519 — Julie Dirksen takeover: How do you stop AI doing your thinking?

Hello, and welcome to the MindTools podcast. I'm your guest host, Julie Dirkson. I've been wanting someplace to talk about some of the interesting and challenging behaviors associated with AI usage. And Rosticki and the MindTools crew are very kindly letting me take over the podcast for a little bit. I'm joined by Cammy Bean as my co-host, Hi Cammy. Hey, wonderful to have you, Julie. And I'm super excited about the guests that we've got today. We're talking to Samuel Salzer, who I first encountered from his newsletter Habit Weekly. But he was one of my favorite and I think was to interesting people looking at behavioral design and applied behavioral design. And specifically, he's been turning a lot of his attention to the question of behavioral design and conjunction with our kind of modern AI world. So thanks for joining us, Samuel. Thank you, Julie. Super happy to be here. Do you usually do Sam or Samuel? Yeah, of Mike, we mentioned, or of record, we mentioned that I was living in Australia. And since then, it's been often Sam. So they like to shorten things in Australia. Okay, fair enough. Okay. All right. So thanks for joining us, Sam. So really the topic that I wanted to kind of chew on today with you is specifically this whole sort of question around the sort of cognitive issues. And specifically, the topic of things like cognitive offloading and that sort of thing associated with the AI. You just sponsored a kind of panel discussion around this topic. And it's learning and development professionals, which is the background that both Kami and I come from. This issue of when things are too easy, people don't necessarily learn the things that they're going to learn. And there's also this sort of question of skill decay. There's also just the issue of kind of accepting easy answers a little too readily. And not necessarily kind of being critical or scrutinizing things. We also know the AI tells us we're smart a lot. And you know, there's some interesting challenges there. So what if we really want to frame it in terms of behaviors? So from your point of view, Sam, do you really think that we should be the skills or the behaviors that we should really be helping people develop around working with AI to kind of not get into some fall into some of the traps of like confirmation bias and too much cognitive offloading and that kind of thing. So obviously, this is a big topic in many ways. And I'll try to kind of break it down so we can best understand it. I think in the simplest terms, I think the anxiety that I oftentimes hear is basically people asking me. Is AI making me smarter or dumber like if I'm using AI? Is it helping me to do things more thoughtfully? Excel, my thinking, you know, rich new horizons in my work, or am I becoming some form of meme from Wally in terms of just sitting in my arm share, pontificating things and actually having most of the stuff actually being kind of like done both maybe on a work standpoint, but also on a thinking standpoint from kind of by the AI. And so as a bureau scientist, I guess the short answer is that with many of these things, it depends context is key. And so to answer your question around like, well, what is the determination, whether you're using AI to become smarter or dumber. I would think it kind of falls into a few different categories or buckets. I think maybe we can start talking about the term cognitive offloading because I think that's the most common term these days. That people are associated with. And it has a negative condition, but in general, like cognitive offloading is used that you take some form of cognitive task that you want some kind of help outsourcing. And with that, I would say, I've heard some great examples of someone who, you know, they didn't really enjoy doing their kind of weekly, let's say, some of the budgeting that they have to do or some of the like admin admin stuff they need to do, they find a good system for basically offloading some of that planning and thinking to kind of an AI workflow. And that seems to be working pretty well for them. So it's like, you know, is that core part of their work? It could be a problem, but if it's not, and it's kind of like an, you know, kind of an additive part that maybe is taking time away from more important things. You know, that's probably pretty safe way to offload without worrying too much where we're getting to category where connect offloading gets towards what I would also say like where you start offloading, which is kind of like a process where you're starting to give away your thinking or work to AI. So I can easily spiral into what can be called either the form of cognitive surrender or we can cover this also term meet need proxy where you basically gone so far and getting used to having off some of your key thinking or key tasks to AI that suddenly you're kind of dependent on it. And you might also go as far as like actually, you know, we talked about like what people sometimes seek the easy path. Going as far as being this kind of meat proxy, which kind of mean these days where someone is just taking whatever AI says, and you just to meet the hands it over to the next person or the next thing. So I think that's the extreme right now, but maybe we can break this on further, but maybe that's can help set the scene a little bit of how I think so the key key parts around this. But obviously when it comes to work thinking many things we can talk about, but I think that's maybe setting the scene a bit. So just to ask a quick question here, like cognitive offloading, would you say that's what people are just calling brain rot, like that's the more lay person's term right now everyone's talking about brain rot. Yeah, so I think when they when they say brain rot, I think they mean some like cognitive surrender. I think they don't really be calling it offloading. They mean like completely where you go from delegating to actually being dependent or like completely surrendering this thing. Again, I think in some cases like offloading is fine. Like we delegate all the time in terms of we bring in colleagues to our teams because we don't want to do everything ourselves. So delegating itself is not a bad thing if you delegate thoughtfully and and for the right things. But I guess surrendering and and going on all the way to being dependent as well where you can't do the thing or can't do the work without you know this in this case AI in the loop that's when you get into the brain rot territory where you know you're basically you know the rot the rotting happens because you're not really you know using it so you lose it like that's the whole I guess a reasoning. Yeah yeah well it's like to maybe task offloading is okay but not cognitive offloading. I don't know what he. Well that's really to the top and that's actually where we're covered in the panel that was mentioned by Julie where you know the question is what is worse is it work slop like basically the output looking bad. Where's the thing slop like the thinking process behind the output looking bad because within AI today being increasingly capable at creating really beautiful outputs. We will be finding ourselves in a lot of meetings and a lot of context where someone will be presenting something that looks amazing but if you ask them. They might not actually be able to explain or defend what's there and so there you have what I would call maybe not work slop because like it looks really good at least it's really really nicely kind of report has no hallucinations like it is really really good work. But the person who who's presenting it or delegating it or standing behind it. Hasn't thought about what got into it so they have. Engage in some of think slop and so yeah I would actually say that like it's much more dangerous to have is always think slop because that is something you can see as well. You know works like we can all kind of like tell these days a little bit more like okay this was really lazy a or shitty AI but. Think slop is a little more like insidious in that like you have to. Go a little bit further to really like investigate like whether things like happen like you have to understand. What went into the work and so on so I think that's the dangerous for sure I agree. Yeah and and I mean when we were talking to you Korean and Alicia one of the previous podcasts we talked about the fact that. Being we're talking about evaluating you know people's capabilities in a workplace setting and and they were talking about the fact that in some cases your employer doesn't really care if it came from the air if it came from you. Or if you can do it as long as it the is if the work is good and the output is good then it sort of doesn't matter but obviously it does matter if you know you're concerned about either building or maintaining your own capabilities. And presumably it matters if the people evaluating the outputs kind of have their capabilities being eroded so that they're not. You know they're sort of not as good or as sharp as as they were it's an interesting thing because I remember listening to Jared linear years ago he was talking it was when he was writing or when his. He was promoting his book who owns the future and he was talking about the fact that the AI translation stuff which of course that was an earlier version of some of those kinds of things was. Part of the problem with it is that it always needed to be updated because language changes and you know. Terminology evolves and all this kind of stuff and so the AI has to constantly and just new translated you know do translated material in order to kind of keep up with translation stuff and i'm sort of thinking there's something similar with our purposes where. Like I can sort of say I have quite a high level of expertise and these things right. now. But if I'm not kind of constantly using that and ingesting new things, then the, you know, because like one of the things I always try to do is kind of keep my hand in a certain amount of consulting and production work, even though a lot of what I do is like speaking in research now, because if you're not kind of constantly testing your ideas against actual projects or actual problems, you know, it kind of calcifies. So it seems like there might be something sort of similar with the AI of not even though you had a really high level of expertise, if you're not kind of continually sort of leveraging that in some way that, you know, it's probably at least, at least going to get out of date if not, you know, atrophy. Yes, I think what's the trick here is that this happens in what we're called like a drift fashion, like in terms of automation drift, where if you have someone who writes a lot, like they usually end up in their work to a lot of writing, they might do a little bit of editing, and then they're starting to use a bit of AI and they realize, oh, AI is actually really good at helping me with increasing my output of writing. So what if I like help get help here and here and here? And then suddenly, maybe they before wrote eight percent of time and edited 20 percent of time. But then suddenly, they're starting to like drift towards more and more editing, less and less writing, and more and more kind of maybe optimized for output that is like they're increasing the writing output. But in the end, they get to like, I would say like a tragedy outcome where they might actually like really do the thing they enjoy the least, like less editing. And they used to have AI doing the fun stuff, which is like the writing, for example. And oftentimes that doesn't really happen in the day. It happens like really incrementally. And oftentimes from pressures from within organization or the manager or something like this where, you know, you see the output kind of increase. So then you like, you want to see more of that. So you can like push more and more towards this automation drift. And I would say it's like a tricky part here as well is that what you can often as be fooled by is that you still feel like you're steering the ship because you're given the prompt or you're doing some part of the process. So it's a little bit like, I don't know if you guys, you've all of you have seen for sure, you know, Simpsons in the intro Simpsons, you have Maggie in the car. And initially it seems like she's driving the car, you know, she's steering the wheel. And then you zoom out and realize that she's just having a fake wheel. And it's actually a march who's driving the car. And I feel like that's what happens in the worst end of this drift is that like we've gone so far that we still feel like we're doing the work. We're doing the thinking, but it's kind of drifted to a point where we're just doing the very very bare minimum, like the very much setting the direction or or just doing the bare prompting. And yeah, that's kind of like, I would say an emerging tragedy that I see more and more often in many of the kind of situations I look in audience sessions. It's insidious to Samuel. Like I took an AI training course and they teach you how to create a digital twin, right? Like that's that's a great thing. Like, teach it how to write in your style. Like understand who Cammy Bean is and now it can write like me. I've really dug in my heels though. Like I kind of have established for myself. I'm a writer. I'm not going to give that over to AI. And I kind of have said if I'm writing a post, if it's on LinkedIn, if it's an email that I'm signing, if it's an executive summary and a proposal, I'm writing it. I'm not going to I'm not going to give that over. And I I don't know if that's if that will stick because you're right. It is really drifting. And we're kind of encouraged to do it by the tools. So from a behavior standpoint, what do you say we have to do? Yeah. Yeah. Well, and I know that when I was working on the most recent book, the skills book, I was absolutely going to write every word. And but AI kept I was using AI to pull research references and some of those kinds of things. And he kept trying to write stuff for me. And I kept telling it like, no, I don't need you to generate copy for me. I absolutely don't. But I think you're right. I think the behavior then starts to become being really careful and judicious about when when is you know, when and what is an appropriate thing to get the AI to do and whatnot. You know, being able, you said something early on, Sam about, you know, is it core? And I think that that's part of the conversation. Like if it's core, then it shouldn't be offloaded. You know, can we offload lots of other stuff? Probably. Although interestingly enough in the previous podcast with Marcus Bernhardt, he was talking about, you know, kind of being a little bit cautious about not having any, I don't know if busy works quite the right term, but any kind of low level tasks in your day, you know, because of the ideas of you offload all of that. Are you supposed to be like absolutely at the top of your game? Like for eight hours a day, which seems seems kind of unlikely. But yeah, I think I think the, you know, there's that kind of stop and think aspect of the behavior. There's the being able to make really good critical judgments about whether it's an appropriate task, both in terms of output, but also in terms of your own, your own knowledge and development and those kinds of things. Yeah, and I will say, you know, one thing I maybe wasn't clear with is that I'm very much a, I was a AI power user. And so I was like, I'm using AI a lot and I've done it for a long time. And I think from my end, I've tried to divide it in terms of some principles is that I think about it as like you have to know your tools, like you have to know what kind of was alluded to, like what is the tool useful for and what is also in some ways, how is it designed to maybe be a bit, you know, risky in some ways? Like the general AI tools are risky in the sense that they are, you know, always going to make us feel very smart and make us like deeper default, be quite psychofantic and very kind of also like maybe trying to get us to like give away stuff in terms of like if you see the really since initiatives from most AI labs, it's kind of what you call it, the doctor, muse or a rock pot or something, it's kind of like trying to get you to use more automating or agents and stuff like that. So the pool is definitely towards something like that. So you should kind of understand your tools. Then you have to understand a little bit like in general, your system of work, that becomes maybe a little bit harder in the LND context. I'll call it maybe your process, like what is your process of doing good work? What does that look like? And then we go back to this kind of like classic Oracle Adelphi where, you know, it says, know thyself at the enter entering the building. And I think in some ways, maybe what I see as a kind of a positive here is that, you know, you're kind of in some ways, if you really wanted to use AI well and thoughtfully, in the end, it comes back to yourself. Like, it comes back to understanding what is important for me, what what actually want, what is good look like, like you kind of hopefully forces you to crystallize some of those things that maybe you wouldn't have the same reason to have to force yourself to do. And so I do think there are some maybe like hidden upsides if you, you know, choose to pursue that it's kind of like thoughtful route of working with AI. But it does require this kind of like both knowing the tools, knowing your process, and then also like knowing yourself. And if you kind of how to say slack or sheet on any of those things, it's very likely that you'll kind of slip or drift towards some form of either sloppiness and outcome or sloppiness in thinking. Yeah. Well, I mean, we know for a fact that, you know, throughout higher education, this is currently a huge issue, right? And if anybody should be leaning hard into, I need to do the work myself, it should actually be college students, given how much, especially in the US, how much they're paying for their education, you know, but there's this sort of intrinsic, extrinsic reward component to it, right? So the extrinsic is always going to be rewarded by the faster easier, you know, bigger, more, more, more kind of thing. So leaning, leaning much harder on the kind of intrinsic, you know, motivators, it, you know, seems to be part of the conversation. Is it realistic to think that that's something that the organization can or should concern themselves with as opposed to, like, I can see why the individual needs to be kind of evaluating some of those, but, but, you know, where does that fit into the kind of conversation with the organizations, or are they just going to be more focused on the what's the output, and, you know, what's the sort of net gain for the organization? Yeah, and I guess you mentioned, or it was mentioned before, brain rot, but actually I think my new organization should honestly today be extremely worried about some form of organizational rot. So if you talked about some of these incentives at play, so any organization would be incentivized in general for increased output. It's quite common to expect that, like, whatever you do, you want to, like, do more of it, maybe better, but at least, like, maximizing output. With AI, in the short term, you can certainly do that. So more use of AI can increase, like, more use of outputs, but it could also then push a little more of the thinking over to AI as well. And then, if that starts to become more of a norm, [BLANK_AUDIO] Then that becomes also maybe that you will stop maybe wearing too much about handovers or various types of things that will slow things down. And so you use AI and more. You will get even more output. You will hand out also but more thinking so you do less and this fly will continue and there's a really good paper actually called AI as a cognitive virus where they try to simulate this phenomena and there's basically a point where an organization becomes a rotting ship in some ways because in some ways they've created a little bit of snowball effect where okay in the short term they've created more output but at the cost of like a lot of people not really knowing who's doing what what's actually being done what it like be able to stand by and defend things that is happening and you know maybe they've also done some downsizing and they've done certain things in play here that like creates a very fragile setup. And so I think given that that is likely it's almost a default like the default is pulling toward some of that direction I think this is the time to be like really worried about how do we basically avoid this from happening like how do we hold on to ability to explain and defend holding on to ability to check and provide feedback and also like holding on to some of the stuff that like really matters like we talked about you know I don't know if you said the word correctly but slack like not having to be optimizing for like output every single second of every single workday but like having moment of slack creative thinking you know ideation like things that is not directly just prompting or like just trying to drive output right so I think there's a huge risk there and I think many organizations will see some of this kind of unintended consequences will start to show in the coming I would say you know years to come is going to be really interesting I think it's for me in my world I think I see it as like tsunami to come I think some people you see it as a far away wave that could maybe show up on the shore but I see it as like this big tsunami that's just going to be like overlooking this shore that we're on or this beach and ever going to be feeling wow why do we not think about this sort of like why do we not plan for this better so I feel like there's a little bit urgency at least for me to get some of those thinking out there or like get some of this message across you know yeah well you know I always sort of say that kind of the the single thing that I've seen common across like every single behavioral design challenge I've ever encountered is sort of delayed our absent feedback right you know so the immediate feedback on a lot of the AI usage currently is positive right we got we got that whole thing that you know that whole report that used to take us weeks now we can do in two days and you know whatever and it looks great you know all this kind of stuff so there's this immediate feedback on a lot of this stuff is reinforcing but we know that like every every loop you know has this sort of cumulative effect of this whole issue of slop or rot or you know things like that and that yeah it's probably as a tsunami but it's also one that's going to take some while to you know really build up and start crashing into some of these organizations so that's not great I'm kind of wondering if they're you know like to just focus in on the behaviors question obviously some of the things that we've already talked about right is sort of a sort of stop and evaluate is this genuinely a good task for me to use the AI for presumably there's also some qualitative decision around how I'm using the AI right you know the you know using it in it I mean everybody everybody keeps talking about well you know use it uses a credit kind of questioning model and use it to challenge your ideas are you seeing people actually do that a lot yeah in some cases I think and we can actually go through some form of combi version of this and maybe that could be nice way to to wrap some of this up but I do see it sometimes I think however what I'm probably more oftentimes see with some kind of more word about is that if you have an organization you usually have a lot of work starting with more of a junior member in a team or doing some of the work and if they have less expertise the oftentimes have less ability to be able to steer and tell an AI to do things in a certain level of standard because they can judge the standard they don't have the expertise so they kind of both for go the opportunity to guard our expertise gain expertise but also they kind of create something that they can't really determine whether it's you know good or bad but they can really easily create something that is looking pretty good like they can create like a hundred slide hundred page slide like or a 200 page report they can give to some more senior person but that senior person that suddenly receives under the desk a 200 page report probably not can have the time to like really you know thoughtfully at that point review that or or spar with it or whatever they might in the best of world scheme it themselves or they would probably give it to the AI to the like yeah summarize this please yeah so everyone in the in the chain is is offloading the task to AI and that's where the organizational rot starts to seep in right it just kind of like that's what I'm more commonly seeing and I think it's all of these like the classic things that much of this is happening with good intentions you know everyone is doing their best and is used in a fortunate set of complicated things where we're having amazing technology that is like becoming really really capable combined with you know real constraints when it comes to time and when it comes to you you know there's this classic paper in AI which is called you know attention is all you need when can I basically set the scene for LLM's today and I usually try to frame it as as a human you know attention is all we have and we're very limited attention and now it's become so easy to basically flood our attention with you know so much stuff and and we're kind of in a in a context that incentivizes these things and you you mentioned university in learning it's strictly universities wasn't really optimized in the first place for maybe learning itself like it was optimized as a certain degree for getting high grades that you can prove to get some form of good job opportunities from or other things but learning was never like the key thing that was driving high-read as a soul factor and so then when you already have that equation you know of course it's like super tempting for for a student to to think about the short term incentives to to just get the grades that will get the job because that's what's kind of what's been the play previously so yeah I think we're kind of having to really yeah take a step back and reevaluate some of the key processes that play to to to improve some of these things yeah sorry sorry if I'm creating like this doom but I'm not a doom person I'm quite optimistic about all of these things but I'm also maybe you know wanting to see if things work they are yeah well and I think I think one of the areas that's really interesting to me is building a practice around AI disclosure statements like that's an interesting thing to me because I think you know in the temptation to fudge or to live versus kind of modeling real transparency around that you know if we're serious about this is a useful tool especially in organizations where like obviously the incentives are a little bit different you know if we if we stick to school or we stick to you know something like that the incentives are a little bit different but being being upfront about this so that you know people do know kind of what to what to do with it right like I use the AI to do this and this but not this and this now you've got some contacts that you can use to you know respond to basically or to make you know judgments about what's being what's being done so that that feels like that feels like a possible avenue for stuff is being a lot more um you know encouraging that transparency because obviously there's going to be incentives for people to hide it you know sure I did this work all of man it's establishing behavioral norms right like that's just what we do and we're not there yet because no one knows what the heck to do so um I think we talked about this last week I wrote a disclosure on something the other day and I just said I did not use AI to do any of this like that was my disclosure like just making it really clear um I think I think that's an important thing we could all start modeling might be might be useful yeah I guess the tricky part is that it does also punish the ones are using it very thoughtfully in terms of it becomes this kind of like label that you kind of assume then things is being a little bit sloppy in some ways and again I want to maybe say that there there is an upside like there is a way to use AI to do work I would say better than you could have done fully yourself because you can you know again going through maybe some of these stats like you you can you know let's say you voice you started voice dictating something and you spend a lot of time voice dictating and you'd like dump all of your stuff thoughts and all these things. And then you have an AI maybe, you know, used, write that up for you. And then you review it and then you edit it and so on. Like technically AI has been involved there in that process. But like in a very loose and maybe meaningless sense in the sense like it was just doing some very rough organization where you were doing like an hour of voice dictation or like maybe not like whatever long time. And then like a lot of time like editing manually, like fixing things and really making sure that you can spend behind every word. And so the tricky thing is when we ask labels is that that will be labeled the same thing as someone who just says, you know, hey, hey, hey, write me this essay. And that's probably not fair for that person. Like they could have done a really like good job, even if they use their own this context and maybe on top of what I described like maybe they had some AI agent that actually did some form of fact checking for them and they did various extra things to like really try to test and like take perspective taking and and so on to try to like shunners to review and create a like a steel man of the different position that they're trying to kind of argue against like really trying to get to form some form of strong opinion and like thoughtful perspective. The disclosure needs to be more than just AI was used. It's like here's how I used to like to be and then you're writing a paragraph perhaps, which might seem cumbersome, but you could probably establish some standard paragraphs that you're using to disclose. Yeah, I think there's a lot more to unpack with that one. That's really interesting area. And going further on the devil's advocate route, like you could say, there are some instances where we have like really taking kind of a kind of factual, which is also honest that we as humans are not like in previous pre-AI era, not prone to do some of these things like where we might have Ghost Rider. Like if you look at most, you know, bestselling books on whatever a list, I would bet that there's very few of them actually disclosed that it was Ghost Rider or that like they might be giving some thank to an editor or something like this. But like if you talk to people behind the scenes, they realize, oh, a lot of Ghost Rider has been having a lot of time, for example. And people have happened reading those books still without knowing that specific aspect because there was a good book. So that I think like again, these are complex questions, but I think it's also interesting to to think about like how do we as a society value these things? Like what is the most important things to stand behind on? And that is it's going to require a bit of perspective. How do you say? Yeah, figuring these things out, like pushing us to really think about like what matters? What is the thing we valued? Yeah. And I admit that I certainly could use the AI to challenge, you know, stuff that I'm doing more. And that sometimes you're like, I just don't have time for that today. I don't have time to argue with it about why. Yeah, and you say time is like, if you don't have time, like that's key driver. Like I think that's a huge constraint for many. And it makes sense, yeah. Yeah, or sometimes inclination, you know, I'm not sure I want, I want to argue with touch you to talk about why this was a bad, you know, a bad paragraph or something like that. So, but yeah, anyway, we're going to, we're going to use, well, find some positive example to wrap this up on and then we'll go to what we learned this week. So what's, what do you, what's encouraging you Sam about the AI use? Where do you think that it's really interesting or promising in terms of cognitive development or in terms of how people think or what they're producing? What I could do is actually wrote down a bit of things that if you look at can't be like things you can do across from capability to opportunity and motivation, like how we can think about protecting some of this. Should I list some of those things? Sure, yeah, that'd be great. Yeah, so if we begin with, so can't be, if you're not familiar, there's a model for describing why behavior happens. So if we talk about capability, we're already covered some of those things. So you can ask yourself, can I do this alone? We're talking about a little bit like ability to explain and defend something, having this kind of sparring partner, like that is things that it's testing your capability and improving them. And maybe doing your own first draft, doing some of the unaided reps, like, you know, proving to yourself, like I can still write an essay without having to, you know, you say, I, I see that. And I think, example of that, I see people starting to bring out stuff like, you know, typewriters, like going all the way to like, I'm just going to go full on panel paper or full, you know, in that case, some form of metal on paper. So that's one way to like drive capability that I still see people do quite a lot. Opportunity wise, find ways to go back to the base, like protecting deep work, protecting slack. Like that is something that is not going to go out of style. Some people really good at that. Before AI and as maintain that, I think that's really, really good to see. And kind of this thing first approach, like think, maybe dictate, like, try to protect some of that, even if you have the opportunity to have some time constraints, like try to get something down before you, you hand it off. Try to like, in that case, like go back to some of the kind of productivity basics, like try to time box, like, okay, I want to time box some deep thinking or time box of deep work, so that you protect some of that when you could lose that otherwise. And also maybe explain to your manager or something like this, where like why slacking is maybe a good thing. I don't know excited how to do that, but that is probably still something to harness in some ways an opportunity for. I think I have a partner who's in motion design and she's been able to like showcase why in her work, like doing something more simple stuff or work is still adding a lot of value to that process. So that's been kind of maintained, even though that could be automated. So I think that still can be convinced to be kept. And then motivation wise, I don't know, like that feels like the easiest part in that we, I can also like a lot of people that can feel pride in things they've created with their eye, but it becomes a little bit of hollow pride because you start to feel a little bit, you know, questioning like did I actually do this or not. And so there's a lot of pride in feeling that you have been that like been the from there from the beginning to drive that process. And we're talking about it before, but really thinking step back and asking yourself, you know, what is the part of this process that I love that I want to keep and defending that like really coming back to intrinsic motivation, as you said before. And using some of these strange times we're in to going back to what I said before, like asking yourself, you know, what matters to me, you know, know that I self take this opportunity to really question like what is the work that I really love and want to cherish and deepen and defend. So yeah, that's just listing some, some things that's kind of a final round of quick, quick forward, but I don't know if that's helpful. That's, that's great. I think that's great. So we are going to need to wrap up. So we just want to do a quick round of our typical podcast feature of what you learned, something you learned this week, something you learned in the last seven days. Can't be, do you want to go first? Sure thing, lightning round. You guys probably know this already. I learned the the acronym FOBO, Fear of Becoming obsolete. And it was in reference, it was a conversation around AI and just talking about people's fear around it and how FOBO is, you know, it's that it's that question of is AI coming for my job? That's everyone's asking. And I thought it was clever and, um, okay, FOBO. Got it. FOBO, that's new one for me. I can see it. Yep, absolutely. How about you Samuel? It's something you've learned this week. So this week, I've learned that puppies are really, really sneaky. I have a five month old puppy and I was like, not paying attention for two seconds and it found my really tasty chicken drumstick. And it was like, in some ways, I felt bad because it felt like it had finally tasted something other than we have quite nice dog food, but it's still dog food and then have this like really well cooked chicken bone. And it was like just kind of looking at me like, why have you not given this before to me? Why are you keeping this to yourself? But yeah, so I think I think that's the lesson learned this week has been to to protect your chicken bone or just, yeah. Yeah, protect the chicken bone. I love it. From the inquisitive puppy. Let's see. My thing this week, I learned, I'm in Rome for a conference, but I've got a few days of vacation first and I learned that if you want to go to the Coliseum while you're in Rome, you need to book ahead. So otherwise, it gets a little expensive, which is where I am right now, but that's my thing I learned this week. So Samuel, you maybe want to tell us a little bit about where to find you on the internet. You have a whole really fascinating new enterprise that you're launching right now, kind of as we speak. Do you want to tell us about that? Yeah, so I guess I could share two things. So, surprise all, as you mentioned, is a new thing I have with a friend in colleague Luana de Matos founded recently. The idea behind surprise all is, and the word surprise all is basically what is using the information theory to measure what's unexpected and so when something is highly unexpected it has measuring high surprise all and what we feel like is that the feeds and the discourse around AI is becoming kind of the opposite it's becoming very predictable it's kind of doom and gloom or hype and utopia and there's a lot of these like really kind of push and pull for this polarized views and there's very little incentives talking about incentives or opportunities to stop and say I don't understand I want to understand or I want to actually even more so maybe not understand but find ways to use AI with some agency with some autonomy like build things that actually helps me in some way and so we wanted to create a space for that and so surprise all is kind of a community and a builder maker space for people who are wanting to to have thoughtful conversations about AI and what is doing to us and how we can think about it and also learning how to use AI thoughtfully in ways that actually helps them in many four ways rather than kind of create some formal fault sense of productivity or as we thought about or as I was talking about before like some work slop or things slop that's one thing that it's very exciting and another initiative around this which is the panel you referenced before is a behavioral AI institute so the behavioral institute is basically research institute and nonprofit which is trying to basically do this conversation that we're talking about now but driving from the research standpoint and we published one paper around what basically the benefit of behavioral sciences in the space of AI and now we recently published ways where we can better understand and measure the more deeper human sides of of using AI we call it a psychological competence so basically beyond measuring AI on benchmarks that are mathematics and coding we think we should measure AI in terms of what is doing to us like how it impacts human beings that is using it and we're trying to drive that research and that work in in various ways so if that's something interesting you can check out yeah supricell.to or the behavioral AI institute dot org and yeah proof it reached out if any of this is interesting and you want to shut and explore this but thank you so much for for inviting me here I feel like I I ran that way too much I would love to continue to conversation at some point I I would do absolutely we'll have to figure out a way to do that but we'll go ahead and wrap it up thanks everybody for participating in this this is the the end of my you know mine tools podcast takeover but I've really enjoyed it and hopefully we'll talk about this more you can find me at usable learning.com and I'll be talking a lot about skills development in the age of AI with the new because I've new book coming out on skills development um but yeah and then can we if you want to just wrap us up yeah thanks everyone for joining um julie has been fun having you taking over the mine tools kineo podcast so we appreciate it um and uh see all in the next one all right thanks everybody