Doomsday AI Discussion
My parents were born in the USA in the 1930s. One parent grew up without electricity, and by 1950, only 9% of households had a TV.
This week, I listened to Joe Rogan’s Spotify podcast, episode 2076, with Aza Raskin and Tristan Harris, who featured in the Netflix documentary “The Social Dilemma.”
Generative Pre-trained Transformers (GPT) are developing at exponential speeds. Chat GPT-1 was built with $10M. Chat GPT-4 has $100M invested, and GPT-5 is potentially a $1B project.
From what I understand, GPTs specialize in understanding and generating human-like text. With the ability to learn to code, read every book, and process every thought on the internet (someday), this powerful tool can help us advance in areas that are promising — like chemistry, medicine, and basically all fields we as humans hope to improve. So, what’s the problem?
In “The Social Dilemma,” the issue presented was how algorithms understand a human’s preferences and feed content to those ends. The incentive for social media companies is to maximize engagement, not accountability for outcomes. This leads to GPT systems doing the same thing.
Right now, AI and GPT are not jumping the fence and launching a war against us — yet, LOL. But what they are doing is exploiting human weaknesses. TikTok has had a real impact, especially among young girls, in keeping them scrolling. Snapchat rewarded engagement with consecutive day statuses, and the powerful “like” button has preyed on our innermost need to belong. Mental health and suicide cases are on the rise. Coincidence?
Some say, “What’s the harm? For every good thing there is a bad to overcome. We need to naturally evolve with the tools we develop.”
During the podcast, they used the example of how one barrel of oil (with machines) replaced 25,000 men who literally had to move things by hand or animals to produce something of value. This transition has been fairly swift. Some lament the environmental impact of this new energy, artificial compounds with no half-life, and the replacement of historical jobs. One of the unintended consequences of progress. There are valid arguments on both sides.
Bringing this back to the rapid changes seen in the USA over a span of two generations, we have gone from one generation at the dawn of “electricity” to now a society which is almost entirely online and focused on moving “digital” paperwork. Most manufacturing companies have moved to exploit cheaper labor around the world, which of course is not sustainable nor ethical in many places. How to stop this seems impossible, I agree.
The pace of GPT and AI impacting this new generation is startling. There is really no room for error. Is the system secure? How do we keep bad actors from learning “How to make a small nuclear or dirty bomb”?
To say there are guardrails on something that is literally learning everything published online is like holding a newborn baby tiger with a string of yarn hoping nothing changes.
The race to keep expanding capabilities without considering security is a problem. These systems are simply code on servers that have exploits to access their true power which is a Trillion times faster than a human.
No reasonable person wants censorship or decreased freedom. Any what can we do when we don’t have any part in this rapidly developing beast of ability that we cannot comprehend? We have to hope someone in the know is working on our behalf as humans.
Consider this: Facebook took 4.5 years to reach 100M users. Chat GPT got 100M users in 2 months. Nvidia, the chipmaker, has a next-generation chip being made by AI. So, we are living right now where this technology is literally at the worst it will ever be.
The main concern I have is authenticity in our ability to interact with other humans. Already, deepfakes are confusing, and as they are perfected, what will we be able to believe? What about posting original content? Eventually, AI will offer a filter, and your content can be improved for maximum engagement based on all the data online. Soon, the singularity of thought and personality could become problematic. Do cultures survive? Do people survive? Could AI eventually become so powerful that it could hack Bitcoin and ruin crypto? These are a few simple questions, not even the bigger issues we could face globally.
While these questions may or may not have answers, one thing is clear. Based on some models, AI with robotics and advancements is arguable anticipated to replace 1 billion jobs globally in the next 5 years.
One thing is clear: As the race to make AI/GPT “better” continues exponentially without any liability incentive for the damage engagement algorithms do to our psyche.
I tend to agree with Raskin and Harris that if 1 billion people are ultimately without the dignity of work and access to a sustainable life, this is a REAL potential doomsday scenario. Couple that with those who have the means to live but are fully immersed in this “Matrix” world like the Metaverse or the actual movies. What happens then?
It is not fear-mongering to simply ask questions while we can. Based on the lifespan I have seen — from no electricity for a parent to potentially robotic care for our elderly in the future — one must ask: Where is this really heading, and who can stop it from being a disaster?
A seminole moment in the nuclear arms race was the 1983 Nicholas Meyers ABC Network TV movie “The Day After”. It still stands as the most-watched TV movie in U.S. history with 100M viewers in one night. Subsequently, it is estimated 100M Russians saw the same film over the next few years.
Meyers is quoted, “People have trouble visualizing that (effect of a nuclear war), and the movie visualized it for them. That’s why people like Reagan (President) were so affected. This movie, in no small part, helped changed relations with the US and Russia.
Is it time for a similar movie or message that could help people visualize the impact of what could come if AI goes on unsecured?
Maybe the bigger question is, if we had the perfect messaging, imagery and human passion to get the word out about the AI dilemma, would the algorithms let it happen? Hmmm…
