Safely Scaling Artificial General Intelligence— heystartup (@heystartup) December 6, 2016
Greylock Partner Reid Hoffman and OpenAI co-founder and CEO Sam Altman on the transformative potential of AGI.
The AI revolution has only just begun. Over the next several decades, numerous industries will continue to see significant transformations as a result of advanced AI research and applications. As AI technology ramps up, so does the potential for artificial general intelligence, AGI, which refers to autonomous systems capable of human-level intelligence. It’s challenging to predict when AGI will come into reach or even comprehend the impact this technology will have on society. Early on though, it will be important to have organizations that can prioritize a beneficial outcome for humanity.
OpenAI’s mission is to ensure that artificial general intelligence benefits all of humanity. This episode of Greymatter features Sam Altman, OpenAI co-founder and CEO and Reid Hoffman, Greylock partner and OpenAI investor and board member. Recorded at OpenAI’s “all-hands” team meeting, Sam and Reid openly discuss the factors to consider before releasing AGI publicly, the societal shift that will need to happen, and the conversations required for collaboration with competing nations.
Prior to leading OpenAI, Sam was president of today’s iconic accelerator, Y Combinator. Sam mentored numerous startup leaders and turned the accelerator into an epicenter of innovation. An accomplished entrepreneur and executive, Reid has played an integral role in building many of today’s leading consumer technology businesses, including LinkedIn and PayPal. As a Greylock investor, he currently serves on the boards of Airbnb, Apollo Fusion, Aurora, Coda, Convoy, Entrepreneur First, Gixo, Microsoft, Nauto, Xapo and a few early stage companies still in stealth. Reid is the co-author of Blitzscaling and two New York Times best-selling books: The Start-up of You and The Alliance.
Below is an edited transcript of Sam and Reid’s discussion. This is the second of two Greymatter episodes featuring OpenAI co-founders in conversation with Reid on the impact of AGI systems. Our previous episode featured OpenAI co-founder and CTO Greg Brockman discussing the prospects of building beneficial AGI.
“Where we are between the linear development of this technology, just about every industry is going to be transformed in fairly significant ways. Once this technological transformation happens, the question turns to what happens next? How do humans treat each other and how do humans react with this transformation is important. It’s vital for people to know that they have a positive role in the future that is in front of them and that this future is something they want to be a part of, and something they want to work towards.
It’s almost like looking through a very deep fog over a minefield. The key is to understand the path of what the future looks like and know how the fixed points change the probabilities of this technology from dystopic outcomes to utopic outcomes. We know AI technology is going to have an impact on our lives so we need organizations like OpenAI to put humanity and public trust at the forefront of the mission to push a positive outcome of how this technology will be distributed and who will empower it.” — Reid Hoffman
When To Release AGI Publicly
“Well one aspect that’s interesting and hasn’t been fully explored is the ability to understand how these decisions will be affected by introspection. I think the targets you’d want to build to is something that is essentially symbiotic with humanity. Something that helps us be better and we help it be better. This is a good, healthy dynamic, whatever that shape is.
One thing I have talked about a lot is the paradigm for how to train this technology with competitive games. But maybe, we can also figure out how to add in the training paradigm of cooperative games. Fundamentally when we think about which human ethics we want to implement, we like the systems by which we’re all winning and all evolving. This helps the future look healthier together because we both want to be in symbiosis.” — Reid Hoffman
“There will always be competition, but it’s important to reach a point where different groups, whether it’s a competing AI lab or a competing nation, come together in significant ways. I believe the lines of communication and the mission that we are collaborators versus competitors is extremely important when acting as a servant for how humanity improves.
In terms of AI technologies among different nations, I think there are positive aspects with the U.S. leading progress in opening dialogues with other nations. Right now I tend to be very positive. What’s important is that there are smart people who are listening carefully to the potential impact and include this in their own decision making.” — Reid Hoffman
A Learning Paradigm
“A post-AGI world depends a lot on what the AGI looks like and where the AGI is. What I would hope all the way from now through that point is that we figure out how to increment the likelihood of a symbiotic relationship. The paradigm for this is that learning machines means the technology is dynamic. We cannot just rely on programming the machine to not do what we don’t want it to do. We have to be learning and adapting with the machines as the systems evolve and grow, and adjust with that.” — Reid Hoffman
Start With A Dashboard That Measures Positive Human Impact
“I’m often asked when it will be time for tech to be regulated. This is a very detailed question, but how I respond is “start with a dashboard that you want to measure the technology to.” In the case of AI, the items I would want included on the dashboard are a humanity-positive analysis or sentiment analysis. We need to figure out how we are tuning the knobs in ways that lead to more fruitful conversation about the impact of the technology and have people understand each other. Having a type of measurement like this and aligning AI to that could create very thoughtful cooperation and discussions.” -Reid Hoffman
Incorporate Broader Perspectives When Building AGI
“Large projects succeed with small groups, but it’s critical to include broader perspectives and understand the larger concerns and challenges of others. Implementing these perspectives is important to include as a value, a principle, and as something you’re investing in. We need to be understanding of emerging markets for example that have greater challenges with governance and resources and realize how these factors play into our work building these technologies. We do this by always asking the questions and incorporating that insight into how we make decisions.” — Reid Hoffman