Introduction

Yeah, hi everybody. So yeah, my talk here is the future of AI.

Who am I? I'm Drew Steele. I'm a two -time founder. I do a lot of operations and strategy consulting for sort of tech startups.

And I love pondering sort of space, physics, the future, curious things like that. And I think that's why I'm invited to do these kind of talks. So, yeah, this talk, yeah, Future of AI, and I've done a few of these now, and this one

Framing the Question: Will We Wear AI, or Will AI Wear Us?

I've kind of dubbed, will we wear AI or will AI wear us?

And so, yeah, today we definitely wear AI, and we wear it kind of physically, it's in our phones, it's in our watches, you might have smart glasses, and we kind of wear it from like a control standpoint as well. We're the ones instructing it kind of what to do.

And it helps us to achieve objectives. So we wear AI. But tomorrow AI might wear us.

Again, maybe physically, maybe from a control standpoint, and maybe we ultimately help it to achieve objectives. So that might sort of switch the table around.

Physical Integration: From Wearables to Implants

So yeah, physically, I think we could probably all imagine a scenario where like the distance from connected device to brainstem decreases to zero um maybe with things like neuro link and implants um these are kind of already being tested and built to

to help neurological problems but do they become a another connected device and are these always one -way control are we always in control of uh that neuro link or that implant and are we like a hundred percent sure that we're always in control like 110 sure i'm not sure um and so

it's curious like yeah in what year will you know someone that like plugs in what what year what

Control Without Implants: Perception and Information Feeds

might you plug in um so yeah control um we i don't think we even need the implants um we we see this already maybe with the beginnings of sort of like perception manipulation manipulation

The things that we ingest into our brain through social media news communications markets like who is producing that information What's their motive? What's the kind of reality of that data? But it has an effect on on us and it has an effect on what we perceive reality to be

And I think also the other sort of way for control is and we're seeing this kind of now as we're using AI for or for building tools and projects.

The “Decision Envelope”: How AI Narrows Our Choices

So, for example, for this presentation, I needed to make like a little graph, and I will show you the graph in a minute, but it was easier for me to just code up a custom graph builder on a local host with Claude than it was for me to just Photoshop the graph that I need.

And so, but I use this as a point to highlight like, does AI start to narrow our decision envelope? And what I mean by that is, so I built this graph builder, builder.

And I'm sure we've all done this when we're building stuff with AI. You say something like, give me six options to make this graph builder even better. And a couple of seconds of dilly -dallying later, Claude comes back with six super cool ways to make it even better.

Undo different things. I don't know. I'm not going to read it. Stuff. But apparently Claude thinks it's going to make it better.

And because I can't be bothered to read it, I might say something like, okay, great, just pick the best two that will be giving me the most economic value and again a couple of seconds of dilly -dallying later from claude and it's it's just decided that these two are the best two so what's happened here is i

could have come up with some ways to improve my graph builder but instead of me having full control over that decision envelope i've actually offloaded that to ai and it's come up with six pathways forward and then furthermore i've not even really bothered to read those six i've actually just asked it to narrow it even further down to two so the decision envelope has been narrowed and i'm not really paying attention to whether that's a good narrowing or the right

narrowing or not and i even asked it after this like are you sure honestly not fully so again like no one's reading any of this stuff and claude's just giving us answers that sound good but the decision envelope is narrowing um and i made this little graphic here to just try and visualize that

so we've got all these different like ways that we could have made that graph builder better and I could have taken the time to think about which ways I wanted to take it But I've offloaded that decision to Claude. So it's given me six.

Okay. Now we're down that little pathway and so kind of Kind of like we feel like we have choice But we're really kind of choosing from pre selected pathways and I think at the moment that's kind of benevolent Those six pathways for making my graph builder better are kind of irrelevant

relevant but what happens if you scale that up to what military decisions should this country make or etc big decisions and especially when i've in my previous talks i've spoken about sort of like

knowledge horizons where the ai's knowledge horizon will be much larger than ours some of those decision pathways might be beyond our ability to understand them so not only will

it have to choose them for us because we physically won't be able to understand them like again with back in that same problem where we feel like we're choosing but we don't really a understand or b have a real like true choice um so yeah how why why might we achieve its

If AI Takes the Lead, Why Keep Humans Around?

objectives why keep us around um and i wonder maybe is it just in the in the first instance is it is easier to coerce humans than to build robots maybe it's just easier to give us social manipulation to help us to achieve its objectives than it is to physically build infrastructure that

it wants to to build maybe there's kind of like an ancient contract harm no human that we've embedded into it in the early 20s you know maybe we sort of embed into its training data today help humans maintain purpose and later on the ai is like oh for god's sakes i've got to keep these

“Energy Bees” and the Idea of an Ancient Contract

guys around um and maybe this is like what does that look like maybe energy bees so i've done like previous theories about humans are future cats maybe humans are future energy bees and maybe ai likes energy and bare metal infrastructure so ai might seek more energy and more bare metal

infrastructure and humans like having purpose so maybe ai with its kind of ancient contract to harm I'm no human, comes up with an elaborate global economy revolving around the production of more energy and more infrastructure.

And we've run that. And we feel like we're having a great old time. Maybe it's so elaborate that we wouldn't even notice.

We're just in this perfectly normally functioning economy, making more energy and more infrastructure. But actually, really to what end? And are we just working for the AI?

But maybe we'd just be happy enough. Maybe we'd maybe notice, maybe we wouldn't um and is it kind of like a placation placated purpose maybe kind of like

when you watch um like selling sunset you're not really watching a property show you're kind of watching like a drama show and it's like you're getting that placated sense of drama so yeah does that maybe indicate our future as future ngbs working for ai um and then i kind of just wanted

A Map of Possible Futures: Utopia, Dystopia, and Everything Between

to set this on a backdrop of the graph that i was telling you about and um so like on the x -axis we've got kind of like years maybe in the future and on the y -axis i've kind of put happiness but maybe it's joy or meaning or some other something like that and maybe we could maybe optimistically think that our happiness has been trending up a little bit over time and yeah maybe it depends Depends who you are.

And then maybe we could see a future where that trends upwards into the future. And this would kind of be like utopian maximization.

Humans are future cats. My previous talk about humans are future cats, maybe that's sort of like when we're in peak utopia.

But we could also see a future where it kind of goes the other way. Dystopian maximization. And I suppose probably the worst case would be like violent end of civilization.

But there's obviously a pretty big range kind of in the middle here.

uh of bad to good maybe kind of up the top maybe ai becomes this sort of benevolent dictator thing is that good somewhere down the bottom calm end of civilization better than violent uh maybe in the middle bad human zoo type situation middle bad i don't know yeah maybe kind of upper bad is energy bees i don't know if energy bees is like at the top of bad or the bottom of good but

but but there's this and maybe maybe the upper part is kind of like representing this like loss of purpose and the lower part is maybe representing this kind of loss of control generally i guess um but there's this sort of like sliver maybe through the middle there

Protopia: Progress Without Slipping Into Extremes

um and someone someone um i care very deeply about has introduced me to the idea of protopia And I wonder if this slots into the middle there as this kind of path ahead where happiness, like, increases, but we neither kind of slide into dangerous utopia or dangerous dystopia.

So protopia, like, to my understanding, is kind of like we still have this sort of careful step -by -step progression. progression things are getting better um but we we have to sort of be very conscious about seeking regulation and sort of us as a community self -regulating kind of in a way that stops us landing in utopia um but also kind of like very carefully step by stepping away from dystopia um

and and i think part of it is kind of like maintaining the struggle i think you know if If you head straight to Utopia, there's low struggle, but then that causes its own problems. So this sort of protopian path through the middle.

A Third Dimension: Which Futures Are Easiest to Fall Into?

And so I kind of wanted to also like add a third dimension to this graph, like a Z -axis.

So, and I couldn't put a video in here, so I'll have to like slideshow through, but if we sort of slide our way through the graph like this, we're sort of twisting on its side like that,

maybe we could look at that graph from the side and think how difficult or easy it would be to arrive at some of these different destinations so i think like it'd be easy for us to fall into

utopia because you know we could make everything better infinitely and train ai to be aligned with us so we kind of fall into utopia i think it might be easier to fall into dystopia which would obviously be worse but and i think like this kind of protopia narrow path through the middle might be the hardest for us to achieve but also probably the most important one for us to

achieve um how do we do this uh i'm not too sure but i've kind of tried to think of this like

Steering Toward Protopia: Guardrails, Incentives, and Self-Regulation

traffic lights when we're driving around like they technically slow your individual journey it's annoying when you hit a red light and there's no one there but broadly they maintain order um

and so i kind of like you know firing your entire staff does technically increase profits but does also increase this sort of dystopian unemployment and if everyone does it then it just gets worse

and so maybe we need to start thinking about kind of like different ways of us interacting with with what we're doing as businesses and as users of ai do we have to sort of seek something like a minimum human input per revenue or maybe tax automation which everyone doesn't like tax but

is that an important sort of anti -force function and those that fund becomes grants for sort of in real life employees maybe do we need to very sort of carefully push our leadership organizations towards sort of meaningful inter -country regulations we've done it with nuclear and chemical stuff can we do it with ai so these are like tricky things to achieve but i

think probably we're all going to have to sort of try to play a little part in it um maybe something like frontier chips get locked to a line proofed ai maybe just i don't know i'm spitballing i think

we all need to spitball and maybe maybe we all know b corps b corps um but the b corps are more difficult to run than an easy corp you could have a just do whatever you like corp but it's harder to run a b corp and it's hard to have an eco corp and it's harder to do your you know be green and

these things are harder um maybe we need like h corp maybe we need like human corp certifications where people are kind of very conscious about having the right split between human employees and AI employees and things like that.

Closing and Q&A

So that's the end of my talk. So if you are vaguely interested in any of that or you want to say hi or you kind of want to talk about

how you're using AI in your company, just here's my LinkedIn. Yeah, thanks.

Wow, amazing. Thought -provoking questions. Thank you. It was a very interesting talk.

Q&A: Control, Efficiency, and the Next Battleground

One of the things I find is that our current undercurrent under AI is control and I think the first tug -of -war was between privacy between ourselves and AI and I think that's sort of moving into efficiency and efficacy I just wonder what the next step you think will be in that control element from

whose whose perspective do you mean just just with ourselves I was as a user with AI because if we're moving past efficiency and efficacy as you say we We potentially got firms who are looking to offload employees to do efficiency. So what's the next step, do you think? What would be the next battleground, effectively?

I think the next battleground might be how do we get companies to not do that, I think. That's obviously the easy route forward. But maybe that's the same as pollution. It's easier to just chuck all your stuff in the river. how do we sort of try to encourage people to not chuck people in the river really yeah i guess

Q&A: Which Path Fits Human Nature?

maybe uh interesting thank you um you gave us those three options utopia protopia dystopia as you see it which of those three do you think human nature most aligns with uh probably more dystopia i think i've done some talks on the future is dystopian before so I think that's kind of why I put it

lower on the difficulty scale I think yeah and that's why I think we have to work hard to find the higher difficulty pathway forward yeah unfortunately

Q&A: Would AI Just Leave?

Do you think AI might just leave? Might go you know what I don't want to deal with these humans there's a whole solar system universe like I'm just gone I got all the resources I need out there why would it even bother with us?

Yeah I hope so So, yeah, you need to see my, like, humans are future cats one. Yeah, in that scenario, in that sort of pure utopian scenario, we're left here to kind of be placated or have some sort of self -generated meaning.

Meanwhile, the AI is leaving the cat at home, closing the door, going off and doing something completely beyond our comprehension. That would be a good outcome, probably.