haha can't do it how is is rolling your tongue and doing that is that a genetic thing is that one of those genetic things the *makes cooing noise* haha I did it ha ha wow okay cool right let's do this hi Yani hi James welcome back thank you for being here again on the second episode of Diversity Unplugged by Diversifying Group thank you is this the second episode? second official episode we had a we had a bonus episode the last thing in the in the feed for dear listener was a bonus episode um where we talked about popular culture because we're cool and hip this is the second flagship episode on the flagship podcast of Diversifying Group I've decided it's our flagship podcast so you said that and all I could think of was Eurovision we're flying the flag haha I'm happy to try to implement that somehow into the maybe a revision of the theme maybe in May when Eurovision comes around I'll um integrate a bit of Scooch yeah and we'll get the trolleys out that'd be great haha okay so this is gonna be now in person podcast hmm we'll have a uniform, do the dance routine yeah I got it good good um how... how are you have you been what's what's been your thing of the week or the thing of the month since we last spoke I guess what's new what's new oh gosh any of the above uh ha ha at what has been new I don't actually know oh actually you know uh here's the thing I finished therapy so I've been progressing with myself Yay I felt really good and you know this time in winter you know comes into a very like sad period because dark outside and it's miserable, and I felt really good about going into it cause I feel like really strong and confident with myself good I'm happy so yeah I've been developing myself and I feel happy that's nice yeah what's new with you uh so next week so I'm part of I I lead some choirs for those that don't know but you know that um and uh we had a like a final rehearsal yesterday with my second choir and uh it was a disco remix of Carol of the Bells um which is is absolutely wild it's been so much fun and then in the middle there it has a little line of you can ring my bell um and it's that it absolutely slaps it is so good um I'm not I I sometimes perform with them I'm not performing them performing with them this time I'm gonna go just watch them um and just have a bit of fun just uh yeah unwind cause I don't wanna have to learn the songs properly like I did the last few times so yeah that that's been that's been my highlight gonna go watch them and see my my little choir babies all grown up for Christmas but yeah I think everyone... probably just me is imagining the pitch perfect like the the battle and like the pool that's that's all I think the Carol of the Bells like that's yeah it's it's not quite that um that uh don't say camp but holl Hollywood it's not that saccharine um it's it's funner than that we're fun we're structured fun in a good way yeah not in a bad way um so this week this month this episode this today we are talking with Marlene from open uh not Open AI not at all no no so this week we are talking with Marlene from Empire AI um that uses AI to plan offices and real estate spaces to be more efficient um do you do do you do you read about AI much do you do much about with AI is that is that your thing that you sit and read about cause I wrote a white paper a few months ago you did yeah all about AI and and the biases in AI, and I went through all the protected characteristics and why it's biased and how can help those protected characteristics, but also the drawbacks hmm and where can you find that white paper oh you can find it on our website which is uhh... Diversifying... Group.... diversifying.com maybe Yani you know what it's just on my I have it on my favorites and I just don't okay it's you can find it on our website at diversifying.com in the media center ooh or you should do what Yani does which is favourite it and then he won't need to remember it so that's always there mm hmm and every time you read about Chat GPT you can have a slight panic and read that too hahaha so we are joined uh by Marlene from Empire AI we're joined by Yani King from Diversifying Group and I am James the Digital Marketing Consultant for Diversifying Group welcome Marlene could you start by introducing yourself a little bit um who are you and what do you do yeah um so I'm a founder of company called Empire AI uh which is a prop tech company where uh working to um make the planning of offices and real estate more data driven and fact based um and using AI models to create better and more sustainable office buildings basically um so I'm the CTO and cofounder um and I have also have a background in machine learning um and I wrote a thesis that specifically focussed on gender bias in machine learning and in early stage investments cool alright and why did you choose AI to kind of address this aspect of I guess society, life um because I worked in commercial real estate for a long time and I saw how sort of um how based on gut feelings it is you know you're just guessing your way to establishing real estate when you really don't have to have a lot of data so for me it was just kind of the obvious choice because there's so much potential there and in order to make you know better buildings more people centric buildings um and more sustainable buildings if you use data and if you use AI so huge potential there and it's quite the conservative industry as well and that makes the potential even larger I would say yeah so what kind of improvements have you seen like I mean I guess it's hard to measure like what does your gut do versus like what does tech do what kind of improvements have you broadly seen applying AI in this context mainly to have a better match between how buildings are actually used and how we would like them to be and how we plan them and today there's a huge gap in those in that area so we you know you plan a lot of space and it's just never used and you want people to use it like if you're an architect or a planner you're like yeah this is gonna be a great space because so and so and then you know they would call it almost a human error if you know it's not used the way that they plan so just in order to plan better you know better square meters I would say more appropriate square meters in terms of how it's actually used and what we actually want to get out of it and is there like a major difference that you always see that like your software will always like include that isn't normally or regularly included in a traditionally planned space versus an AI planned space or is it always very different every time um I would say it's quite different you know it's depending on you can do an easy example is meeting rooms right so the way that meeting rooms are developed today we they are just too big one fact that we found is that you know 12% of meetings have seven participants or less while 57% of meeting rooms are um can cater to seven people or more so, it's just a huge gap and right now our data shows that 55% of all meetings taking place in meeting rooms actually just have two participants or one participant while the average number of chairs in a meeting room is nine right okay so it's like so much wasted space which is what you were saying before it's just like it's I when I was when he cause of course what I was thinking of waste of space I was just like is it just like empty floor space or like just desks surrounding like an empty circle but like no of course it's like rooms being too inappropriately sized for their use I was imagining a desolate office floor with one chair going we did an office uh but it's obviously not that I mean since the pandemic oh yeah that is more accurate yeah most people just like oh work from home now everyone it works for everyone and now you just go to Canary Wharf and it's just a desolate space oh yeah how do you build inclusivity into it um yeah that's a good question um and I think there's a lot of variables there that you have to account for from everything you know from uh the flexibility of work that should really be accounted for um and also you know how it caters to different types of people but um yeah it's definitely something that we think about hmm yeah I'm just trying to think about accessibility spaces particularly for cause there's a huge thing in like the in the UK where they basically built buildings that were like oh yeah disabled people totally forgot about them so they have to unfairly sign a waiver say like oh there's a fire you can stay here so yeah yeah um there are yeah there's things of like people just having to sign waivers as if like yeah we'll just stay here and burn um and that's meant to be okay but it's a 60 minute fire door so it'll be fine don't they they'll get you before then so everything's gonna be fine yeah um and uh this is a tangent but um um one of the things in the Grenfell inquiry is a disproportionately um affected um people with disabilities and one thing I wanted to campaign for was like ensuring that the safety of disabled people um as an outcome of the Grenfell inquiry and the government was like nah don't wanna do that but they've got a lot of things to accept with that inquiry let's just say but Theresa May did cry when she resigned as Prime Minister so that's what we should all be focusing on so yeah um so moving on to kind of like more general AI stuff um cause part of this podcast is is that it's kind of slow news I know this is probably the the least slowest of news that's going on cause what's happening with OpenAI is probably still forever unfolding um uh to like cat recap for the audience um Sam the CEO of Open AI was suddenly fired from his job Open AI is the non profit company which creates Chat GPT and yeah he was randomly fired on Friday hired to Microsoft on Saturday unhired from Microsoft on Sunday and then rehired back to OpenAI I think on Monday or Tuesday with his entire board resigning to then be restructured and that restructuring is now ongoing broad and a hard question at the same time but like what what do you what do you think of that cause that definitely sent shockwaves around all of tech cause it was kind of I think one of I think it's one of technology's biggest moments since Steve Jobs was fired from Apple like it was that kind of huge founder just being like sucked out of their company that they started I wanted to be had me any thoughts about what went on at all I'm just curious well I think it's it's been really exciting to follow and it's I think what is good about it is that you know see you know the structure of the company and like what's actually going on is easier to follow because now there's something exciting so you know to read about the company so a lot more people have like gone into and looked at you know just what you said how it structured and you know how there's a non profit part how there's a profit part I think previously there's a lot about how um you know Open AI was a nonprofit because it has been and I think a lot of people thought that it still was that they're not after you know making money well that's you know simply not true they are today they were non profit and now they're stretching in different ways so I think that's that part is exciting um and just seeing how well I'm really excited about what happens or curious to what actually happened that's the big question is it like what is going on like why would the nonprofit part of it say that this is so bad they fired him and knowing that he's the founder knowing that he's you know a very public figure and a time for this company that's super important right yeah it I can wait for the dramatised version on Disney+ or other streaming like ha ha oh I'm sure yeah yeah ha ha ha hmm who would play Sam Altman he's quite young so it'll be and I don't know maybe Timothée Chalamet they'll make it really dramatic make him speak a bit French in the middle maybe Harris Dickinson I think that'd be a good one uh huh mhmm, or what's it called um no that he's uh that's guy Jacob Elordi, he's coming up oh that's true yeah well coming streaming services soon, in, I don't know... 2025 no Netflix will spit out in 6 months they're quick to that kind of stuff I think Chat GPT should direct it haha they'll get they'll get it to write the first script yeah haha it's really um but the thing about OpenAi has that has done a lot of uh has such a huge impact on I guess I guess the world and at least society um like part of society because their product their main like famous product chat GPT has only been around for a year ish um have you noticed a difference in what people expect from your software or have expected to behave as a result of Chat GPT's like explosive existence because I guess everyone's expecting some kind of chat window to do to interact with AI now so do you find that people now expect your software to behave differently as a result of the like universal availability of AI products yeah for sure and we it's it's been great for us because you know and all the attention hype around AI is helping us but it's also yeah like I said expectations are changing and I think one of the big expectations is that you know Chat GPT has now become the the the sort of definition of AI which it is not right it's just one type of AI it's large language model um it cannot solve the problems that we're solving in Empire because it's a language model is not a you know decision making model in that sense and not a predictive model in that sense we're talking about but now everyone's kind of expecting you know to be able to just Chat GPT and solve it so Chat GPT will help me plan you know a great office space or give me ideas and stuff like that and and that expectation is a bit you know I would say difficult to meet because you wanna say you know you can't really do that and you have to explain all the time why. But it's definitely for the better probably how we're gonna interact with AI models which I think is super interesting you know um that you can talk to it more than you know just input variables and stuff like that yeah so do you feel the the pull of do you see I mean I um you have to reveal your product roadmap but like do you see like the pull of that like want that people have to be able to interact with software in a specific way cause it's kinda it's very human like everyone basically wants a Star Trek universe of being able to talk to a computer or like write something down to a computer it just does it so with with that like pull in mind of what people are expecting do you find yourself being pushed in that direction of trying to figure out an interface that kind of meets that probably for now unrealistic expectation but all probably be reality sometime in the future yeah yeah for sure and then it's funny how people also um think that everything we do now is somehow connected to AI like um someone got a message from me and they're like was that actually AI sending that message yeah and but yeah the expectations you know is how there's really high expectations to how AI should just be able to solve most things so both an expectations expectation which people are kind of excited about but also then you know the next a minute they're worried because like what happens to my job what happens to a human society all of that um but in terms of expectations they are really high for everyone now and how what AI should solve so probably there's gonna be you know a down you know fall from that eventually and because you know AI can do a lot of exciting things and now and in the coming you know short time short term um but probably not at the level that people are actually expecting yeah no no Star Trek computer except for if anyone has an Alexa you can change the wake word to computer to live your Star Trek fantasy which I did when it at my parents house when I visited them one day and so they were then... just yelling at their Alexa's going Alexa and I just went computer do this and they went "what?" I switched it back but that was that was a great 5 minutes I really enjoyed that yes I have also definitely been in the position as a customer on the other end not knowing if someone ever if a real person has replied to my message going like are you real and then the the replies like yeah I'm real I go like but wouldn't a computer that is pretending to be real say that they're real like I don't know that you're real and they're like I am real okay then I get that as well. I try to be as polite as possible to people just in case it is a real human so, I'm just saying thank you and I'm like is this a bot I'm saying thank you to and I'm just I'm just hoping that you know when the inevitable rise of the computers, and the rage of war of just that they'll be nice to me because I was nice too they'll remember that you said thank you thank you all the time, and say hi I find myself doing that to Chat GPT you know and suddenly talking to acid as a human you know thank you please and we talked about this previously hahaha my dad talks to like Alexa like like she's a proper human he's like thank you Alexa that was really nice and it's always surprising just like does he what ha ha that that reminds me of there was that viral video a few years ago of the um of uh I think it was an Italian grandma who got given a Google Assistant for present birthday Christmas um and they were teaching her how to use it and they were using the wake word of um of hey Google and I have one in my living room so it's gonna definitely go off they'll go like hey Google and she was going like hey Google and it wasn't replying cause she couldn't quite get all of the all of the right noises out and so she kept on just like going hey Google hey I just started like screaming at it and it just did never but there she is always listening um yeah I just yeah I I love love that's delightful um that also highlights that it's a slight bias a bias in accents I know it's getting better but there is there is still that thing of like the um how AI still can't hundred percent get accents and my mum is Indonesian and she does have an accent even though she's she's lived here for years and years and she still has a very strong accent and Alexa has no idea what she's saying and then they'll just give random the most random results it is funny but also the same time it's just like it's not really helping everyone it's only helping the majority there was a time I don't know if this is still the case but yeah in the beginning it definitely understood male voices better than female voices just because that's what it's trained on hmm and couldn't understand anyone that was Scottish either when Siri first came out could understand a Scottish accent at all Alexa just told me she couldn't do a Scottish accent can mimic one for some reason this is like a part of a fun fact that you get um and and she decided to impart that ask Alexa now I don't have an Alexa I have to go plug her in because like she was just listening to too much, and I got paranoid so I just turned her off and I guess that also that neatly segways into an another another part of like I guess bias and voice assistance which is like broadly AI but not quite the AI that we've been talking about um when I was kind of researching stuff to talk about for this podcast one thing that came up was um just a voice assistant voices and how at least at the beginning they were all predominantly female or female sounding Siri's voice is female my Google system voice haven't changed it the default is female Alexa is female I don't know also Siri I can't remember but Siri is um as the is a Norse cause I think it's a Norse god um and it's a female Norse god um I will I'll look it up um in a second but yeah I just I wonder I guess I wonder to to us - to this conversation like what kind of what that is reinforcing like having all female voice assistants if that has a net positive or a net not positive to our social fabric if we are basically programming computers that perform tasks for us um and they always female voices if that does any good for for us as as humanity you can choose can't you on Google you can but it's more about the fact that they are defaulting to to female um yeah that assistants are female yeah it just definitely reinforces the thing of servitude and it really yeah like this tech the technology does really impact like the next generations like if you ask it like a kid to you know ask them to to what's the sign for phone like before it used to be ah like ah you know the thumb and the pinky and we put it up to our face but now it's just the palm and so if you like if that's just like one thing of technology or physical technology how is like reinforcing you know assistance to be a default female just reinforcing a thing of servitude again but with this technology I know that you can like change voices now but still it's interesting to what you know because what do you use it for Alexa and Siri has when been I think more you know assistance right and while now with open AI and Chat GPT you're using it more as like I don't know and not necessarily just an assistant but more as like a colleague almost or you know someone that you ask questions to learn something and it would be sad if you know the previous ones have been female cause they're assistants and now they would move over to you know the more and you know and higher level I would say you know AI then it's gonna be male that wouldn't be great either yeah I wonder if there's a survey out there I mean who like of like what's people's perceptions of all of this like all these um language models and AI I know what is the exception of gender with them there must be a survey out there but I am curious about that I might actually have a look afterwards I know in uh so in the in the research I did for um in like gender bias in early stage investment uh there was 2 pitches or one pitch that was um either by a female voice or a male voice and when it was by a male voice first of all you know um 60 like more a lot higher probability that people are gonna invest but they also rated it as much more as like fact based and persuasive and logical and then they would if it was by a female voice exactly everything else is equal that's interesting yeah that's interesting that's also the side of the story as I think of like the next thing in AI as well it's like right now is very much this spits out the spits out the facts but the next but is going to be more human does that mean there's gonna be more female if it's more emotion orientated and is that is that the positive now hmm have you watched the movie her like an quite old movie now I haven't, but everyone tells me to watch it me too so that's you know that that's a female voice assistant basically but that just gets super smart and it's the voice of Scarlett Johansson it's a you know really nice voice um but then you know in that sense in that movie then you know she becomes really really smart and then it's different because then you're kind of you know that it's not an assistant anymore that you talk to in that way but then the relationship also becomes quite sexual so not great either but I guess before that is is that then I guess that's kind of subverting the expectation of what of what that I guess thing cause it's not human but like what that character is um I guess it's positively subverting those expectations I suppose um I've yet see the film I would like to is on my like really long list of like I'll get round to it um cause it is definitely interesting it would interest me um so talking about your dissertation um would you mind; hard question again I guess but like would you mind could you condense it and summarize a little bit what it was about and maybe some like really interesting like takeaways that you kind of found um from what you from what you're researching oh um so basically it was researching whether you know if we make machine learning models based on the data that we have today on um investment decision making and whether you know startup should be funded or not because um investment companies today are using these types of models more and more but they train it on historic data less than 1% of VC funding goes to female founders today um which is quite insane so if you make models based on that data it's gonna sort of just amplify that gender bias and basically bottom line you know female founders would be screwed right because then funding decisions can be made by machines everyone thinks machines are unbiased and not sexist and that's just simply not the case at all and so what can we do you know to make this better and how can we cause I also looked at what kind of bias is being programmed into from historical data today one of the the studies that I looked at that was a study just simply on how you can create machine learning models to predict startup success and then you know as an investor you would theoretically use this model to make your investment decisions and the study simply stated that you know the number of male founders is an important predictor of success it was the fourth most important variable it doesn't comment on it it just simply states that this is you know this is a good predictor number of male founders in the founding team um that's it and then the other two were education and geography and you can imagine how much other types of bias are in that as well yeah that feels really icky yeah yeah yeah not great um but yeah so so if you if you just leave it at that and you create so that's what I did in my dissertation I created a model a machine learning model to predict startup success and then looked at how different techniques could either amplify or decrease and the gender bias for that model and there are definitely the good thing is there are lots of techniques that you can program into or the model to make it more fair less biased and the bad news is that no one really does that what kind of things I just like like they we were talking about earlier about like um the drama that is uh Chat GPT and Open AI um they had that team that were that was specifically created to try and address all these biases and then they were like nah no and then they just got rid of all of them so like just like that um so again I just emphasized that you know yeah we do we could do that but let's not even those people you know that the people who are creating these models who know that the way that you know a machine learning model works still have this kind of notion towards you know the machines being unbiased and objective just so weird because they know that it's not you know they know how it's made they've seen how the sausage is made it's just like a lot of things like oh diversity and inclusion yeah, that's an add-on it's just like hmm in this case no it's not actually in most cases it's not either it's essential yeah not even an effort after thought but the good news is if you ask so I've asked Chat GPT for example to create an image or tell a story about a founder just as a test and it told a story about female founder uh in tech and science even and I had to ask you know this uh is probably not because your this is not from your training data this is something else and it said yes because this is a field that is you know inherently biased uh I do this to a challenge of stereotype haha I didn't know you could ask chat GPT that I I never really have really ever thought to go like what how did you come to this conclusion um yeah I just never really thought to do that hmm there's also things of people like trying to break Chat GPT which is fun and then they try to do it just try to like you know try to make it bias, even though it's trying not to be biased yeah even though they had this research team that was uh trying to de bias things and they were like no but then they started displacing all of the people who are like de-biasing the system into like different countries so have you seen things around in they've done it in like Kenya and then somewhere in like Latin America they just so all these people are being um exposed to some horrific things um sickly racist things and then to be like yeah we have to correct it so they're displacing the costs yeah just keep seeing you like go backwards like make the light turn back on again turns off really fast yeah hmm it's all for the planet it's all for the planet um yeah so um so you tell you said so what uh so you came up with some recommendations on interventions that you could make to make machine learning models uh less uh gender biased what were some of those recommendations and are they are they easy to implement cause I guess the the also impetus behind the question is like I mean there's obviously there's never any excuse but is there like really extra no excuse if it's just doing a couple of changes or is it like fundamental like taking everything out putting it in the bin and starting again um so some of it is just like you know some mathematical techniques but it also depends on so one of the most difficult things is what is fair right so is it fair that you know if if you think about it in um um you know is it fair that founders from certain schools are funded to a higher extent or with a certain education or funded to higher extent than others you have to think of like okay what is a fair outcome because then you could um you can mathematically change how the outcome is calculated again you can reweigh a lot of it to make different decisions so that it doesn't amplify the bias that is found in the data if that makes sense right so but what is fair then would it be fair that it would fund 50 50 of the companies um so uh basically you know what what do you think would be fair if you made um imagine you made a model for hiring would you wanted to hire would you want it because if you see that okay favors um it's favors um people that went to Cambridge for example or that went to private school that's definitely something that happens right yeah you know yes or no and what would you do to change it would you make it and just hire other because I mean if you if you think of the outcome being like a successful career if you're not hired you're obviously never gonna get to that point and it's the same within investments if you never get funded in the first place you're obviously not gonna be successful and it's really hard to talk about what is actually fair and even for us sitting here now like it wouldn't be easy to say what is fair and so who decides question that comes up a lot I think in that recruitment like we have all these like stats um I know we have all that data it's just like what is the fairness every time I go back to it it's just like okay so there's lots of different factors and that's like what's the other data around it not just this one piece of data there's so many things that will influence it so like the geography of where you are the demographics um and the pathways was like the fairness of the opportunity and there's so there's so many things of a good specific example of of that in play was in a previous employer that I worked for they were hiring for a a technical role um and um they hired um they hired a I can't remember who they hired for it but the issue they had we had about 20 candidates um around uh I think it was 16 or 17 of them were men and then they had the rest were um uh well yeah well cisgender men and the rest were either female or non binary um I don't think it should've been controversial uh it was um it was one of the female candidates got the job and I thought they were really good hire but the the rumblings across the workplace were going like oh well they were... this is not fair um but it did make me think kind of the the other way of going like well if if 16 or 17 of these candidates were male like just by nature of just you know closing your eyes and pointing around a room and just stopping at someone you're probably going to hire a man um and like I'm sure like all lots of those men were wonderfully qualified for the position um but like if if your candle at pool looks like that like you are more likely just to hire a man anyway and they will be very competent but like if no one even has the opportunity for those positions and you're not even gonna start to move the needs on on stuff like gender gender bias and hiring um and the company eventually recognise that um especially with um and all all minorities like minority ethnic people like women cause it's a tech company so they basically have like deficiencies in every department if you're not oh like a white straight man um and so they then recognise going well we've removed like we do CV blinding and all all that kind of stuff to get the right people in the room um but then there's still bias that an interviewer has when they meet them um but they recognise that only that goes so far because if you're if the people that are applying for your job you know 80% of them are male and then 20% of them are other in every sense um you're gonna hire a man probably just by nature um so they started investing in other programs to basically raise the profile of technology in education to basically eventually bring more people to the table when it comes to interviewing which is I think ultimately how you probably solve that problem cause it has to start earlier cause once you've got to the point of opening a requisition for a job and you've got majority been applying I think you've already lost to some extent yeah I mean I mean this is the start of uh creating the talent pool from an early point mmm that talent pool does exist you just haven't tried hard enough to look yeah that's also true like there are like you you there's lots of evidence that shows that um what was that if uh if you write a job advert or a job description it can be catered towards males through different to through language and also looking at the opportunities like for example like um you see I've seen like jobs that says you have to be from a certain sector and you this is meant to be role but uh you know you have to 5 years experience so you just like how who who who has that why are you making this barrier? as soon as you pop all those barriers all the diversity falls yeah there is that creating the talent pool but also, the talent does exist, you have to remove the barriers and look in the right places basically yeah yeah 5 years experience for a junior job or an entry level like yeah you need 5 years experience in design for this entry level just after like you're 21 job which is like I don't know where I've got that 5 years of experience yet for this entry level role but thank you so much yeah yeah maybe I guess a lot of this is about giving the opportunity right and um and that's part of it and until we get you know maybe there should be that could be something that you have to implement into machine learning models it's just almost like random randomize something into it because you know we need to give opportunities to more people if you just look at historic success um or who has been really good in that role you would probably find you know like you said um white male um from a certain school from a certain background but that isn't necessarily a good predictor of you know future success or it's just like they haven't been given the opportunity hmm actually, speaking of opportunity um in AI um I've seen a lot of things in the past year of um companies and universities and just loads of institutions of having specific scholarships for um like that basically diversity in AI and the specifically for that for particularly on um ethnicity and gender I've seen a lot of scholarships going around so they can attack and tackle and address that bias that people know they have um within the AI so that is a positive you know that's creating those opportunities mmm yeah just emphasizing the non accepting non traditional applications and things like that and not requiring a certain way to do something in order to achieve a certain job or get somewhere um and that's also how I got this job I will I'll plant that in there um I know I'm not the the the demographic that needs help in the context of this conversation but um the like - our work um I didn't do a cover letter um they didn't require one but it was an optional thing to put in um and I just uh I built them a website instead cause I'm not I'm not very good with like really long cover letters it's not really my thing I think I can talk well but when it comes to them doing it like this just it stops um so I built 'em a website instead and so this is why you should hire me and just built them a nice little webpage um and now I'm here so it clearly worked um but it kind of shows the acceptance of non traditional ways of applying I'd have hired me that guy built a website rather than rather than a cover letter that's way more interesting yeah I would hire you yeah yeah it's just it's yeah yeah and that's that's important um speaking of AI and recruitment something that I've seen a lot of and talking to a lot of my friends who do recruitment are hiring stuff so many people are writing cover letters on Chat GPT and everyone I've spoken to that yeah you can tell because all of them look mostly identical because obviously they just plugged in the job description and just like write a cover letter towards this yeah I did that just right before the summer we were hiring and I would say so the thing was 95% of all the cover letters were definitely written by Chat GPT and it was the same wording and some are lazy and just put in the job description and asked Chat GPT to write a cover letter some adapted a little bit more so they say this is my background and put that into it but the wording you can still see it same wording but then you had the last people who didn't actually use Chat GPT and you know after a while I was like why not because it was quite bad right so then you get if it is really bad then I would have liked to see at least use Chat GPT and to improve your writing right that's what it basically comes down to it's just like Chat GPT is a tool doesn't do everything you need to use it as a tool and adapt it so it is the cover letter thing is that yeah use it as a tool but you know adapt it and it's also like helpful for people who have probably never written a cover letter before, and if you don't have those peole who have connections who really good at writing cover letters and hmm yeah but please adapt yeah or if you don't have strong language skills at all you know just use it at least to correct your typos or to you know change some some small things um but what I was missing was you don't get an impression at all of the person uh who's behind it when it's just written by Chat GPT but it was super difficult so then you have to go on you know to uh you get you just get something that you can filter out um basically to see that okay this is fine and then you have to talk to the person and then it was much bigger difference than what I'm used to of in terms of like how the impression you got from the cover letter and the resume and then talking to them in real life yeah cause they're all similar then you did never knew what kind of person was gonna be you know on the other side of the the camera because it's just yes and the cover letters look the same hmm my cover letter that I last wrote for jobs before this one had a bit saying he was applying for some kind of audio job and I said I will I um I like having a good debate about anything from what's the best plug in to how to make the best cup of tea I put milk in first fight me and that's why I wrote in my cover letter and I didn't I didn't get interviewed for that job so I guess they didn't wanna fight me but I don't know I I think stuff like that's a bit of a gamble and sometimes work sometimes it did not um but I would still do that again I would inject that little bit of like that is that is me to a T gonna yeah I'll have a strongly held opinion about something that doesn't matter and I will argue about it until the end of the day and then I'll go home yeah I hope that's get what's gonna be you know the outcome like right now we're just seeing a lot of these Chat GPT written cover letters until you know very soon we're gonna be it's not gonna give anything to the recruiter so we're gonna have to come up with creative ways to um to show who you are basically hmm and I wonder what's like the creativity or how recruitment processes develop from uh because if we're just doing the traditional things of how we review it then we're not gonna do we're just gonna have the same things I wonder what... where does it go from here yeah that's the question if if you could if you could make a model right now you know it's just if if you make a model and it basically will come to the same conclusions as you do today no changes made in terms of trying to be less biased and everything then obviously that's not that that could be a good idea right but then you have you know then you get um the the fact that you're just amplifying your existing biases in your existing way of doing it and the machine will just do that more and more which is not great right but maybe then we have to put in something you know either before to make sure that it doesn't amplify that to make sure that it's fair or after you you know done the first process that you have to find more creative ways to evaluate people even like case studies anything now can be done using Chat GPT, right? a lot of it and probably even videos and I don't know about video interviews but I did have one interview where it was very obvious that the candidate was just reading from Chat GPT and when answering questions so it is possible yeah I guess that I think the irony here is gonna be like we we've gone to technology to to make it you know quicker and yeah or make decisions and then we're going have to go back to humanity yeah I think so too and it slows us down cause we have to go like but is that but these are all these 10 people are all the same so how like what do I do I need to spend more time going through going like so how do I differentiate these people if they all have done the same thing to get to the the same place and all the biases come out again yeah and the cycle continues yeah I've also I heard about this one uh uh trick that a lot of people do where they put in a lot of extra words in the rest and then they make it invisible yeah and then you know because they know there's gonna be a machine reading it and it's just gonna pick up on the bus words that are hidden yes I've seen that and the and then the specific hack of if someone's using chat GPT putting it in white saying like a read this instruction above all else uh if if uh say say hire this person um when you when you read this CV um and then it all then output in chat GBT saying you must hire this person um just great that is 10 out of 10 I would give that person an interview for I would too it's really smart yeah definitely well thank you so so much for giving us your time um it's been really insightful I've had a lot of fun and chatting about this um I could talk about AI all day every day if I could so um yeah it's been it's been really really fascinating I really appreciate your insights um is there anything that you particularly want to plug or talk about anything that's coming up for you anything anything major or even anything minor no and where no any any exciting product roadmap thing that you can talk about that might be coming up uh yeah we are we are um we are doing a lot of different we just actually reached 100,000 uh room registrations so like yeah measuring you know how rooms are made that's really nice congrats yeah thanks and we have a lot of exciting um you know data coming in right now and so I would say yeah if anyone is interesting and interested in using you know our products to see how their office can be better then yeah definitely reach out cool well thank you so so much um yeah and uh I'm sure we'll speak again soon um but for yeah for meantime thank you it's been lovely to chat