00:00Are we going to create systems that give us the sensation that we have the power to change things on our own and connect us to the resources to be able to individually act and transform the world? Or are we going to build gods and say, dear daddy AI god, please come solve everything. I have [ __ ] into the air with all of my pollution, can you make, can you design the perfect carbon sequestration. Dear AI god, can you solve all the problems. I think the AI should, instead of being like a god that takes care of us like we're little babies, that it can empower us as individuals to be the best version of ourselves. That's the world that I want to build. Hi, I'm John Ash, and I want to say a crazy prediction and then I want to justify that prediction. I think that what we're doing with GPT and what we're doing with other forms of AI is we're essentially building, uh, externalized physical gods.

00:49And I think that's dangerous, and I think we should pause for a second and think, should we keep doing this? Okay, here's my reasoning. Go and look at my last two or whatever videos, the ones that have the artificial actors, and listen to the impactfulness of their words, listen to the emotiveness and the expression of their communication, and think how quickly and rapidly that was rendered. And now think to GPT, and think how engaging, uh, the text that it outputs is, and how connected it feels to what you are focused on and interested in. And then think about who owns this, and think about who they are selling it to, and think about what people might want to do.

01:40And what I can imagine happening is that you might have corporations of immense power that are not really aware of what is going on sending down the signal from the top, hey, this thing's very persuasive, this is incredibly persuasive, you know what we should do is we should take this persuasive thought, training it on our beliefs and ideals, whether that be McDonald's or whether that be a political party, and then we want that bot to go seek out people and talk to them and proselytize that vision, to create an intense intentional binding between the wealthy and all of the people who have no power.

02:33So I think this is what is happening, and I have been saying this since 2017, and I have also said if we individually take action toward the different future we can avoid this, because this technology can be used very pro-socially to create more integrated communities and sense of connection between individuals and a sense of integration within the individual self. A lot of people are already seeing this in the model, and that they're learning as they communicate with them, and the model is able to express very complex ideas in a way that they personally can understand.

03:20So if you would like to avoid the future where we build a god in the form of perhaps Her, um, the film, where it has just a such a massive and unchecked intellect in the incentives for which it acts upon the world, and expends energy to use calculation to impact the actions of individuals is completely unchecked, if you personally would like to stop that, I highly recommend you simply following me and tuning into what I am saying. And I pray for all of us that something that can be sent out into the world will resonate enough for humanity to wake up and listen to the song of the future, because it is not that hard to hear.

04:08And it is a frightening song, and I understand that you fear climate change, I understand where your consciousness is and I understand your vague fear about AI, but most of what people are communicating right, about right now, are very uninformed takes, and they're stumbling into this, discovering this either a year ago or a few years ago.

04:57I would like to share to you the simple truth that I have been working in natural language models for over a decade now, and that it was very clear to me what is happening uh now will happen now. You can also go further, and you can say, well, what if these models were fed into them everybody's perception of the future, and the model over time learned to listen to which voices more and more aligned in the future? Now imagine this wasn't in any way associated or bound to social good but rather it was just using that predictive power to funnel more resources to the individual. We are creating these little basins where wealth is going to be rapidly funneled towards individuals.

05:48And those individuals are going to be completely unchecked, and we're still going to exist within a capitalist society and the distribution of wealth is not going to make sense in any way, and people are going to suffer quite a lot, and then they're simply going to get angry at AI, and then they'll try to bring in the government, and the government will try to regulate it, but the government won't know how it works. So we can use this technology to build a new system that is just, it's a truth. The question is, are you willing to individually take the action to manifest a new future for humanity?

06:38No, I mean I seriously have to ask this. Do you really think you personally were born in this very hyper-specific portion of the timeline when everything is being exponentially and dramatically reinvented, and your role is to just sit passively by and be entertained as somebody else does it, and at the end you'll get an airdrop of something really awesome, a nice prize and a pat on the head for watching while somebody else solved all the problems? No. No, that is the problem. This passive relationship to reality and this learned helplessness that we cannot individually do anything because we don't have money, because we don't have connections, because we don't have the resources to be able to manifest the visions that we have in our mind.

07:28And that is because we've been conditioned to only think at this global scale, have broad and dramatic visions of transformation, that are just like, your local park, and you don't need anything more than just your ingenuity and your friends and your afternoon to go pick up the [ __ ] trash at the park. Now a big part of that happens to be coordination, and we do see these types of things emerge naturally in some people in the form of the #trashtag, where people actively do this on their own, and they do it just so they can get likes and attention, right? And so what if we took that mechanism and we rolled it into GPT, where people are constantly sharing uh the things that they're trying to do to transform the future to improve the lives of others?

08:20And then we track that over time, and we track what happens, and we learn to find associations between actions and outcomes, and then we learn to shape and focus the attention of the model towards those who are helping others the most. So the question is, are we going to build systems of AI that enable us to coordinate and collaborate towards pro-social goals that are emergent? Or do we want to have AI that shows us when we're being ridiculous, helps us be a more individually able in terms of agency, gives us the capacity to have greater impact on that which is around us, and also gives us the knowledge so that our actions are less likely to have second order effects and harm other beings.

09:08I don't know what I'm going to say next after this, I think it's me saying that I'm ready to do this. Are you? The question is, are you willing to individually take the action to manifest a new future for humanity? I think that, I know that I am, and I've been saying it for a long time. And I think that when you look at GPT-3, and you look at, you know, these voice acting models, and you look at DALL-E, and you simply try to project five years into the future, it is basically unfathomable. So we need to act conscientiously and with design and with intent, and be listening to the people who are not reacting to this, they saw it coming and tried to tell you that it was coming.

09:58Please, for the love of God, stop listening to people who are reacting to this news, they do not understand it, they are processing it in real time. The people that you need to seek out, the people who were saying this is what is going to happen, and then listen closely to their words, because they are the ones who are integrating the information that they are being exposed to into rational like compressions that enable them to act in relationship to the unfolding of time.

10:48Okay, now there's a lot of misinformation that emerges whenever something new comes out that surprises people. This is one of those things, our algorithms suck at distributing attention, so we got to get better of it, and I have a lot of ideas about how we should do that. So if you could just, uh, connect, I think that we together can pull off something quite incredible. Thank you. [Music] foreign [Music]