InvestAI etc. - 10/15/2024
InvestAI etc. - 10/15/2024
AI for investors, explained in simple terms. An open thread updated weekly.
Topics discussed this week:
Reality is breaking down
Human population ↘️, robot population ↗️
Visions of paradise vs AI-optimized compounding
The public LIST updated
AI links to consider
Reality is breaking down
>
RICHARD >
You might have seen that realistic (but fake) photo of a distressed child holding a puppy in the aftermath of Hurricane Helene. It’s obvious why it went viral — it touches all the heart strings — and you can understand the inevitability of a future where such fakes overwhelm traditional news photography. We are clearly the last generation who took photographic evidence seriously and this will have far-reaching consequences.
Lars Daniel, a security expert, writes in Forbes about the many ways that this can undermine society, especially through desensitization. As we become overwhelmed by these shocking images, many of us will simply stop believing what we see, even when a real image demands that we pay attention.
A similar example is Ethos_atx, the Instagram account for “Austin’s Number One Restaurant.” If the photos of their food seem out of this world, it is because they are fake and AI-generated. There is also the problem that this “restaurant” doesn’t even exist. It’s just an Instagram account.
Ethos now has 70K+ followers on Instagram, a level of popularity that is bound to attract copycats. When compelling images are this easy to create and spread, what will become of the more mundane photos we take in real life? Would you rather keep an album of poorly-framed, blurry family photos from last Christmas, or some AI-optimized shots of happy kids acting out your parental fantasies of the perfect family?
>
SAMI >
It’s too bad that Ethos doesn’t exist. The dinosaur croissant and the feetzza look like winners.
I agree that the inability to trust photos is going to be a problem, but even before AI, we were increasingly looking at photos with a skeptical eye and wondering if a real photo had been staged or photoshopped. For example, the photo of the child with the puppy is indeed gripping and too perfect in its ability to stir emotions. So much so that I would wonder, if it was a real photo, whether the photographer staged it or at least photoshopped it.
>
RICHARD >
True but AI enables a whole other dimension of “fake”, and we are not just talking about fake images from news events. You already don’t know if that seemingly well-crafted email from a colleague is composed by an LLM. Soon you might not even be able to tell whether it is really them on a video call.
Zoom demonstrated a new feature that lets you generate a video “clone” of yourself that you can manipulate with text commands. I guess if you’re not feeling your best some day, you can stay in bed and let your clone do all the talking.
>
SAMI >
I am trying to think of offsetting factors and of possible benefits. One offsetting factor is that there will likely be new companies, perhaps also run by AI, that offer authentication services. For example, in the case of the child with the puppy, they could determine via geolocation or other technology that no such scene was possible at the location and time claimed by the “photographer.” And with cloned avatars on zoom, companies that care about having the real person on the call may require fingerprints or other forms of authentication.
From a broader perspective, one possible benefit of all this is that we will gravitate back to sources that we can trust. I remember reading that in the early days of radio in the 1920s, it was the Wild West when it came to reliability of the news. But in the 1930s new regulation and a sense of crisis (with the Great Depression) made radio more reliable and made people turn to trustworthy sources. We may see the same here. News organizations that genuinely put trust ahead of everything else are going to be big winners.
If we take this thinking one step further, news organizations that are blatantly partisan will lose their audience because they will be overtaken by sources that are even more fake. It is difficult to compete with “totally fake” when you are merely in the manipulation business.
>
RICHARD >
AI can do subtle manipulation too. In a new study from MIT, researchers got 200 people to watch a crime video and answer questions from either a standard questionnaire or an interactive genAI chatbot. The chatbot successfully imprinted them with false memories.
The chatbot was designed to deliberately mislead. It would ask a leading question and then give positive reinforcement to an incorrect answer. You might think this isn’t a fair test—the chatbot is trying to mislead. But what if the chatbot does this accidentally?
>
SAMI >
In real life, for example in a courtroom that includes prosecution and defense, a judge can interrupt questioning that is deliberately misleading, and there is an opportunity for cross-examination. Future AI systems will need to incorporate something similar, some kind of “critical thinking module” to help mitigate these issues.
>
Human population ↘️, robot population ↗️
>
SAMI >
In the same week that Tesla showcased its new two-seater robotaxi and an improved Optimus robot, Foreign Affairs published an article by the demographer Nicholas Eberstadt, The Age of Depopulation. It includes this passage:
Just how depopulating societies will cope with this broad retreat of the family is by no means obvious. Perhaps others could step in to assume roles traditionally undertaken by blood relatives. But appeals to duty and sacrifice for those who are not kin may lack the strength of calls from within a family. Governments may try to fill the breach, but sad experience with a century and a half of social policy suggests that the state is a horrendously expensive substitute for the family—and not a very good one. Technological advances—robotics, artificial intelligence, human-like cyber-caregivers and cyber-“friends”—may eventually make some currently unfathomable contribution. But for now, that prospect belongs in the realm of science fiction, and even there, dystopia is far more likely than anything verging on utopia.
Eberstadt’s central thesis is that the global population is about to peak and to start declining, a phenomenon that is already happening in Japan, South Korea and other places. The question for our purposes on this page is the one posed by Eberstadt in the above excerpt: to what extent will robots replace family and friends. Instead of marrying and having children, will many people just have one or several Optimus robots in their households?
>
RICHARD >
I confess that I didn’t pay much attention to the Optimus announcement, or to the claims about humanoid robots in general. The concept feels too predictable: you have a bunch of tasks currently requiring a human, so make a human-like robot that can do those tasks instead.
That seems to go against the way technology usually brings its biggest efficiency gains: by eliminating the activity itself. The classic example is an automatic shirt-ironing machine. Is it better to have a machine that can iron your shirts, or to redesign fabrics so that shirts don’t need ironing in the first place? Or maybe even a step beyond that, to have casual offices and workplaces where pristinely-ironed shirts are no longer required?
Your bigger question is a philosophical one about whether people will replace real human companionship with machines, and I’m afraid the iPhone generation has already decided that the answer is yes.
>
SAMI >
Families are already experiencing this shift and many have banned phone usage at the dinner table, a small act of resistance against the tech onslaught.
About Optimus, there is also no reason why a robot should look like a human, with a head, two arms, two legs, a torso etc. The human body was optimized for a completely different type of existence, through millions of years of evolution. Instead of one Optimus that will cost tens of thousands of dollars, we are more likely to end up with a number of less expensive robots that look nothing like a human and with each designed for one or a few functions. We are already seeing this with the robot lawnmower and robot vacuum cleaner.
I also don’t see a robot as an effective replacement for a companion but perhaps future generations will feel differently.