A lot of resources will be needed to create an Artificial General Intelligence. There are many efforts to make various narrow super-intelligence. Narrow Superintelligence are AI-related software systems that are beyond humans in some area of activity.
Google makes a beyond human capability search engine. Google is a trillion-dollar company.
Tesla is leading the way to self-driving cars. This will be beyond human capability in driving. This is taking billions of dollars and the data from millions of cars.
Deep Mind was acquired by Google in 2014 for $500 million and had losses of nearly $500 million in 2018. AlphaZero is a modified version of AlphaGo Zero. AlphaZero handles any two-player game of perfect information, gained superhuman abilities at chess and shogi. Like AlphaGo Zero, AlphaZero learned solely through self-play. Google has spent more than $1.5 billion on Deep Mind and it has 1000 employees. AlphaGo technology was developed based on the deep reinforcement learning approach.
Deep Learning and Deep reinforcement learning seem to be lacking key concepts needed to enable full AGI systems.
Scaling and boosting AI capabilities are benefited from more computing resources. Billions of dollars of computing resources are needed to put an AI system at the leading edge.
Even after an AGI is created it will need to have access to data and knowledge to grow its capabilities.
Any promising pathway to AGI will thus need to be profitable and useful as it is being created, as iterations are made and the system leans, grows and develops.
It still seems like we are many years from achieving general intelligence. This means when AGI does emerge, there will be many highly profitable and capable narrow super intelligent systems.
A fast and capable, minimally intelligent general coordination system for many narrow intelligence systems will be useful and profitable.
AGI will come from very powerful and highly profitable companies. The companies may not be the current major technology companies. Any company with more rapidly improving AGI will become highly profitable and will attract more resources. I do not think the learning and super-intelligence explosion will be scaled in only a few years. I do not think a superintelligence explosion would happen in secret or in stealth mode.
We are seeing the emergence of a dominant electric car company and a dominate space launch company now. These emergences are taking decades and are highly public.
If there were an emerging AGI company then I think the rise would look like Deep Mind and Google or self-driving within Tesla. The accumulation of the needed resources would be visible.
The only exception would be a very well-funded combination molecular nanotechnology-AGI company. They would need to develop a powerful leapfrog in molecular nanotechnology that would be applied to supercharge and merge several narrower AI systems into an AGI that suddenly had a trillion times more compute resources than any competing system.
If it was AGI with just new software, algorithms and concepts that were leveraging the same computing systems as competitors, then the gathering of resources to develop and perfect the system and feed it data would take many years. A leap in maturing AGI capabilities could still be sudden but it would be like a running long jump. The run leading up the leap would be long and visible.
Below are several interesting videos discussion superintelligence and pathways to AGI.
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
I want a wearable math coprocessor with an ocular user interface that can provide real time analysis of the environment and items in that environment. The operating system of such a device might be a good use of artificial intelligence.
If AGI isn’t dependent on exotic hardware:
Support artificial narrow super intelligent systems for a successful future, avoid the race to the bottom AGI will bring.
The reason I, in 2005, suggested to Marcus Hutter (http://www.hutter1.net/prize/index.htm) a prize competition to compress Wikiped was similar to the reason I:
was that objective criteria for prize awards permits cash-rich but expertise-limited managers to assure themselves that expenditures return value rather than ending up feeding parasitic “projects”.
It is trivial for a bureaucracy of any size, no matter how ignorant of the details of narrow or general AI, to advance AGI:
Put billions into the Hutter Prize purse.
Every dollar awarded will represent an AGI advance more rigorous than any of the other so-called “SOTA” benchmarks out there.
It really is that simple.
Even with nanotech, developing new computing systems would take time. That would also be a visible process. Those new systems are likely to be widely useful and profitable, so they’re not likely to be kept within the leading company. The nanotech itself is also a profitable product with many uses.
So if nanotech arrives before AGI, we’re more likely to see a global advance of off-the-shelf computing, rather than an internal leap within a single company.
Just manufacturing a bunch of already existing computing systems isn’t much different with or without nanontech. Such systems are already mass produced, and it would still take time to set up the server room and get everything installed.