Shakespearean take on it. Why? Whizz bang…and all the Shakespeare plays are in the algorithm’s data base.
- A word caveat: It appears AI is off and running and there is no turning back. Ahem, however, I question whether or not “artificial” is the right descriptor. “Artificial” has a lot of connotations—most of them negative. I suppose “man made” is at the core or “unnatural”…”synthetic”…etc. … Also: “intelligence” may be generous.
- In the NPR story tonight, they admit these beta versions often fill in the gaps indiscriminately—no fact checks—and thus incorrectly…the nascent industry terms this: hallucinating.
I gave it a go….thought I’d try an obtuse question: Here’s our dialogue.
Subject: ChatGPT experiment
–Dave
Ha. Interesting line of questioning. ChatGPT seems to contradict itself about tighty whitey issues. First it says there’s no evidence of ball-huggin’ clothes affecting reproductive rates, then at the end it says that tight clothing that restricts airflow can lower sperm production. I would say a Google search engine gives ChatGPT a kick in its AI nutsack on this fact-finding mission.
Okay. I suppose the next logical question would be: “does a lower sperm count result in a lower ability to reproduce?”
Dennis C.
Or I’m wondering if a hot tub can function as a contraceptive device. With champagne, of course.
Geoff
(The High Road)
Seeing that Geoff and Dennis took the low road on tighty whitie sperm performance, I feel compelled to relate my own brush with AI. This anecdote operates on a much higher moral plane.
Remember a few years ago, when IBM was advertising nationally for their “Watson” AI system? Here’s a link to one commercial, and it’s actually fairly entertaining:
https://www.youtube.com/watch?v=8xYvwcnHn9k
Well, this piqued our interest over at Northwestern Mutual. Actually it would be more accurate to say that my director was looking for a way for our IT Investment Support department to look modern & nifty. So he charged me with figuring out a way to use IBM Watson AI.
Here’s what we came up with: analysis of Private Placement investment offers. Quick explainer: In addition to stocks, bonds, and Real Estate, the company also invests in so-called “private placements”. These are investment deals which are NOT structured in a standardized way, like a public stock or a publicly traded corporate bond. These are special deals, put together with all kinds of caveats, bizarre capital structures, and weird payoffs. Companies will go to sugar daddies like NM with a private deal precisely because they DON’T want the loan or stock IPO to be a public matter. Maybe the company wants to stay private. Or maybe they don’t want to adversely impact their stock price by issuing more stock or more public debt. What’s in it for NM is that, because these are special deals that take much more pre-deal due diligence, and which have higher debt service “operational” needs once the deal is struck, there is a premium in the return. They have to offer a higher yield–without being intrinsically riskier.
The Private offer starts with a big fat descriptive document from the prospective borrower that lands on the desk of some Private Placement investment analyst at NM. He or she has to read the offer quickly, and decide whether it’s even in our ballpark. We only invest in maybe 10% of the deals that are offered. So there is a major winnowing process.
Our idea was that we would have WATSON read the deal document, and it would predict whether the deal was something that we would be even remotely interested in. If so, then a human would get involved. This would save a shit ton of time on the very front end.
Well, we brought in the Watson sales team from IBM, and we explained the concept. The idea is that we would process a few hundred of these deals manually, and we would also have Watson try to decide. Watson would then look at what our humans decided, and “machine learning” would allow its artificial intelligence to take over the filtering process, eventually. This would have been a huge help, because the Private Placement dept is always constrained by lack of personpower. If they could do more deals, they would, but they just don’t have enough people. And in the years of low interest rates, which encompassed most of my 34 year career, the Investment Department was forever pushing and straining to get MORE YIELD on their investments (without taking on additional risk). This fit the mission perfectly.
The first problem was that we weren’t really serious. There was no way that we could have gotten, say, a half million dollars allocated out of the capital budget for this. We were really looking to do something on the cheap (though we never told IBM).
But the other big problem was that IBM listened to our pitch… and then tried to sell us on one of the solutions that they had already created for another company. A solution that only remotely resembled what we were looking for. They never even seemed to TRY to understand what we were trying to do.
So I came away with the opinion that this fabulous Watson AI was mostly “vaporware”. They had no ability and clearly no interest in addressing our issue.
Therefore, I must draw the inescapable conclusion: All AI is terrible! If it don’t work for me, then it don’t work.
Mark M.
Interesting AI story Mark. It brings to mind my working life experience in “business”, or more specifically sales and deal-making. As you allude, 90% of the time spent on pursuit of possible deals amounts to nothing. I suspect that is a rough approximation of all sales or deal-making activity in our economy. It certainly reflects my experience. I had to think that true business acumen consists of the ability, whether analytical or intuitive, to discern whether the party “across the table” from you is serious and has the funding to back it up. I could never make it in “business” for lack of this acumen and also for the innate revulsion at “wasting” 90% of my time. This is a very small-timer mentality. A true big-timer knows and accepts that they will waste 90% of their time (or the time of their underlings) in pursuit of the Big Payoff down the road. The Big Payoff that can move you into the fortunate 1% of society. This is a gambler mentality. Sometimes I wish I had a gambler mentality but that’s not the hand I was dealt (see what I did there?).
MarkO
“I wish I had a gambler mentality, but that’s not the hand I was dealt…” cue rimshot. Ha. I have to admit that the salesman/gamblin’ life never had much of an appeal to me either. Maybe I’m just too much of a Wisconsin Thrifty Bowler or maybe “Death of a Salesman” and “Glengarry Glen Ross” affected my psyche too much at a young age.
Which brings me back to AI and a question about MM’s interesting Watson story. Did IBM have any highly-commissioned go-getter sales people like “Glengarry Glen Ross'” Alec Baldwin or Al Pacino going to any lengths to hammer home a deal? I’m just wondering how that stuff works–what incentivises businesses to embrace new technology or reject it.
On the rejection side, I’m sure there are lots of working stiffs who feel threatened by AI moving in, even if AI is supposed to lighten their workload. Was that maybe an underlying issue at NML–spoken only in furtive whispers? I’m wondering, if I worked in the private placement department at Mother Mutual, would I think it best for my cronies and I to undermine Watson moving into the cubie next to ours?… John Henry “died with a hammer in his hand, lord, lord, yes he died with a hammer in his hand”.
Dennis C.
In the IT “space”, as the cool kids would refer to it, AI advances are not yet a threat to the IT working stiff. The real threat is the Indian IT developer, whether on site (on an H1-B visa) or offshore in India. They literally do the same work for half the cost. Quality & project communication suffer somewhat when offshore teams are used, but IT management doesn’t really care. Their problem is that ALL IT development is just ungodly expensive, and IT needs to find a way to cut costs to satisfy their corporate masters.
When it comes to new tools, almost everybody in IT welcomes anything that can relieve drudgery and increase productivity. It will be some time before actual AI starts whacking jobs in IT.
Those Private Placement guys? They are all winners in the meritocracy—high achieving players from privileged backgrounds—and they also want to focus on deals, not drudgery. That’s how they are compensated.
I would say the biggest threat right now to US workers is not artificial intelligence, but “consumer” intelligence. Everywhere you look, work has been pushed downstream to the consumer. We check out our own groceries and book our own flights, for two examples. Those used to be jobs. I was at McDonalds today and I picked out my own value meal on a large computer kiosk!
As far as high pressure sales. When you are selling a complicated product that requires sales engineers to demo, hard sell doesn’t work. You need to show that your shit actually works. But I did encounter one instance where the tech team successfully sold us, but then the business negotiations broke down cuz the vendor was too hard nosed. Our biz guys told them to take a hike cuz their definition of a system upgrade basically required an open ended financial commitment on our side. So the ABC (Always Be Closing) guys blew it on their side!
–Mark M.
Nice work bots…MM’s Edison tale was fascinating. (my how IBM got passed by ay?….yet another example of substance trumping hype) I also liked “whitey tighty” (DC) and the gambler…..not (MO) lines.
A minor eureka for naïve Dave from Madison was what other think tanks have known for decades—and perhaps Orwell, Huxley….et.al. too…. Namely: Robotic mechanization umbrella includes AI, not just spot welding on automotive assembly lines—the initial fear of the “robots are coming”. 40 years ago, (yeesh btw), I remember a little point counterpoint on a stairwell bulletin board in the ME building in Madison. Someone had taken aim at a Robotics presentation—ala “the lefty elimination of jobs angle, only to have a rebuttal, stapled on…”of routine, dehumanizing tasks”. Can’t remember what I thought at the time…probably was knee jerk anti-robot…but now, with the wisdom of hindsight, I now believe that jobs morph, don’t get too romantic about spot welding. The horses out of work after cars came along is the classic “change is inevitable—deal with it” rebuttal to “oh no, no one is knitting mittens any more….”
About the time of the Edison hype I recall someone saying much of the lawyer trade being eliminated. Not the few trial lawyers, rather those heavy hour brief searching jr. lawyers. Similar distill it for me angle as NML’s. Has not arrived yet, but all this hype about chatGPT means the tsunami is building—IMO.
For now I’m fine with the bots doing the grunt work…just as long as those MMMs (marauding Mequon mom’s), don’t learn Python.
Dave S.
https://en.wikipedia.org/wiki/Python_(programming_language)
Oh yeah. That reminds me; I’m thinking my new stage name as an old white man hip-hop artist and major influencer is gonna be “Y.T. Ty T. “. (Still workin’ on getting the initials just right)
Err. No Wait…. “Ty T.Y.T”
–Dennis C.
I did get a good chuckle out of Denny’s one-liner about “Google giving ChatGPT a kick in its AI nutsack”. Fits the generational stereotype to tee…also evoked that John Henry folklore in my head before I saw someone make the reference.
What you’ll probably start to hear more of is a term called Artificial General Intelligence, or AGI. The big push towards AGI ties back to shortcomings that exist with current AI solutions–similar to the frustrations Mark mentioned from working with IBM Watson at NW. IBM Watson was developed to be more focused on narrow use cases, which only works well when the criteria and parameters relevant to decision-making are well understood and codified into algorithms. But all of the work that developers put into developing that ML architecture are inherently limited to the narrow use case they are solving and also don’t incorporate the probability weighting we intuitively assign to decisions in the real world, as your friends also mentioned.
As opposed to narrow AI, which says “let’s develop a machine learning model that can automate this specific process”, AGI says “let’s develop a machine learning model so vast, it can replicate the cognition of the human brain to automate all processes”. While the latter certainly seems a lot sexier than the former (and hence why it’s getting a lot more attention and funding from VCs), I think we are still a ways out. My hunch is that developers & biz people will need to start putting their heads together and get their hands dirty learning how to develop automation for their own specific use cases, as opposed to throwing their hands in the air and praying that AGI will come to solve everything before they fall too far behind their competitors who are putting the work in to automate and become more efficient.
–Son of Dave
(The Low Road)
Interesting topic, Dave. I remember when I started teaching and computers—the Apple I and II mostly—were first being installed in language labs and classrooms, and the rank and file teachers were freaking out. “They’re going to replace us!! Run!!” Of course, that was hyperbolic hysteria and laptops are now an indispensable classroom tool. But it does beg the question of whether–eventually–AI devices could replace teachers. Could classroom management be handled by a droid—anyone short of the Terminator (maybe in today’s classrooms he might be the solution)? Could AI handle the subtleties of student engagement and motivation?
Anyway, since I’m on the low road, Mark, I thought I’d bring up the topic of AI sex dolls. I didn’t (really, I didn’t) look into this topic until this thread emerged and the advancements that are being made by some of these companies that design these “intelligent” sex partners is just plain spooky. The company Luxury Humanoids design robots that not only provide sex services but are developing 3D technology to clone loved ones who have passed away, or—I suppose—celebrity models. Here’s more from a Psychology Today article:
https://www.psychologytoday.com/us/blog/sexual-self/202209/can-todays-sex-robots-offer-relationship
This stuff is the fodder for tons of sci-fi novels and films like Blade Runner; Millenium Man: Star Trek: Next Generation: I, Robot; and Wall-ee. They seem to ask where human sentience and emotional awareness begin. And these films are just the tip of the iceberg. Despite the assertion that AI will simply make life easier for the rest of us, the general trajectory of some of this research seems to be moving toward doing a Frankenstein—creating human life where none existed before. If this is true, and at this point, it’s only my speculation, it begs the question of why? Why do we want to make a Mini-Me?
Geoff
Kris is going to wonder about all those large boxes being delivered from Amazon.
Exactly what IS “artificial intelligence”? An internet definition says “
- the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.
But I think it misses the key component: The ability to LEARN.
And this is the doorstep to all the scary sci-fi. They learn & learn and then become our overlords.
Oh, speaking of which, you MUST listen to Flight
of the Conchords “The Humans are Dead”. You can find it on YouTube.
–Mark M.
Will do. And if you haven’t seen Lars and the Real Girl, it sort of addresses the topic of the human need for response—even from the inanimate. It also has an interesting take on human projection on inanimate objects–and It’s really a sweet little movie.
Geoff
Sexual reproduction is such a risky and unpredictable method of continuing the species, surely there must be a better way. And apparently there is. We can use technology to satisfy our sexual and intimacy needs and leave reproduction to the genetic scientists. Why does this smell a lot like the entirely respectable social and medical science of just 100 years ago, eugenics? Sexual reproduction is required for the maintenance of healthy gene pool in the long run, but it’s a lousy method of reproduction for individuals in the short term. Some futurists believe that a simple lack of birth rate will cause the end of the human race, as opposed to nuclear war or killer pathogens. My money is on genetically engineered viruses. Wait, did I just place a bet on the cause of the end of the human race? How am I gonna collect on that?
MarkO
It’s the hand you were dealt, Mark….
Geoff
Quick interject on the film references:
- First the big Amazon boxes Geoff will have to intercept delivery of: Ex Machina. Oscar Isaac—one of my faves, plays the cool intellect to a tee. Utterly bored with his “above all others-ness”.
- Back to lower sperm counts: Children of Men (I’ve mentioned this one before…the Michael Caine smoking dope scene and repeating his joke, ala Gene Hackman in Bonnie and Clyde), but the story is pretty good—I’m a sucker for dystopia flicks. Also the leads on the cast, Clive Owen and Julianne Moore, are (2) more I respect.
-D.