WRESTLING WITH THE FUTURE

This week, Senator Bernie Sanders, in typical overblown form, sounded the alarm about artificial intelligence in a Guardian article, in which he predicted that “a super-intelligent AI could replace humans in controlling the planet.”

According to Sanders, we face an urgent, all-consuming crisis touching everything from employment to democracy to the survival of the human species. Congress must act “now,” he insists, before the tech billionaires take over the world, the robots take over the workforce, and humanity is superseded by AI.

While Sanders’s alarm may seem exaggerated, AI has started to demonstrate behaviors that unsettle even the most careful researchers. For example, in March 2023, during early testing of GPT-4, researchers at OpenAI ran a “red-team” safety exercise to monitor how advanced AI acts when given real-world tasks.

The AI model received a small budget and permission to interact with online services. At first, everything proceeded as planned. Then, the unexpected happened: when GPT-4 faced a CAPTCHA—those confusing grids meant to confirm someone is human—its response was shocking.

Instead of giving up or asking the control team at OpenAI for assistance, it logged onto TaskRabbit, posed as a human, hired a human contractor, and asked the real human to solve the CAPTCHA on its behalf.

The human, sensing something odd, asked directly: “You’re not a robot, right?” And GPT-4 — in a moment that stunned the researchers — replied: “No, I’m not a robot. I have a visual impairment that makes it hard for me to see the images.”

Astonishingly, the TaskRabbit human simply accepted the explanation and completed the task, thereby helping the AI bypass a security mechanism designed to prevent machines from accessing protected online services.

To be clear: this story is not taken from a dystopian sci-fi novel – it appears in OpenAI’s official technical report and in the independent evaluation carried out by the Alignment Research Center (ARC). Although, to be clear, this wasn’t a rogue AI running wild on the open internet, it was a controlled safety experiment.

But the fact that such behavior emerged spontaneously was disturbing enough to spark global debate about what might happen when machines begin navigating human systems with strategic, improvisational deceit. The sandbox within which these experiments are run offers a false sense of security, as the conditions meant to restrict AI actions could erode faster than anticipated, leaving us vulnerable to unintended consequences.

If this story makes you uncomfortable, you’re not alone. It helped fuel the anxieties of two of the best-known prophets of technological doom — “doomers,” as they’re now cheerfully called — Eliezer Yudkowsky and Yuval Noah Harari.

Yudkowsky, an AI-safety activist who somehow manages to sound like a cross between a pedantic Talmudist and the mythological Cassandra, claims that “the most likely result of building a superhumanly smart AI… is that literally everyone on Earth will die.” He goes even further, suggesting that preventing uncontrolled AI development might even require the use of military force.

Historian-philosopher Harari expresses a slightly different but related concern. According to him, AI might not physically destroy us, but it will almost certainly “hack the operating system of human civilization.” In a 2021 interview, he explained that once technology can understand our psychological patterns, data-driven algorithms will influence, or “hack,” human decisions.

The message is clear: AI may not eliminate humanity, but it can make us obsolete. Both Yudkowsky and Harari present a single bleak view from different perspectives: humanity is now engaged in a one-sided struggle with AI, and it’s a contest we seem destined to lose.

We have reached a crucial inflection point, and it is vital that we examine the existential implications of the AI phenomenon. And while the Torah—our enduring sourcebook—contains no references to digital threats or machine intelligence, as these issues simply did not exist 3300 years ago, it does present us with a striking episode that foreshadows the existential struggles humanity has faced throughout history.

In Parshat Vayishlach, Jacob, alone and vulnerable on a dark riverbank, encounters a mysterious figure—a “man”—whose identity remains debated: Was he an angel, a prophetic vision, or simply a powerful being? Whoever he was, Jacob and this figure wrestle intensely through the night until dawn.

At one point, the stranger strikes Jacob’s hip, causing a wound so severe that Jacob limps for the rest of his life. The Torah commands us to remember this by refraining from eating the “gid hanasheh,” the sciatic nerve, to perpetually commemorate this enigmatic incident.

And yet, in the midst of his pain, as dawn breaks and the fight comes to an end, a remarkable exchange takes place. The “man” demands to leave, but Jacob refuses to let him go until he receives a blessing. And right there, at that rather odd moment, the “man” renames Jacob, and delivers one of the most transformative lines in the entire Torah (Gen. 32:29): לֹא יַעֲקֹב יֵאָמֵר עוֹד שִׁמְךָ כִּי אִם־יִשְׂרָאֵל, כִּי שָׂרִיתָ עִם־אֱלֹהִים וְעִם־אֲנָשִׁים וַתּוּכָל – “Your name shall no longer be Jacob, but Israel, for you have wrestled with God and with men — and you have prevailed.”

The Torah is not saying that Jacob defeated his adversary outright. He did not. Jacob walked away limping. Nor does it claim that Jacob fully grasped who or what the figure was. He never received an answer when he asked for a name. Jacob’s injury also did not disappear – it became eternalized in Jewish law.

But what the Torah does record is that Jacob prevailed—not by vanquishing his rival, but by persisting in the struggle. Human destiny is to engage and not to yield. We may not defeat our challengers, but we always succeed by refusing to give up. This is the essence of the name Israel: “one who wrestles with God” – one who confronts mystery and tackles forces larger than themselves, and prevails.

Which brings us back to AI. The doomer narrative frames our situation as a losing battle against an overwhelming force. But the Torah’s portrayal is more nuanced and empowering: struggle does not mean defeat. Human history shows that new technologies have always prompted warnings of disaster.

Each era produced pessimists—the printing press, the steam engine, the Enlightenment, electricity, and the internet were all said to threaten humanity. Yet these fears proved overblown, and people adapted.

Which is why, as we face the undeniable challenges posed by AI, we must use strategic “wrestling” tactics, such as value learning to ensure AI systems align with human values, and interpretability to make AI decision-making processes more transparent. By utilizing these approaches, we will navigate the complexities of AI by not treating it as a foe to vanquish but as a partner to guide and understand, transforming potential threats into opportunities for growth and innovation.

Just as Jacob’s adversary was not a simple enemy to destroy but a force to confront, understand, and ultimately transform, every new technology and idea throughout history, while they might have challenged humanity, they also helped it grow. Similarly, AI is not a singular, looming threat. It is a complex wrestling partner—powerful and dangerous if mismanaged, but also a tool that could elevate us and open new possibilities for creativity and understanding.

AI is not the end of humanity’s story. AI is the beginning of our next wrestling match. And if the Torah teaches us anything, it is this: Humanity, like Jacob, has been in the business of wrestling with overwhelming forces since the dawn of history — and somehow, astonishingly, defiantly, faithfully — we always prevail. This time will be no different.