With our ability to both predict and influence our own futures in an ever-more-complicated world, it can be difficult to disentangle cause from effect or stay above hype and buzzwords. When ChatGPT and other systems that rely on LLMs (Large Language Models) occasionally render insightful comments about the future, it can make things even more confusing – especially since such systems can only extrapolate from existing data, which is often flawed and biased.
Jason Schenker helps people and businesses tease out these distinctions as a futurist, economist, and best-selling author. He is the president of Prestige Economics and Chair of The Futurist Institute. He is currently embarking on a In July 2024, he led a talk with Rabbi Neil F. Blumofe at Congregation Agudas Achim in Austin, TX as part of a series of post-Tikkun Leil Shavuot classes titled “Jewish Ethics, AI and the Future” – called, simply enough, “AI and the Future.”
(This post is part of Sinai and Synapses’ project Scientists in Synagogues, a grass-roots program to offer Jews opportunities to explore the most interesting and pressing questions surrounding Judaism and science. This post is adapted from Mr. Schenker’s talk at Congregation Agudas Achim on July 23, 2024.)
Photo by Alexander Popovkin
Questions for Futurists
Jason Schenker: When I was doing data analysis around 2016 or 2017, I started getting the same question all the time: “What about automation? What about jobs and robots? How do we predict that?”
Now imagine you’re trying to predict what will happen 10 or 20 years into the future. The confidence interval around what might be a reasonable for a forecast of that range gets infinitely wide. The confidence is blown out, and so are the potential kinds of future realities. There are many fractal dynamics that can come into play.
There are also different kinds of futurists. There are some folks who are futurists in a very “everything will be perfect in the future” kind of way. These are the ones that fall into the “robotopia” column – “everything’s going to be a robotic utopia. Everything is perfect.” There are also the “robotal depths.” If you’ve seen movie Terminator, that’s kind of how some folks think.
Chances are that the actual future will be somewhere in between these two extremes. It has a lot to do with our own agency. This is where we get into a concept of preferred futures. By thinking about what the future outcomes are, we have agency in crafting what the preferred future might be.
One of the ways people often do this is look at what’s called the “four futures framework.” Let’s think about one example using geopolitics. In the tensions between the US and China, if we look at that, where would the four futures be?
One would be if things get stuck where they are right now. So imagine nothing gets better, and nothing gets worse, and nothing changes at all in the relationship between the US and China. “What we have right now is what we will always” – that’s one possibility; that they don’t accelerate, but they will continue on the path on which they have been.
But we’ve seen tensions ratcheting up between the US and China. There are two more catastrophic or drastic possibilities. One would be a parabolic acceleration. So in other words, everything that’s been tense, tense, tense – and then, suddenly, a giant conflagration. Things accelerate rapidly.
The other possibility would be a massive reversal. The conflict goes away, and it’s Putin and Xi and Zelenskyy sitting at a campfire making s’mores.
So those are the four ways to think about it now. Are all four of those scenarios plausible? If you had a long enough time, all these things could happen. Which scenario is more likely right now? The answer to that question could always change. But we also need to identify the levers that can differentiate between the four scenarios.
If we look at our future with AI itself, what about job replacement? Do we think that what we’re seeing now, with some jobs going away, will be the same? Could this accelerate? Could it continue? Could it accelerate more rapidly? Or could it reverse? Any of those four things could happen.
It’s important to think to both recognize and manipulate the levers that could get you to those four scenarios. And then it’s important to think about policymakers and regular people – what could we do? What agency would we have to affect that outcome?
How will AI affect the future?
I use AI all the time. Is it good at doing creative stuff? No. One of the three big risks of AI is that it keeps reading and regurgitating all the same stuff, and if you’re firing a bunch of reporters – which many companies have been doing for a while – and you’re just having AI run the stories, it can make everything really, really dull (at best). It’s not going to make anything that’s new and creative.
I have a friend works in the movie business. I’m a producer for one of these films right now, and he came over to my house about a year ago. He didn’t understand what SAG was doing the strike without the AI. I said “Watch this.” I told ChatGPT to write me a movie about vampire and werewolf love in Texas, and it came up with a love story title and everything. And then I told it to write dialogue. The dialogue wasn’t great, but there’s a lot of vampire-werewolf garbage out there.
Do I think it’s going to write an Academy Award-winning movie? No. Could it write the next series in the Vampire Diaries, which is already pretty horrible and is in its fourth iteration of regurgitation? Absolutely. It can regurgitate like nothing else, which is why it’s really great for letters of proposal, press releases, all kinds of corporate drivel.
It is perfect. For creative stuff, not so much. So this puts the onus of creativity on people.
Neil Blumofe: That’s great. The question I’m going to ask you to keep in mind, Jason, is: What’s an important pattern that you have discovered at the edge of society?
This is from the Bible, from Genesis 49:1. It says, “And Jacob called his sons” – before he was about to die – “and said, ‘Come together, that I may tell you what is to befall you in days to come.’” So the idea that we could say that, in the moments before death, we gain a sense, not a futurism, but a sense of prophecy, is a very important thing. This episode Jacob called Israel, telling each tribe what’s going to happen, is particularly important in the mystical world. So I’d like to think of Jacob as the “first futurist” in that way.
So the idea of this perhaps being understood only later is like any sort of fortune that might be told. If it’s big enough, you can find a way for it to be true. I’m a hypochondriac – so when my wife was in nursing school, we had to hide all of the medical books because I thought I had everything, But the idea is that if, you know, it says, you look at your horoscope or anything else, I can say, “that is going to happen,” or “that did happen.”
So I want to just think about: How can what Jacob would have told his kids actually have sunk in? In our analogy, how could it be that whatever you tell the companies, or you tell the industry, or what AI tells you, is actually helpful for the work you’re going to do, as opposed to reacting to the prompt by saying “close enough”?
The imperfect art of prediction
Jason Schenker: As a futurist, you’re looking at trends of the past and data of the present, and you’re using that to craft a vision of the future – so there is a certain kind of systems thinking involved. You’re connecting the dots of what you know. If you’re looking at a statistical average, you’re taking a bunch of data points and you’re drawing a sort of line down the middle. The up-and-to-the-right regression trend that you might see in a financial market is because over time, the data points generally go in one direction – but this is all just from drawing lines between points.
Anyone who is looking at the future is likely to do the same. And this is why being a futurist is really important in thinking about those different scenarios, because there is a tendency to say, “Well, this is the trend,” and finance people tend to believe “the trend is your friend, and the trend is just going to continue.”
Very often, the trend just doesn’t continue. Things suddenly happen either a lot faster or a lot slower, or they just stop, because suddenly everything gets crazy. If everybody thought tomorrow that a certain stock was going to go to the moon in three months, would it actually take three months for that stock to go there? No, it’ll go tomorrow, because all of us have heard “in three months.” So there’s an arbitrage of information.
Neil Blumofe: So maybe what would happen, if you’re one who can connect the dots of the facts. Tell me what you think about this text, in terms of reading it as a futurist quote. This is a fairly well-known section from the Talmud from Menachot. You have Rabbi Yehuda telling the story that Rav told (which in itself has a very interesting theme of attribution). But he says it like this:
“When Moses or Moshe ascended to the heavens, to where the Divine is, Moses found the Holy One, the Blessed One, sitting and attaching crowns [to the letters of the Torah].”
Moses looks at it and says, “I don’t understand this spreadsheet. I don’t understand this budget.” And he says, “Master of the universe, what are you doing? I don’t understand. It doesn’t make sense.”
And God says, “Moshe, Moses, there is a person in the future who will exist after many generations. And his name is Akiva, the son of Joseph, who will in the future expound upon every crown piles of piles of interpretations.”
So even though I may not be able to read your ledger, there might be somebody, or something, or some program like AI in the future that can do it very well. So, in your own future, then, are you actually setting the stage for someone else to read the tea leaves of your future projection?
Jason Schenker: Before, I said there are three really big risks in AI. The first one is that everything becomes really boring. Number two, which is much more insidious, is that biases become codified. Garbage in, garbage out. If you think a robocop is going to have better social justice than an actual police officer, I have bad news for you. They will look at historical data, and they will behave in a certain way that is tied to that data. Unless you’re intentionally filtering for poor actions in the past, you’re just going to do what’s always been done. You will get the most racist robot police you could ever imagine, or worse.
So, the codification of bias is a problem. A futurist looked at the past 100 years of New York Times articles and created a mathematical function where an LLM would use this information to cluster together words. It came up with equations like “man + plane = pilot, “woman + plane = stewardess,” “man + medicine = doctor,” “woman + medicine = nurse.” If you’re looking at articles going that far back and weigh them all equally, this is what you’ll get. So someone has to go intentionally to break that pattern, because the society we live in today is different than the one 60, 70, or 80 years ago.
And there’s a third risk, which is the proliferation of errors. AI creates a reliance on technology. Technology is great when it works, but the problem is that when I have ChatGPT do something for me – like write a press release, or a legal draft, or a letter of proposal – I’ve done those things for 20 years, so I know what they should look like. But if I hire somebody brand-new and tell them, “Use this to write press releases,” and they’ve never written one, they won’t know if it’s good or not. Now, people are going to rely on AI to do all kinds of tasks for them, but they have to be able to do them by hand first.
Education and preparation
Anybody here do long division on the regular? No? That’s because we’re probably using Excel instead. But you know 1 million divided by 5 isn’t 3.89 – you can spot a glaring error. That’s important. This means for education, they’re giving people papers to write at home, and they’re just having ChatGPT write for them – all-too-efficiently so.
What they should do is have students keep doing their homework out there, but have their actual grade be based on coming back to class and writing it by hand in the blue books, with no computer, no tech. We can expect to use tech in the future, but you have to be able to do the things you ask of it by hand. It’s just just like how in a math class, you can use the calculator at home now – but when you come into the classroom, it’s all pen and paper.
That’s how we’re going to have to educate and train people – at the collegiate level, high school level, and in the workplace. If you’re hiring people, before you let them loose, you’ve got to see if they can actually write.
Neil Blumofe: I think that proposal is an interesting example of saying that whatever you want to do, one day you’ll have to deliver on your own brain. This recalls Pirkei Avot 4:16 – “Rabbi Jacob says: This world is like a hallway to the future world. Prepare yourself in the hallway that you may enter into the banquet hall.”
So, just like in today’s technocratic society, “the more you learn, the more you earn” – everyone should be like Rabbi Akiva, right? Learning those things, how to put it together, interpreting the crowns on top of the letters and all the latest technology.
But let’s find out what happens to Akiva. This is BT Menachot 29b. Moses gets to go into Akiva’s classroom, and he says, “Master of the universe,” this person, Akiva, exists and you convey the Torah through me?”
In other words, “I’m not worthy. This Akiva is this incredibly well-constituted being who knows these things.”
And God must have had a bad day, because God says: “Shut up. Such is my will. You don’t understand.” (God doesn’t say, “Be quiet, my son.” That’s not our tradition.)
So Moses then said, “Okay, great. So since you’ve shown me Akiva’s great learning and teaching and adjusting, now show me what his reward is. He’s obviously on a fleet of yachts, right?”
So God tells Moses to turn around, and upon doing so, Moses sees Akiva’s flesh being weighed out on the market. An ignominious end, right? And Moses asks the masquerading Hebrews, “Can this be the Torah? And this be the reward for such a high-functioning reaching?” And what does God say? “Shut up. What do you know?”
So I would turn to text #6 BT Sandhedrin 90a: “Every person has a share in the World to Come.”
Jason Schenker: I think is important that parents talk about this, because there’s a lot technology can do, but there’s a lot it can’t do. It can’t write sales scripts. And I’ve had it do this. “Give you 10 reasons why people should buy this. Write me 10 objections people might have. Write responses to the objections. Write this in an informal way.” Anybody worked in sales, you’ve done this. But you probably had to either have somebody train you or you probably made a bunch of mistakes. So if you can use the technology to avoid mistakes, fantastic. But you know what the ChatGPT won’t do, is it won’t sit down and have a warm conversation with another person. It won’t build rapport. There is no replacement for human relations.
So, the robot is not coming for you, rabbi. I know there are online rabbis, AI shabbatons, people having theological discussions with ChatGPT. But this is not applicable to you. And the reason for this is because AI hallucinates. It does what I was talking about with the data, where it just draws lines between points.
Once I tried asking it who I am, since I have a decent public profile online – it might happen to know who I am. And it told me I have degrees I don’t have from universities I’ve never been to, that I’ve worked in places that I’ve never worked, and that I’ve written a book that not only I haven’t written, but no one has. So it just connects the dots and makes stuff up.
Because things move so quickly and things can be manipulated, how do you know what’s real? Trust and relationships with people are more important than ever. We live in a physical world. We don’t live in the metaverse, despite what Mark Zuckerberg might have you think. We live in a world of physical stuff moving around. We eat physical food, we do physical things. We are carbon-based, not silicon-based, life forms. This means a lot of what we do has to happen with other beings, other people. Relationships and trust, sitting down with people and kibbitzing a little – there’s not really a great tech substitute for those things.
0 Comments