Cognitive Inversion: Amplify the Mind, Don't Offload It

Hello and welcome to another episode of Applied Intelligence. I'm your host, Roderick, and I'm joined by Kelly Sarabyn. Thank you so much, Rakhi, for joining us today. We have a pretty exciting show taking a very different approach to artificial intelligence. And the insights that we expect to get into, I think, are going to be quite exciting for the audience. So I'd say let's get into it and have an exciting debate, discussion, what have you, on applied intelligence. So

Rakhi, let's start with who you are and how is AI part of your world these days?

Thanks, Roderick, Kelly. Very nice to be here.

My name is Rakhi Rajani. I am a computer scientist and a psychologist by background. And I've worked in everything from the big research labs in the valley to operating companies in the biotech space, in the automotive space, in the city space, all sorts of things. And I was studying human cognition and machine intelligence way back when. And I was fascinated in how these two things came together.

And that's been at the core of all the work I've done, regardless of where it is. At the moment, who is not talking about or leveraging AI? Good question. I think everyone is. And in my work today, it's coming through in the clients that I serve and the conversations that I have. But more and more, it's interesting to me that it's becoming less about the technology and more about the interaction and the, if you like, human machine teaming that needs to sit around it and all of those issues that are starting to spark when technology

kind of tends to find its way into the world. And I think those for me are the interesting conversations, partly because of my background and partly because of where the world is going.

That sounds fascinating. And the idea of being in this field for so long, you know, really pre-AI, is particularly interesting to see this a a sort of a revolution happen that is completely different, I imagine, from the your field of initial study.

Yeah, absolutely. mean, the term AI and the words and the theory around it have been around for quite some time now, right? mean, it's some, what is it, 70 years, someone said to me the other day, which I wouldn't, which I wouldn't put past anything, or at least a very long time, that might be the wrong number. But anyway, the point is the terminology and the thinking has been here for quite some time. I think the instantiation of it, the application of it, the ability to access these,

models and ways of thinking and technologies has shifted dramatically. Now I've worked in tech for a very long time. I've invented tech. I've been in that world for quite some time. And the speed to adoption has massively increased. And if we think about what's helped that with AI, I think putting an interface on something that is quite complex, so an easy to access interface on something that is quite complex, that feels like you're having a conversation and responds like you're having a conversation.

has made the technology, has made artificial intelligence so much more accessible and so easy to play with and use that it's very different to my days of programming or building technical solutions that require us to be in the code all the time. Access here has just become so prolific that it's unsurprising it's taken off. And I think that it's an interesting conversation around

technology and design and interaction models and all of these things are coming to light in a way that perhaps haven't with the technology in the past.

I definitely think it's the democratization that has shifted, right? The tech I mean, obviously the technology's taken leap forward itself, but I think that the as you I think you nailed it, what's really changed is is the democratization. So the access across t pretty much anyone who has access to a computer.

Have you s do you spend time thinking about this in the personal front or is it more on the work front? So I think that we a big topic of conversation is how humans are going to work with AI, but then there's a separate conversation, which is what about a people using AI as their therapists? What about using it as their relationship counselor within their personal relationships or their life coach? these are I I saw a a while back therapy was one of the highest trending use cases of

Yeah.

Okay. you

for for chat bots for people. so just curious, is is the personal side something you've given thought to or are you really more laser focused on the human machine and work?

So much more in the work context, I think, is where I spend most of my time. The personal side, though, is hard to ignore, right? I mean, as you pointed out, Kelly, it's like people are leveraging this as relationship coaches, as tutors, as life coaches. I I'm guessing that the number of times an LLM has been asked what's the meaning of life is probably hundreds of millions of times, probably at 2 a.m. as well for most people. And so I think it's a...

I think it's really growing in that sense. And that brings about its own questions and perhaps opportunities, but also worries for a lot of people. think that if I come back to the interface again, there's a belief that comes through when someone or something converses with you in a way that feels convincing. The question for, I would imagine for those fields, for therapy, for coaching, et cetera, is

what should people believe and how do people believe and all of those processes and cultural processes and the way the brain works kind of comes up in different ways to how are we using machine, human and machine teaming in a work setting, right? So my focus is much more on the work side of things, but I just out of interest and because of how things are moving, it's hard to ignore the other.

The belief is an important part of both dynamics. You know, in work what you need to believe is say is the analysis that's being presented back to you something that's accurate, should you push back on it, that sort of thing. And same thing in the context of a life coach.

Yep, I think so.

So when the situ you know, I recently had a situation, it kind of makes me think about this. Now it wasn't necessarily a work context fully, but I was at a conference where at the end at the end of the conference they let a lot of students in, just to kind of go around and talk to various vendors. And I was there with my company and the students come in and they were wanted to grab pens and things like that that we had to to give away. And they were you know, like little jackals. I had to hold them at bay. And so I did is like, well, all right, if you guys can answer some questions like what do we do, how do we do it and things like that.

And so it got into that our software was used for making movies and VFX and things like that. And I asked them, well, you know, what what how do you do that? And and the kids all said, you just use AI. AI. And and nothing more. And for them it was like AI was this massive catch-all. and it could do almost anything. And it was an interesting and somewhat disconcerting assertion on the part of these children. It's like you need to solve a problem, you just use AI, and then you get your solution.

Mm-hmm.

it makes me think like what does that mean in the context of the workplace? You know, if you do an analysis and it was via AI and it's wrong, can you just blame AI? How do you push back as an employee?

I think that question is coming up so much. I think that the process that we're hedging toward and what you're describing is what I would probably call something like cognitive inversion. So it's a way from the human is generating to the human as kind of approving or attempting to evaluate machine generated content. And because it's so easy to generate that content, we're going at it because we think we can make things faster. My question is,

how our kids, how our adults, how in the workplace is the brain tuned to just approve machine generated stuff versus be the people that are generating the content. So I don't think this is just a prompt conversation or a it can do it conversation. I think this is where for me, what's really interesting is understanding how the brain works and how intelligence is formed and what AI can tell us about how intelligence has evolved over time, right?

And that's not directly tackling what you were just talking about, Roderick, but I think it speaks to what people believe, but also speaks to what we need to learn about the processes that need to shift in a work context in order to leverage the value of AI. So from the human generating content, we've got machine generating content. That question then becomes, what do humans then need to do?

It's not just accepting that and going, thanks very much, I've got my document or I've got my essay, I've got those things. But I think it then is about where and how is our ability to process that information and analyze it and do something with it. And when it comes to you in formats that feel right, it's harder, I think, to apply that lens of critical thinking. Or maybe it's not harder, but we believe more frequently than we used to.

Yeah, it's super interesting. I think there's been a lot of articles recently about how AI can cause cognitive decline and at the same time a concern that people who are coming into the workforce now are not going to learn the domain expertise that older people like us have, whether that's for coding, for writing, for graphic design, people who kind of

did these professions before AI, did them themselves, right? And so if they shift to a purely strategy review, they have a different mental model and and understanding than someone coming into the field fresh, never having done that code, and then trying to just position themselves as a reviewer and learn to be a reviewer. And so I think it raises, you know, there's a lot of concerns is does humans move away from some core expertise that they that they really need and we kind of

A little and I have these conversations unsurprisingly quite a lot in in work situations and I think that if we think

go down in a spiral of no one actually having the judgment that's needed to correct the AI. is that something that you s you've really thought about?

So a few things come to mind. The first is that I think the question should be in terms of how we use AI. The question should be around amplification, not kind of tempering down what the human brain can do. And I think that leads to a question about education and critical thinking and bring back debate teams. don't know, the things that make you question stuff, right? These are things that have...

I don't know how much they exist or not. I was never part of a debate team. But the point is that that banter that you have and the ability to question something I think is hugely important here. The second thing I think is really important is cognition never existed just inside our heads. I mean, we talk about cognitive psychology. We talk about cognition. I don't believe it's ever just existed here in our brains. If you think about work situations,

Cognition exists in the artifacts we use, it exists in the documents we read, it exists in the conversations that we have. That's how stuff gets worked out. If you look at medical environments, which I've been in quite a lot, or even critical environments, whatever they may be, it isn't just about the things you're thinking, it's about you plus the machine plus the thing on the wall. Cognition is effectively distributed. And the question for me is, how does

how does technology like AI become one of those elements of cognition, not the whole answer to things? And I think that to me is a question about where and how does it fit into the work process? sorry, that's a slight tangent from is in cognition on the decline, but I think that it could be, I mean, but I think that's a question for beyond AI. could, know, all of this.

The research that has been done on short form versus long form information has been asking and answering the same question. So has social media made us all short form content consumers, right? And can we absorb long form content question? There have been studies that have been asking that for a while, yet at the same time now we're seeing more bookshops spin up because people want that long form information again. So I think that the, is it possible? Yes, absolutely for that to be true, for that decline to be true.

And I have some hope that the models will, and I don't mean technological models, I mean the systems and the human thinking models and the environments for thinking will have to change, but we need to do that and be on the front foot of that. Which is a question I think for working environments, right? Which is not just how do we build this technology into what we're doing, go use AI, but I think it is about how we rethink how we work. And so that's maybe a slight sitting on the fence answer, but I'm not completely.

I'm not completely in the camp of it's all bad, right? I don't think that that has to be true.

Well, I like how it's a reshuffling and that's a good point, 'cause there's been other reshuffling, you know, just the personal computer alone reshuffled how certain tasks, you know, maybe less math in the head or on the paper and now move to the computer and things like that. But in that context of the reshuffling, what seems to be kind of getting to Kelly's point, little dangers here is that th you could ask AI to do a lot. And so it's not it could potentially be a reshuffling if you're disciplined, or if you're not disciplined, it can end up being an enormous quantity of work.

Yeah. That is it.

and and a lot less done by the person.

Yeah, and I think one risk too is the speed of change. So I think I think a lot of this will happen at work. It almost seems more pivotal at work because that is where adult humans typically spend a lot of a lot of their working hours and a lot of their mental load. And so if if organizations and businesses, and other, you know, employers

Bye.

don't approach this the right way. There there is a risk that this is rolled out so quickly and people essentially use it and disengage in a way or don't kind of have the practices in place that would enable it to be amplification versus almost replacement, you know? And so I guess how do you how do you what would you think of as best practice in terms of if you are rolling this out within a business or an organization of how to think about

Yeah. Yes.

How do we ensure that these are tools that are amplifying human output and and working in a way that doesn't ultimately undermine t people's cognitive abilities versus just rolling it out and either hoping for the best or or having a bunch of employees who kind of sit back and ask Claude to put their deck or their Slack messages or their emails and just let it run, and which over time can decrease people's cognitive abilities, of course.

Yeah.

maybe just throw in one other comment here is when you have the CEO of Nvidia saying if one of my developers isn't using a half million in tokens and I can't remember what, like a year or something like that, then I'm worried. you know, that you have something like that as sort of this other viewpoint that seems to put a lot of pressure on if you're not doing it and not moving fast, you're not moving fast enough.

Great, questions, of Great, great questions, all of those.

I think that there has to be some, having been in the tech space like you guys for quite some time, there has to be some level of playing with it and seeing what happens, right? And I think that unless you deploy it, you can't see what people are going to do with it, deploy these tools into an organization. And I think that that's okay on some level. I think you're the...

you were making there Kelly about, know, what do we do is such an important one because having these conversations right now, folks are worried about are people going to get lazy and just sit back and have this tool generate everything for them and some people will. But I also think that there are very conscious conversations going on at least in the rooms I'm in about the value that an organization is trying to bring and the culture of those organizations.

and the tools and processes that they use. And in most conversations I'm having right now, the question is how do we get rid of those? It's about how do we bring a tool like AI into the conversation? And I think that at least, so one of the examples is one of the clients I'm working with right now, we're looking at how we leverage, in this case Claude, to help us with one of the processes that leads to quite lengthy written documents.

And what we're not doing is saying, you write it for me? Right? Here's some input, can you write it for me? What we're saying is, how can we leverage and how can we build an agent or a script or whatever it may be that helps us, that prompts us, if you like, to ask the right questions as we're working through a problem. And that's been quite an interesting exercise because what we haven't done is say, here's some data, build me a document, I'll be back after my coffee. What we're saying is we run a lengthy process to try and generate content that is

quite complex and quite involved. And can you, it play a role in prompting us to get to the right answers? And so I've been writing scripts and working with people that build agents and stuff that help us do that. And that's a very conscious effort to not remove the human, but to actually leverage the value of this technology in aiding in cognition versus offloading cognition. Because the people that are just doing the write this for me are offloading the cognition to the machine.

And I think that's the thing we want to avoid, not just because it often doesn't generate the right outcome. And we as humans are not verifying the outcome because there's lots of it. So it's a compounding problem. But I think that this is about, in all things, thinking about what you can do with these technologies in the same way that I think we need to be asking and answering the question of what new roles do we need?

So just the other day I was having a conversation with someone about, are we building bias debt into the models that we have? We've talked about technical debt for a long time. Are we building bias debt into the models that we are generating? Because we're probably, in my opinion, compounding hallucinations over time, right? And I think that when you start asking and answering those questions, it makes you think about what potential roles you need. So I was...

And I was writing something the other day, but also chatting to someone about, have QA engineers that are quality assuring the code. What about the introducing QA roles that are evaluating the models and the biases that might exist in them or where there might be hallucinations or all those kinds of things. And those roles don't yet exist, at least officially. But I think, and call them what you want, that might not be the right terminology. But I think this is also a question, not just about how work is done.

but what kinds of thinking we need in organizations to ensure amplification, to verify differently, and to assess what we are building and the impact that it will have from a debt point of view that isn't just technical.

So something that there's a couple of points there that I find really fascinating. One of them is you're talking about a an AI that essentially works to the strengths of humans. so you're kind of leveraging what the strengths of the humans are and the strengths of the AI. So you're saying the AI doesn't just do everything, it becomes a sort of an augmented experience. But how do you find out where the real strengths of the human are relative to the AI? You know, what does that process look like?

Oh gosh, that's a great question. No idea. I think...

Let me think. I think in organizations that are really good at doing something, they have built the kind of mechanisms to do something. So whether it's how to write documents or how to structure documents, like good writing has a process that goes along with it. And I think that when those things are strong in an organization, it's easier to say, how can you help us augment that process rather than like,

take it over. The question about how do you identify where, where, or the question you asked about identifying the, the humans that are doing this in a, in a really compelling way and, and bringing the tooling into that, I think would make a great research study. But also I think that that has to be about, actually, I think where I'm going with this is, sorry, way to where saying,

What are the leadership models that we need in organizations that are leveraging AI? And I think that this is about who the leaders are and how they're building teams and how we are assessing capabilities in a much more thoughtful way when we're bringing on a tool like AI, which is completely a non-answer to your question. But it's a fascinating conversation to have had about leadership. Yeah.

it's something that's not yeah. But it's totally ad hoc, right, at the leadership level.

I think it ties into something. I'd be curious 'cause there's some of the areas that I've thought about and I feel like a lot of people in in different fields are agree that IA isn't quite there. I think one thing that you really think about is the things that are closer to what one would describe as the arts. And I think when that comes up in business would be brand story, brand strategy, truly compelling writing.

Not like perfunctory everyday writing, which I think AI has nailed down really well. but really distinct and I think that ties into leadership because what inspires people, what inspires humans is strong narratives, not generic narratives that everyone's now become kind of attuned to the to the voice of of AI as as like

Yeah.

And and I will say like on a per this is a personal note, but I think it connects to this ability to tell stories within work and to come up with like really strong distinctive brands and also really strong distinctive brand writing is on a personal level I often put poems into AI and just ask them, hey, what do you can you well how do you interpret this poem as as like a someone to talk to? And I will say it does a good job about it, right? It identifies what the poem is.

But unsolicited multiple times it w has said, let me write a poem for you in this similar to this style. And I'm like, your poem is awful. Please stop. Please stop trying to write a poem. I didn't ask for it. And so I think it really highlights for me the limitation there. While it can write an amazing research paper, although to your point the veracity of certain points need to be validated, but in terms of of of the abilities and the tone, aligns very much with academic papers.

Yeah. Yeah.

But when it comes to this more distinctive voice and and and kind of things that fall into that, I would say higher level of creativity and human resonance that I do think is necessary to successful leadership within organizations, especially large organizations, but really any size, that seems to be something AI just doesn't have the capacity for now. So that seems like one area where humans are gonna continue to need to own and excel.

Yeah, I think you're absolutely right. I think that creative processes, I feel like are kind of coming back into the folders as we do this right now. So I think brand storytelling, you're absolutely right, is one area where it's about connecting, right? mean, these stories written well are about the connections you make between

words on a page and what people are thinking and feeling and all those kinds of things, which is harder to do with, well, the models don't do that. The AI models just struggle to do that for reasons that we can probably all understand. the arts are having a great conversation, right? The art world is having a great conversation about AI and what role does it play? Does it have any role? Is it going to take over all of these kinds of things?

I think the higher level creative thinking, but higher level cognition and where this isn't just about the transactional outcome, but about feeling and emotion and connecting with something are ones where it's a question for me about how technology and how AI can interfere in those or collaborate in that sense or change that dynamic.

And I think the more people experiment with this, the more interesting it's going to be to understand how that plays out. I was in a session the other day around building an agent for something. And the folks who were in the room that were great at building agents, I was surprised at how much they enjoyed a Post-it note session, where we were scribbling these on Post-it notes, putting them up on a wall, and then synthesizing. And I was shocked when somebody said,

we need to bring these like physical techniques back into our work. We've never done that before. And it blew my mind because, you know, I think on paper and I think by scribbling things down, but it completely blew my mind and, and it shouldn't have done. But this is, this is, this is, this is, you know, a group that is sitting in front of a computer all the time, which is absolutely fine. I've done that a good chunk of my time, but those moments where you let the brain breathe and you let your thoughts breathe outside of a screen, it's not even about the screen, but

in the world comes back to my point about cognition being distributed and when you can move bits of paper around and you can think more laterally because just by its very nature doing this and being blinkered means that your brain is also blinkered. And I think the creative process that you're talking about and the things that humans can do in that sense require a physicality for want of better phrase that is harder to do with these tools.

I think that's absolute No, I was just gonna corroborate that to some degree. You know, I I one of the fields that I I work in and worked in much more intensely previously was architecture and I was talking to someone about this and they said if you put a physical model of a building on in a table with a group of people

Yeah, and I think you're checked Go ahead.

And there is no better way to get consensus amongst that group than through a physical model. And it seems like and it would make sense, right, over you know, hundred thousand some odd plus years, couple of hundred thousand years really, you know, our brains have adapted in such a way that we we are gonna conform to understand the physical world best.

Okay.

and so you can imagine that in some ways, you know, this type of human interaction or stories that truly resonate in the ways that stories have been told for generations is gonna work better. And I I I wonder if there's an element of AI where you know it's trained on everything. And if you're trained, if you have everything, then it's very hard to be exceptional in some ways. And I'm wondering if there's an element here of just the nature of the way AI functions, is that it's very hard for it to be exceptional in the way that that humans can be, you know, a leader of a group.

Great company is oftentimes exceptional because they can tell great stories. And that's something that I imagine the sort of the mass training of AI makes it difficult.

Okay. you

Well I think it's especially exceptional when you get out of side of like raw intelligence, right? So I think we can see a line of sight too, maybe AI does become an exceptional physics professor. Maybe they don't, but maybe they do, right? And I think if you think about a computer and and raw analytical intelligence, which is one way that humans can be exceptional, but there's a lot of different ways that humans can be exceptional. And AI doesn't necessarily have abilities in those arenas.

Okay.

And I think you bring up a really good point, which is ultimately work, the economy, whether that's private businesses, the government, nonprofits, they are there to serve human need, right? And so there are a lot of human needs that can't be necessarily met through a computer. And I was on the treadmill the other day at the gym and I saw an ad from Chase Bank and the whole ad was

Come to us in person. We have people available to talk to you. You don't, we're not just bots. Like literally, this was the ad. And I was like, wow, we're already at that point where large businesses are saying, here's our point of differentiation. You can talk to a human. Well, in the past, I don't know if for people nodding to our RA should probably remember, it was great. I no longer have to go to the bank. I can do everything in my app, including up those checks. But now we're seeing this backlash where our human beings designed to just constantly interact.

Yeah. Yeah.

With bots or the digital world, the answer is no. And I will say in B2B marketing, I see this trend of live events, right? And and enterprise sellers, right? They're still very much like, hey, we offer this human component. So I do think it it touches another way where you think about where does what use cases and what skills does AI fit into and where you're do humans want a human? And I don't think we know the answers, but I think we know it's there, right?

It's a.

I think you're right. I think that if you look at the retail world, I'm seeing a lot more artisan brands pop up, right? So folks that are, know, clearly I'm not a fashion person, but the people that are kind of making clothes by hand and, you know, they only make this many pieces a year or something like that. Like the value of those businesses is growing, I think, and the need for people to have that

and provenance, if you like, around a thing is growing. And that is outside of the machine interface. That is about real things in the real world made by real people. And I say that because to me, that speaks to what you're talking about, which is the need to not just see a human, but to feel human and to have the physicality of a story and provenance and all of those things. And we're losing some of that with data. Like the two words that I think of

destroyed a lot of things in the world are data driven. We're all data driven, data driven, data driven. And we all know that AI is trained on data. So, you without the data, isn't it? But this notion of what's the story in there, what's the provenance in there, all of those kinds of things are becoming much more important. And those are human traits, which is what you were talking about there, Kelly. So the fact that you can go to the bank and speak to someone in person sounds glorious in many ways. And at other times,

The fallback of the app is fantastic too, because you don't want to speak to anyone, But the question is, how do all of these things, how do we, what are the systems that we need to build? To me, that's the question. What are the systems that we need to be thinking about in education, in work, in healthcare, in all these places? And how do we design those as a whole versus just the single component parts of them? And I think that we aren't thinking that. We're thinking about embedding a tool.

Exactly.

And I think if you think about this from the perspective of embedding a tool, that's all you do is you embed a tool and then you hope that people know what to do with it. But in critical situations in healthcare, for instance, where we're saying, know, believe this technology, I think that the need for human...

I don't know what the words are, expression or belief and things like that are going to become more important versus what people think is going to happen, which is just weird. We're just going to trust the AI. And I don't think that's happening because I think belief is a very strong part of this and getting to belief is different.

That's a really good point though, because you think about what motivates people or what persuades them of a given argument and oftentimes it isn't data.

you know, it's it's a compelling story or it's a compelling narrative. And those are things where if you have something that is purely analytic driven, purely data driven, you're gonna be missing out. And I'm thinking also with a doctor, you know, what's gonna motivate them to give the best quality of care? I mean certainly I want them to have the best data, but I don't think even having the best data is necessarily gonna get them the best motivation.

Yeah, and I think this is why we're seeing conversations around adoption and interpretation coming to the fore more so. That's not to say that data is bad, right? We need great data and it means that we can know so much more. But I'm seeing more and more conversations about interpretation, adoption,

human machine teaming, all of those kinds of things, which are, I think, very positive. And I think they speak to how both we will trust information, we will trust people, but also how the culture of organizations will shift as well. So those conversations to me are really great, right? The adoption conversation or the interaction conversation, I'm finding fascinating. I think that they're also getting, it's also helping people question

how the tools can be embedded and used versus just saying deploy.

I lo I I love the verification or quality assurance point that you brought up as a as a future role because I actually don't think that's being talked about very much and I think it's it's like a almost existential problem because if you think back before AI tools, right, it was you could see in terms of this concept of disinformation spreading on social media. I I don't know if you've ever known this, but sometimes you'll see there will be a study done that

resonates with people's biases and it ends up going viral and accepted as part of like common vernacular and it be it it becomes cited by legitimate sources like say the New York Times but if you trace the original source back it was to some random publication that didn't actually have like a statistically significant sample but because what the learnings was resonated with the culture, it became just accepted. So that's been happening for

fifty plus years, probably all of human history. But what's what's scary about AI, right, is w I've I've seen now like some I think the majority of content being published on the internet na now is from AI, and so I think you have this velocity of misinformation that goes even past social media where AI is tr continually training itself on content that has now been produced by AI, which was originally mass sourced from the internet. And so

Hmm

you w you enter the possibility of having disinformation at a speed and scale that we've never seen before. So I feel like this this role of how do we have humans validating and verifying information so we have that quality control is just is just vital. And I don't feel like organizations are talking about it enough or even the the conversations about AI I think are more about strategy or art or leadership versus

you

How do we actually implement processes to to validate that we're not going down this feedback loop of disinformation which is which would be scary, right?

Yeah.

Maybe just to ask a question on top of what Kelly just said, because I'm kind of really it's an interesting point. You know, one side is you're you're training off of this data that has misinformation. So maybe you're learning something wrong. But the other side is are you also, because of the way the model's structured, are you starting to learn failures of logic in some way? You know, so there's not only the data's wrong, but potentially the capacity for analysis could also be wrong.

I think so and I think that for me that's where some that's where some of these what I would just call new roles which I've made up in my head but the new roles are about

Yeah.

questioning the logic as well as questioning the output. I think that's absolutely spot on. You question the provenance of data, you question the logic and you question the output. You don't question it because you're saying it's wrong, but in the same way that you would question other things, right? And I think that notion of, I don't know, what should we call it? Like the ability to probe into these things I think is so important. I keep using the phrase at the moment.

which again I made up, but compounding hallucinations. my point there is we talk about AI hallucinating, I think people hallucinate too, but if the out there is what it is, right? And it's been there, the data AI has trained on has been there for many years, it has various biases and it has various truths in it. But models are trained on human interaction with models unless you tell it not to train on those things.

I think that we're in, we are in a place or we will be in a place where let's just call it, know, machine hallucinations and human hallucinations are compounding themselves, compounding each other, sorry. And over time, whether it's disinformation or whether it's just this set of information that is slightly off kilter, there has, that's not just about the content, it's about the logic that got to the content. So to your point, I think you have to interrogate all of these things in the same way that,

back in the day, back in my working out maths on a piece of paper day. At school, my teacher was always looking at the logic by which I got to the answer as well as the answer. And I think that your point is a great one because the logic in how we come to a conclusion is so important in how we explain how we got to something. And it's also important in generating belief and in generating trust in the answer.

So if in a critical situation somebody came to me and said, you know, I think you have X disease, see you later. Without giving me the logic that got there, how much would I believe that versus not? And so the logic point is huge. And I think the positive of that is that it is, from an educational point of view, I think it will surface back, which is why I mentioned debate society again, It will surface all of these.

these disciplines and ways of thinking that have perhaps not been at the forefront as much as they used to be. And for me, that's a great thing because it then speaks to what are the important topics in education that we need to be focusing on and or in the workplace. Logic and how you get to an argument is one of those. Expressing how you got there is another one. And when we're just saying, here's the answer, without that logic, I think there's something missing.

It's funny to think of Kelly used the example of a physicist as being somewhere where maybe AI can be competitive. Because one of the more well known physicists out there is is Richard Feynman. And he was so well known for telling stories in ways about physics that were understandable by a lay audience. And I think it does show that, you know, you need like, you know, that explanation is is hugely valuable, not only from the analytics perspective, but maybe even for the communication perspective as well.

Absolutely, I mean I love Carlo Rovelli's books and so I would say he does the same thing as I, yes absolutely. Comes back to the art point.

So I would love to hear thoughts on if the ways that AI can if it's deployed correctly can improve humans' ability to have higher cognitive fun ch function and and more success. And I think, you know, for me at this stage,

There's a couple things that I think about. One is the ability to access such amazing research so quickly that would have taken a long period of time to to collect without the technology. I mean, when deep research was first launched, that was the first time I actually personally started engaging with AI because I'm like, this is just incredible that you can get a whole research report. Now of course you have to validate it. I think the other hope and potential, right, and I don't think we're really there at any scale with this one, but is

Okay.

That AI can qu sort of take a lot more of mundane task off humans' plates. So a lot of stuff, you know, if you think probably far as along when you think about like a customer service agent within an organization, it can answer basic questions that a human was very rote for a human to just repeatedly answer the same questions over and over and over. and certain automation and and project management within organizations. So in theory, that could free human up.

this human up for this more creative, this more strategic, this leadership level, this this validation, focus on logic. So those are the two ways I'm kind of thinking about there's potential to kind of unleash more human potential, but would love to hear how you're thinking about that if if if things go well.

I think it's a great question. think I agree with you. I think it can unlock human potential. I think some of the areas that I've been working in or investigating are ones that I think are a huge draw for that. The first is in medicine, right? So in understanding

human, like understanding health data or understanding the genome or any of those kinds of things, like accelerating our ability to do that means that we can provide better medical care, sometimes faster, more accurately, things like that. And I think that that massively increases the potential for human beings to live healthier, not necessarily longer, but healthier lives. And that's one where I think it has huge potential. I think

to your point about research, right? Being able to pass huge amounts of research that would otherwise have taken months, years, whatever that may be, is a great way, not necessarily to just accelerate thinking, but to access content that would have taken you a very long time to find. I think this isn't about, and I think in the situations you're talking about, there are either,

examples that you mentioned that were about just quickly answering questions without needing to spend too much time thinking about it, right? We have the answers, let's get them out there. That's one of them. In situations where judgment is important, I think the purpose here is not to offload that judgment, but it is to augment the ability to get to that judgment quicker or more accurately. And that's where passing hundreds of thousands of research documents is a great example of where you can do that. And you can see that happening in any discipline, right?

and

It's something I've heard is that people are automating the the sort of the repeating of meta analyses that might have been done previously. So as new research comes online, it's really easy to take an older study and incorporate n new data that's being generated. and I think what's nice about that is you're right, it it it does it can stop at the point of, okay, now the human can use their judgment and interpret and apply the these results.

Yeah.

But it you know, it's it's synthesizing and presenting the data as opposed to coming, you know, completing the loop, if you will.

Absolutely. And I think just a riffing off that second, I think the other area that I think is fascinating to me and I think where AI can accelerate understanding is just in understanding the brain and intelligence. I think this notion that AI is here to replicate intelligence in my mind is wrong. I feel like that actually AI will help us understand intelligence and the brain.

if you study both of those things side by side, because it accelerates our understanding of how reasoning works and all those kinds of things. And so for me, and I'm sure there are great research labs around the world that are doing this, I'm not, but I can see huge potential for us, many of the reasons we've talked about, helping us accelerate our understanding of how the brain works.

reasoning works, how intelligence works, all of those kinds of things, as long as we see it as a partner versus a thing that replicates human intelligence. But back to your question, Kelly, think healthcare for me is huge. Education, defence even, the arts and creativity, mean, even if it's not generating great poems, which I agree doesn't, just the ability to surface some of those things or to...

Ha ha ha.

question things or to interpret text in a way that we no longer do. I think there's, it's not even about huge potential. I think this is a really interesting way to experiment with the information that we have. And I live in a world of wonder most of the time, I'm a daydreamer and I think the world is amazing. And my interaction with some of these tours is about extending that wonder and playing with ideas that I think is hugely.

That's actually a great note to leave on is is think of that, you know, play with ideas and and to explore and and have a bit of wonder in the context of AI. I think that's something good. I know that like for instance Kelly loves doing some biohacking research in the context of of AI and I love the idea of using it to make ourselves even smarter and use it as sort of intellectual sparring partner, if you will.

All of that.

Yeah.

Absolutely.

Well, thank you so much for your time. This has been a great conversation. really inspiring. absolutely.

Thank you guys.

And we look forward to probably another conversation in the future because this field is not standing still, that is for sure.

Absolutely. I'm very excited to hear what else you've been talking to people about. So thank you very much for having me. it's been, I've really enjoyed the ideas. It sparked hundreds of more thoughts in my head. So thank you.

Perfect. All right. Well, take care, everyone, and thank you for listening.

See