- The Intelligence
- Posts
- The coming years will be TERRIFYING in AI
The coming years will be TERRIFYING in AI
According to former OpenAI employee we absolutely have no clue on what is coming
Kain Ditmer
June 17, 2024
The coming years will be TERRIFYING in AI ๐ณ
Hello everyone,
"By the early 2030s, the entirety of the US arsenal, like it or not the bedrock of global peace and security, will probably be obsolete." - Leopold Aschenbrenner
Yes, you read that right. According to a former OpenAI employee, the coming years in AI will be terrifying. He claims that AGI is just a mere steppingstone, and by the end of the decade, humans will have made their last invention. Ultra-intelligent machines will be so advanced that they won't need humans to improve anymore.
But before we enter a Matrix-like world, we've got some exciting updates for you. We've revamped a Plug-and-Play template, and we're here to give you a full rundown of last week's big Apple event. Plus, we'll catch you up on all the latest happenings in the AI world.
Buckle up and letโs dive in!
Table of Contents ๐
Are we not aware of wat is coming?
A former employee of OpenAI who recently got fired, Leopold Aschenbrenner has voiced his concerns on the situational awareness of society on developments of AI in the coming years. He did this with a 165 page paper, comparing this moment in AI with the years before the nuclear weapons became reality.
If weโre lucky, weโll be in an all-out race with the CCP (Chinese Communist Party); if weโre unlucky, an all-out war. Everyone is now talking about AI, but few have the faintest glimmer of what is about to hit them.
Here is everything you need to know:
1. Intelligence Explosion:
Ashenbrener predicts achieving AGI by 2027 and superintelligence by the end of the decade. He emphasizes the unprecedented acceleration in AI capabilities and the dramatic leap expected in the near future.
"There would then unquestionably be an intelligence explosion and the intelligence of man would be left far behind thus the first Ultra intelligent machine is the last invention that man need ever make."
2. Power Consumption:
He also highlights the exponential growth in compute power and the corresponding scramble for power contracts, emphasizing the need for massive increases in electricity production.
"All of these compute clusters being built out need electricity... We literally do not have enough electricity in the United States currently to power everything that we're going to need to power to reach AGI and superintelligence."
Meaning the electricity and energy to power AGI and superintelligence, might be found in other countries than the US, countries in the Middle East or China who have prioritized power generation.
3. AGI and Superintelligence:
The discussion underscores the rapid progress of AI, moving from GPT-2 in 2019, described as a "preschooler," to GPT-4 in 2023, likened to a "smart high schooler." The next step is expected to reach AGI, followed by superintelligence.
"By 2025-2026 these machines will outpace college graduates... By the end of the decade they will be smarter than you or I."
5. Security and Espionage:
Ashenbrener criticizes the lax security measures in AI labs, highlighting the risk of critical AI secrets being stolen by foreign actors and used to gain a competitive advantage.
"The nation's leading AI labs treat security as an afterthought... Securing the AGI secrets and weights against the state actor threat will be an immense effort and we're not on track."
6. AI Research Automation:
The potential for AI systems to automate AI research is discussed, suggesting a feedback loop where AI rapidly improves itself, leading to an intelligence explosion.
"If AI is as smart or smarter than an AI researcher and engineer, then all of a sudden we can deploy hundreds of them, thousands of them, millions of them all working 24 hours a day in parallel."
7. Economic and Military Impact:
Superintelligence will have a transformative impact on both the economy and military, predicting an industrial explosion and significant advancements in various fields.
"By the early 2030s, the entirety of the US arsenal, like it or not the bedrock of global peace and security, will probably be obsolete."
8. Government Involvement:
Ashenbrener advocates for government intervention in managing superintelligent AI development, drawing parallels to the Manhattan Project for nuclear weapons.
"The USG will wake from its slumber and by 27/28 will get some form of government AGI project... This is the most important challenge for the national security of the United States since the invention of the atomic bomb."
9. Future Predictions:
Ashenbrener predicts that by the end of the decade, the development of superintelligence will be in full swing, with significant challenges and risks needing to be managed to prevent rogue AI systems from causing harm.
"By 2030 we will have summoned superintelligence in all its power and might."
๐ TEMPLATE: Your personal AI email assistant ๐ฉโ๐ป
We've shared this template before, but now it's even better. We've improved its performance, allowing an AI to fully manage your mailbox. With this updated template, the AI can draft or send emails based on the information you provide.