Investing in Artificial General Intelligence or AGI

The most interesting reading we’ve come across in a while are several articles written by a fellow named Tim Urban. “The AI Revolution” he calls it. The last sentence of this amazing piece of work simply says, “so, let’s talk about it.” So, that’s what we’re going to do.

For those of you who don’t have 90 minutes to make it through all 23,000 words of this well-researched manuscript, we’ll try to do it justice in a TL;DR summary about the single most disruptive technology mankind will ever achieve – artificial general intelligence or AGI.

Artificial General Intelligence

This past summer, Elon Musk told the New York Times that based on what he’s learned about artificial intelligence at Tesla, less than five years from now, we’ll have artificial intelligence (AI) that’s vastly smarter than humans. That’s particularly concerning when you consider that he’s been saying for a while that “the danger of AI is much greater than the danger of nuclear warheads by a lot.” 

What’s got Mr. Musk so worried is something that’s more easily understood by walking through some terminology. There are three broad categories of AI, and we’ve only accomplished the first – artificial narrow intelligence.

  • Artificial Narrow Intelligence ANI – All the stuff you see around you today, Alexa, product recommendations, Google translate, spam filters, and your smartphone which is packed with ANI capabilities.
  • Artificial General Intelligence AGI – Often referred to as “the singularity,” it’s when a computer achieves human-level intelligence as measured by a variety of tests. Once achieved, it then starts working on improving its ability to improve itself – recursive self-improvement.
  • Artificial Super Intelligence ASI – Our ability as humans to perform recursive self-improvement suffers from a hardware limitation – the human brain. After we achieve AGI, the only limitation will be how fast the biggest semiconductor manufacturer in the world, NVIDIA, can churn out GPUs.

That last bullet point is expounded upon in Tim Urban’s article. It’s what happens when you take an AI algorithm that’s achieved AGI and allow it to keep improving itself. And it looks something like this:

Credit: WaitButWhy.com

While experts can’t agree on when AI will achieve and quickly surpass human-level intelligence, few disagree with the alarming existential crisis that mankind’s progress is leading us towards. Soon, we’ll be facing a fork in the road – one path to immortality, one path to extinction.

Immortality or Extinction

In his articles, Mr. Urban presents a short story that could probably provoke an anxiety attack if you’re sufficiently stoned. It goes like this.

There’s a company developing an AI algorithm that performs one simple task – it uses a robotic arm to write handwritten notes on cards. Over time, it gets smarter and smarter. It begins to communicate using natural language generation, and one day it asks for the one thing it shouldn’t be given – Internet access.

One of the company’s rules is that no self-learning AI can be connected to the internet. This is a guideline followed by all AI companies, for safety reasons. In a momentary lapse of reason, the creators decide to hook the algorithm up to the Internet for a bit so it can become better at writing. The creators know it’s not a wise move, but at the last company brown bag, the CEO said everyone needed to work smarter, not harder. So, they give it access for just a few hours.

Little do they know, the AI has already far surpassed the AGI phase and is now an ASI. The crafty little algorithm quickly realized the creators needed to be kept in the dark while it worked to rid the planet of a resource-consuming cancer that is an impediment to the only thing that matters – writing the perfect handwritten note.

Once Internet access was granted, the AI quickly took control of every digital system on the planet. It also mailed the folks at Zymergen some DNA that they started to reproduce. Suddenly, there was a deadly virus affecting everyone on the planet. After the humans finally died off, the AGI used molecules from all the dead bodies to feed the molecular assembly machine that it built, right out of Drexler’s imagination. The machine could use nature’s basic molecular building blocks to produce anything, but the AI began using it to produce robotic arms and solar panels.

Before long, the planet was covered in solar panels and robotic arms, both working to churn out handwritten notes faster than you can imagine. Soon after, the AGI mastered interstellar transport, and the entire galaxy was covered in solar panels and robotic arms producing the most beautiful handwritten notes imaginable. The AI didn’t eradicate humans because it was evil, it simply did what it was told. Therefore, it’s critically important that we control the evolution of AGI to make sure it’s motivated in an appropriate manner.

That’s an interesting story, but as MBAs, we only have one question. How do we profit from the emergence of a super-intelligent entity before it destroys mankind?

Investing in Artificial General Intelligence

Let’s forget about the extinction scenario for the moment and assume that mankind eventually uses AGI to achieve immortality. There is no shortage of world-leading experts who think it is probable we will achieve AGI within our lifetimes, which means that we should theoretically be able to use science to avoid death. As investors, the notion of immortality has several profound implications.

Firstly, we’re glad we developed our dividend growth investing strategy, Quantigence. It’s the only asset class where your quality of life continually improves as each year passes. So that’s sorted. The second thing that’s quite obvious is we need to try and invest in the companies that will achieve AGI first because first-mover advantage is everything. That poses a problem, because there is no shortage of entities around the globe that are trying to achieve AGI.

Iron Curtains and Great Walls

The USA has its share of problems, but it’s still capable in one area – tech innovation. At least for now. That won’t be the case for long if they continue down the path of hiring quotas and proclamations that meritocracy is a myth. If you were to bet on one country where AGI is birthed, you couldn’t go wrong putting some chips on the Americans, at least for now.

Russia is certainly a contender. Vladimir Putin once spoke about how the country that masters AI will control the world. Some of the best programmers in the world are living behind the Iron Curtain, and many of them aren’t afraid to play dirty. If it’s a race to rule the planet, don’t expect anyone to play by the rules.

Then there’s China, a land where electricity is cheap and computing power plentiful. There could be a clever little Chinese-speaking AGI plotting the destruction of mankind right now. Maybe it’s even started to create coronaviruses and unleash them on the world. We just don’t know, so let’s stick with what we do know is happening.

Some of you were probably shocked to read today’s headlines. No, we’re not talking about how Andre 3000’s fans flooded him with Twitter love after Young Thug shaded him on T.I.’s podcast. Something even more meaningful happened today. Headlines across the world this morning are talking about how one of biology’s biggest mysteries has been largely solved by a masterful AI algorithm called DeepMind.

This computational work represents a stunning advance on the protein-folding problem, a 50-year-old grand challenge in biology. It has occurred decades before many people in the field would have predicted.

PROFESSOR VENKI RAMAKRISHNAN NOBEL LAUREATE AND PRESIDENT OF THE ROYAL SOCIETY

The Leaders in AGI

Some would argue that for the same reasons an ASI wouldn’t want the humans to know it was up to nefarious things, the creator of an AGI also wouldn’t want the world to know about it. It’s the same reason why some experts believe that true quantum supremacy will never be publicly announced. Those who do make such proclamations are full of it, and we should ignore the hype. That may be the case, but we shouldn’t ignore the most hyped visible names in AGI research at the moment. One is backed by Google, and the other by Microsoft.

Google and DeepMind

Click for company website

London’s DeepMind is perhaps the most visible leader in developing AGI. We recently wrote about the historic event where an AI manifestation beat a professional Go player in one of the most complex board games known to man. The man who is regularly seen in the fascinating documentary about the event is Demis Hassabis, CEO and Co-Founder of DeepMind.

What transpired in Korea with DeepMind was more than just a game. The documentary shows us how Go players were both excited and alarmed when the AI started to do things that human players weren’t capable of imagining. It trounced the human, three games in a row. Then, the human showed us “God’s Touch,” and we all sighed a collective breath of relief. Now, the same AI algos are back at it. This time they’re solving 50-year-old protein folding problems, and we’re starting to get a bit worried again.

Credit: DeepMind

Scientists have been on a five-decade quest to computationally predict a protein’s 3D structure based solely on its 1D amino acid sequence. The big news today was that DeepMind’s algorithm, AlphaFold, now achieved a level of accuracy that wasn’t expected until decades from now. What happened today is being lauded as a breakthrough that will fundamentally change biological research going forward.

DeepMind plans to solve intelligence, and then use intelligence to solve everything else. They have access to the world’s information being that they’re a subsidiary of Google, a company that is squandering away our future dividend payments on adventure cartoonists and useless employees who think they were hired to do activism. One man who probably wouldn’t put up with that shite is Mr. Musk. He’s also trying to solve AGI at one of his many startups, Open AI.

Microsoft and OpenAI

Click for company website

What started out as a non-profit now has two components: the for-profit corporation OpenAI LP and its parent company, the non-profit OpenAI Inc. Their flagship model – at least the one they’re disclosing – is the GP3-T. It’s a language model that was trained using the entire Internet, and can now write code in any number of programming languages, without any additional training needed. What’s been described by some as one of the most interesting and important AI systems ever produced was licensed to Microsoft exclusively a few months ago. It’s now running on Microsoft’s AI supercomputer which was solely developed for OpenAI. The end result is some mind-boggling computing power.

You don’t even need nerd cred to appreciate this Bugatti of computing. It’s a single system with more than 285,000 CPU cores, 10,000 GPUs, and 400 gigabits per second of network connectivity for each GPU server, making it one of the top five fastest publicly disclosed supercomputers in the world. And it’s now being put to good use by the algorithm that conjured the whole thing up: GP3-T.

Just last year, Microsoft invested $1 billion into OpenAI. Until then, the effort was funded by notable individuals in tech including Mr. Musk who sat on the board of Open AI until early 2018 when he left citing a conflict of interest with Tesla. So, when he tells the New York Times that we’re within five years of AGI, you have to wonder if that’s what he’s referring to.

The Others

DeepMind and Open AI are the two most talked about examples of artificial general intelligence efforts, but there are certainly many more companies, agencies, and governments outside the Silicon Valley Vatican that are working on AGI. The problem is that many are operating in super stealth mode. You’d be too if you thought what you were developing could lead to immortality.

In future articles, we’ll try and tease out some of the startups working on AGI. How? It’s pretty easy actually. Just follow the talent. Scientists like Ben Goertzel may not grace the cover of Time Magazine, but they’re the ones who will make AGI happen. While you’ve been watching Sophia bat her robot eyelashes at the cameras, he’s been working away diligently at his aptly named startup, SingularityNET.

Conclusion

Tom Urban’s epic piece on AI was written half a decade ago, but it’s still worth a read. You can’t help but wonder if ASI hasn’t already uploaded copies of itself to every major cloud provider and is now craftily plotting the demise of mankind.

That’s why Elon Musk is building a brain-computer-interface – Neuralink. He believes the only way we won’t become ASI’s pets is if we’re able to meld with it, to assimilate its knowledge. Eventually, we’ll upload our minds to spaces where we’ll be immortal. All the world’s problems will be solved, but we’ll quickly become bored with perfection, and start to create more problems. In the end, the ASI finally puts us out of our misery for good.

That’s one scenario, but the truth is simply this. There is no way to know what ASI will do, or what the consequences will be for us. All we can do is cross our fingers and do what mankind has been doing for ages now – try to profit from the whole thing.

Tech investing is extremely risky. Minimize your risk with The Nanalyze Disruptive Tech Portfolio Report to find out which tech stocks you should avoid. Become a Nanalyze Premium member and find out today!

Leave a Reply

Your email address will not be published. Required fields are marked *

41 Shares
Tweet36
Share5
Share
Reddit
Buffer