[Music]
hello and welcome to technically speaking a podcast where scientists and Engineers come together to chat about a common interest share knowledge and satisfy some curiosity I'm Antonia and in this episode I'm joined by Ellie and Jasmin to talk about what life could be like with artificial intelligence in 50 to 100 years and offer our viewpoints on what we think AI should be solving so to start off with Ellie I've heard you used AI in your work can you tell us about
that yeah I'm going to out myself now for being a terrible journalist and um using chat GPT to write questions for interviews I think often you don't have much to go off necessarily unless it's like a paper that's been published or stuff but the ones I've been doing recently I literally just had a topic and to dive into an interview with one person on that topic is quite hard so I put that topic in chat GPT and I said write me 12 interview questions it's sightful and interesting interview
questions just to see what it would come up with but actually some of them were good I mean I didn't use all of them some of them were nonsense but most of them work well and even one of my interviewees complimented me on my questions and I didn't have the heart to tell her I used AI to write them for me um but actually I think if I'd gone to a colleague and said I'm struggling with these questions can you help me what's the difference in going to Ai and saying please can you help me write these
questions so maybe shouldn't feel guilty but I do feel like I didn't use my own brain power but I still thought of going to the AI so I got the job done what more do they want for me personally I think there's an art to speaking with AI or or you know search engines in the right language to get the results that you want because you can't just put absolute nonsense in because it still has to be able to understand what you think you're trying to say and then you still did a bit of um validation it
afterwards you didn't just kind of print out the AI chat and then read that straight off I definitely finesse them so you've used it sort of like as a as a gathering ideas or mind mapping type of tool yeah I would say so definitely is a tool just to help me so I feel like in that way it's not anything worse than like I don't know using Trello or Google Docs I suppose yeah or Google Search right it's still a powerful tool Jasmine you don't write articles but you doite
write Journal papers what's your interest in this so I really like Sci-Fi movies and AI is pretty common in most of the Sci-Fi things that I've really loved so like Star Wars Marvel movies The Matrix you got agent Smith Terminator AI you also got Baymax who I love as well as like w-e if you like cast your mind way way back to the Jackson you had Rosie the robot so like some of my favorite characters and some of my favorite franchises or like series are AI so I just like have a general
interest in them but also I also use AI but not for like writing uh my journal papers but mostly so like Siri on my phone if I'm too lazy to do stuff I'll just tell Siri to do something us usually to set a timer I do that all the time especially cooking yeah and also like Netflix I'll just let the algorithm recommend stuff to me cuz I hope that it knows what I like to watch enough I realized that actually Siri and Alexa are a I don't know what I thought they were but I'm just glad they wen't
it's just part of everyday life now yeah it's not like some fantasy all knowing creature not even creature creature is the wrong word but all knowing being it is just Siri yeah before we like go right down that path let's talk about what is AI I mean we've kind of discussed it in some other episodes with biom medicine and Medicine itself um but do either of you want to explain what artificial intelligence is yeah I can give a crack at it so I actually do know a few people who work with AI so I
picked their brains a little bit and interesting enough they gave slightly different answers generally the consensus is that AI is basically a computer doing something that a human would normally do but there's like different aspects such as like how would you define intelligence cuz obviously it's a computer doing something that human would do but like what would you classify as intelligent the goal poost is always moving in the 60s when they were already talking about AI what they
would consider intelligence really different to what we consider now one of the things about AI is that it's always developing but generally what it is is basically you replace people with coding or algorithms is generally what it is software right it's like code an engineer has written the code for it yeah you have an engineer that's written a code that executes a function and the function has to be intelligent I say intelligent in quotation marks one of the early
examples of something that is a that most people probably don't realize is a is a computer chess so I was reading an article about what's the difference between you know some older AI or computers to today's Ai and they sort of try to Define it as the first computers that was playing chess that beat humans had learned all the possible options uh of each move and then optimized for which one was the best move whereas today they are learning off winning games and rather knowing the rules of
Chess they've just determined if I do this this equals winning more often based on this training data yeah I mean it's the same approach so they're learning but like for different routs so originally it was like you give them all the data whereas nowadays it's like you don't need all the DAT you just need like a subsample and hopefully it's enough that you would like figure out how to learn how to not how learn how to win it's like its own intelligence right
so it's not just you fed it you've told it all the rules and then it can work out you it's more than that is that it can learn almost independently once it knows some of it it can then work out more and more and more yeah that would be the main difference between AI Now versus AI like in the 80s and 9s we've talked about Siri and we've not talked about oh well Ellie sort of alluded to what she used which was chat GPT and putting those aside we found some other
interesting uses for AI were there any particularly got your attention I've got a good one from uh work so they had for the first time in astronomy um built an AI called the bright transient survey bot and it's recently discovered the first ever Supernova so it's learned what a supernova is it's found it in I guess lot of space data um and it's classified it all by itself so I thought that was quite nice that it was like basically taken the job of a person worked out
whether this mystery space object that they found what it is and applied the like I guess criteria for what is a supernova to all this data which I thought was cool yeah that's a really really cool like sciency application of AI I think for me it's going to be really boring I recently like took over like doing all the bills and stuff for my flat I spent a lot of time talking to a bot oh gosh according to a survey about 7% of people trust chatbots when talking to them so um area for
improvement I I don't know if I would trust a chatbot but I might be willing to use a chatbot to get to the result that I want to get to sometimes it's stressful I think for me personally it's because on one at least on one of the chat Bots I was talking to one of the options that they gave me wasn't my issue basically Asos just sent me two pairs of trousers by accident and I wanted to return one so I was like okay okay I can't return once I'll just keep the both of them
then yeah I guess the problem is everything it's learned right it's learned like how to deal with returns how to deal with refunds but it doesn't know what it doesn't know right so it doesn't know unexpected situations or yeah if your trousers had turned up and they were ripped maybe it wouldn't have known what to do about that well that's something interesting about AI is that it is really good at picking up information and spotting patterns like in a supernova but I think that's kind
of the test of intelligence is is if it can figure out what to do if that situation didn't happen in its training data such as someone receiving two of a thing instead of wanting to just simply return uh or issue a complaint about a thing and I thought there was a really interesting YouTube video on this answering progress made their own Ai and tested it through different intelligence tests and found in some situations it was really good and in some it was really poor and it wasn't really like
middling it was just either excellent or terrible yeah that's the thing with AI cuz like it only knows what it knows and even the ones that are like kind of more intelligent I say more intelligent like that kind of displayed the ability to learn like they're still restricted to what they do and don't know if we that's where we are today where do you think we could be in the future in terms of like how AI develops and progresses like a large part of it unfortunately is due to
like where the money is because it costs money to develop these things so generally the places where you would likely see advances due to AI are primarily the ones where you can make money I think also it needs like the way it works to my understanding at least is that it needs like defined parameters yeah so like with the Supernova classification if you've got loads of data that you can feed it and it you can tell it look for XY Z and then it can just do like a huge amount in a very
short space of time so I can see it becoming very popular for like data analysis or you know the stock market or something like that you could train it on predicting fluctuations or something like that where you've got loads of historical data that you can feed it and then it can learn what's most likely to happen or for like climate modeling or something like that maybe I would hope not because then I would be employed sorry jasine I've just taught you out of a job yeah thanks there like
are some things that AI could assist with in environmental sector a lot of it is trying to like run different scenarios on what could happen and right now ai is still not that good at thinking about what could happen the situation with me getting two trials is like that's not something that Asos or whoever was programming the AI thought could have happened I wouldn't say it's like AI will take over but I think it'll be more of a collaboration or we'll be using AI more often in the work that we
do yeah I think that's definitely very very likely it will be more a part of our everyday lives some people are really crazy about these home robots so I would personally would really like a robot vacuum cleaner but that's not AI right uh some of them do count as AI cuz they have to have like the smart sensors to detect like corners and stuff but they always do the same or a randomized direction it's it's pre-programmed I wouldn't really call that intelligent I
don't know I just like read some stuff and like well have like more AI robots doing tasks that humans would do like cleaning and Mowing and vacuum so don't know if that you can make arguments on whether or not some of those would count as AI but um Healthcare is definitely an area where they're seeing uh improvements due to AI yeah they had that amazing um story about the pastries did you hear this sorry no I don't know if this is true I've seen it on like Reddit or something but basically there
was a a bakery shop and they had like so many kinds of pastries that all look very similar that was taking ages to train new staff as to like what each pastry was so this guy built an AI like took pictures of all the pastries told it what all the pastries were so then it could really quickly if you were buying you know 50 pastries it knew one was a donut one was a bear claw one was a Pano shocker and it could do them all really fast and then I'm not sure quite how
this happened but this guy like learned about this pastry Ai and he learned that the similar model could then be applied to loads of pictures of tumors and he could run it through all these pictures of tumors and it was like the same basic core programming and it worked out which ones were benign and which ones were like malignant or something different that's a really good one cuz also like a computer can like process data much quicker than a human especially when you have to do the
manual U visual comparison stuff yeah like another area where we see like changes due to a would definitely be like policing we've kind of seen it in the news not too long ago but like facial recognition and that kind of terrify some people that is quite scary what from like a privacy point of view yeah I guess if you think back what life was like 50 years ago or even like 100 years ago people had specific jobs just for typing and then you know typists would eventually become a really
standard skill for everyone like everyone today knows how to type whereas that wasn't the case hundred years ago so if we have ai it's always around what would jobs be like in the future like yeah you're worried about losing a job or could it be we don't need jobs because the AI has solved all of our problems you know we don't need to work to get fed I mean like in Wally where all the humans they just go around in those robotic chairs I like the idea that an
AI could edit this podcast and then our lovely producer Laura could have a lot more free time she'd be be very grateful yeah but actually AIS and editing are a huge thing I use Photoshop quite often and the advances in AI in Photoshop and in like uh premere like creating stuff from nothing is crazy draw me a picture of a cat and it will just produce an image have you guys seen the photos that someone used AI to generate what different celebrities will look like if
they worked workingclass jobs so one of them was Snoop Dog as a barista Ronaldo the football as an Uber driver amazing well I've also seen like AI generating images in the style of a certain artist yeah you come up with that whole issue of did they create that themselves or were they inspired and therefore should people be paid for how the AO was trained it's like that guy won a photography competition didn't he because he entered with an AI created image and he won he didn't disclose that
it was AI generated did until after we received the prize but like did he make that image did the AI make that image was it a ripoff of someone else and also is it actually a photograph yeah so I think that's interesting and also like the copyright that's why I don't use it at work because we don't know where they come from and you can't like attribute it to who was right image credit this that and the other but if it's not a person taking the image or even you know
a graphic designer saying I created this then who do I credit it to and where is it getting that information to be like you know create an image in the style of pop art or in the style of yeah that's a good point cuz like an issue that I have with some of the um AI generators that create art is like cuz they're they only fed certain data so when you like do stuff like what would this Disney character or like this historical figure look like in real life like they all
kind of like look very similar cuz they've all have like specific beauty standards already programmed in that is just what is put out there by like the Western media so that's one thing I don't really like about AI right now but it could change the internet is biased right so if it's learning on everything that's ever been created it's yeah but AI is supposed to be less biased than humans one supposed to be one of the arguments of towards using it in decision making is like supposed to be
less biased I don't know how they've got to that conclusion cuz surely it's just it's trained by people right and people are inherently biased yeah it's very good point depends on how it's how it's trained but imagine if it it was running like a budget and it didn't you know it couldn't be bribed it couldn't be corrupted it just made the best decision imagine if AI was used to decide whether or not we should like go forward of hs2 or just scrap it it would have gone with
the experts opinions wouldn't it yeah I'd hope so yeah I suppose I guess then it gets rid of the like political of it rather than like is this the best thing for the train Network to be more efficient it would probably say yes but it probably wouldn't take it into account all the nuances of actually building a train Network across the country I don't know if we've trained it enough wouldn't it have say the training data we feed it all the experts opinions then why couldn't it decide based on how
difficult it is to build how much environmental impact it has maybe maybe we should try it one of us needs to like get really good at coding and then actually make AI to do this kind of decision making applications but I also thought what if it's so almost like clinical it doesn't take into account like a smaller minority getting hugely affected by its decision whereas like a human might feel oh yeah you know having to lose your home is actually a really big thing
whereas the AI doesn't see that it says oh well if you move house 5 miles you will benefit from this whereas you know humans are sentimental I think that was the point I was trying to make is that becomes very binary then doesn't it like oh you can just move house but it doesn't take into account the effect of actually doing that I think there are people who are trying to like code computers to be more human there is AI that tries to replicate humans so like Instagram have recently launched some
chat Bots that are like supposed to be very like similar to humans so basically they paid a bunch of celebrities lots of money for their likeness and you can talk to these celebrities but they're not actually the celebrities oh that is creepy there is a def a market Forex I know in Japan like men have virtual girlfriends so they just like don't socialize or leave their house and it's causing a big issue it's like that film um her her yeah when then at the end he
realizes this is Major Spoilers if you not seen it but he realizes everyone is doing the same thing it's so sad like AI that Steven spilberg movie I couldn't watch it as a kid it really freaked me out they're super human like those AI robots that are played by human actors do you think AI robots are going to be a thing because I think of AI and I think chat GPT I think they're like image Generations I don't think robot yeah people have seen all these movies with
like robot Ai and and there's definitely people going to try to want to replicate that but you think that's going to be widespread because like it's going to take so much more to be able to try and fit it into a package I mean and then again mobile phones mobile phones are a case where they're as good as a computer used to be and now it's way way smaller and computers used to be the size of a room it's going to be the like AI version of like a old noia brick phone that's what
people are going to be laughing at in like a hundred years or have you seen like tricorders in Star Trek they just seem to do everything I want one I have no idea what they even do they just like oh it tells me this it's like it's the same picture as it was a few minutes ago and it was doing some complete other job oh I think that's called a PL device rather than I don't think you can blame that on the AI I think you can blame that on the set designer the show Runners whoever
decision it was to have that but I do get what you mean though like I remember watching like sci-fi shows and they used to have like tablet computers and I thought that was incredible and I could not imagine being able to like walk away from my giant home PC that was plugged in that you couldn't use the phone if the internet was on and they had they were just walking around putting in data into their tablet computers and now we have that so maybe that is you know the
way it's going is that we will have walking talking moving AIS I still think we're going to keep AI not corporeal incorporeal otherwise it's going to take over and kill us all yeah like in many Cy movies maybe maybe that would be really scary wouldn't it turns into like iroot or the actual Matrix what would you think would happen if we did teach the AI based on sci-fi though I think it would turn out a a bunch of crazy stuff it needs to be able to do all the software stuff first right before we
give it a body I feel like giving it a body is a bad idea I don't want it to be the same size as me CU surely at some point it's going to be like I don't need her pretty sure there are people trying to make robot companions for people but do you think in another way if we programmed it well enough that it would teach people to like rather then it being a crutch that it actually gets them out into the world yeah potentially I mean educ is an area where there is
like interest in using more AI maybe if like being sociable or like improving your skills is one of the things that it's coded to do cuz generally with what is available out there for that kind of application they're more for like saying what you want to hear rather than like what you need to because they've got like a commercial interest yeah commercial interest and also like basically what you described as therapy to a certain extent like how do you program an AI therapist I kind of
thought of it more as a like a coach maybe yeah but how do you program an AI life coach do you think we could see like sport coaches because sport is could be put down to like rules and like easily programmable features you know the ball goes in the goal that's good that's winning so then you could pile all that in and then you could have instead of football managers you'd have an AI that would know like you know if you do this series of moves you're more likely to score a goal I think that
could work because like is an area where they already they are already trying to like make it like really scientific with like analyzing every single minute detail so I think it could work it's like all about the angles it's all about the speed it's all about which where you kick the ball from which part of your foot the athlete might get really annoyed by all the criticism but I think it could work yeah have you played the F1 game I'm terrible at that I'm so bad
at driving games I'm really bad but you have the engineer over over the radio telling you you are slowing Cor Corner 1 2 3 4 7 8 9 and it's just like I'm slowing all of them except for like two thank you that was really annoying I didn't need that trying to corner a different one now so like you're probably aware of it while while you're driving like yeah I'm slow I was always in the gravel yeah me too can you imagine if we had wearable AI like a smartwatch well I
mean like the Google Glass or whatever other company did the glasses and and you can see everything and picture this I'm going to pitch this to Ellie you're you're filming for wildlife documentary and you see some creatures or you see some movement but you don't know if this is the same creature that you've been tracking this whole months and in said Boop AI assistant says it's a new species you've never seen a species before amazing I'm sold yeah give me those glasses I thought you were going
to tell me it was the same individual and I was going to be really happy about that but if it can recognize new species straight off the bat then let's feed it all the animals and see if it knows sounds like a hybrid of a pokedex and a Google Glasses I want one imagine that for like conservationists though like instead of tagging them physically and you can actually instead track individuals you can just look into the distance and then I'm imagining you know like where you
see it like holographics almost and then like the little tags come up and say this is individual three 58 this is individual 361 yeah be perfect save a lot of time as well as gists of the future you also use it for plant yes yeah that's true can't forget about B even though I always do but we see that in drive in self-driving cars or even normal cars they kind of tag things I don't know if you've seen some of the um particularly high-end cars they have cameras all over
and some of them are trying to park and they've worked out that is the parking spot that's another car that's a pest pedestrian I would love a car that can park for me that is the most practical use of AI anyone's ever thought of I've been in the car and it has like the South Park Park feature and I was terrified when it was parallel parking I'd get out press the button open the door get out it can park the car I won't even watch I'll just let it do its thing what
if it reverses into you oh no it's clever enough right surely if we train it how to park it's not going to crash into the driver that's like that's Ru one one in theory it has all the cameras and sensors so it should know if it's like something is in this path yeah exactly otherwise it's going to be rubbish and it's going to crash into all the cars so there was that one self-driving car that did crash into a woman maybe I'll give it 20 years and then I'll get one what if you just
happen to wear green screen trousers though and it just couldn't differentiate you is that how Green Screen Works it's just like oh there's this like green area that I cannot see but like rest of it looks fine and it has a torso of a human I mean it depends what it's what its sensors are based on you know if it's if it's light and it just sees and Ellie just Blends into the background I don't like how you've come up with ways for my new self-parking car to crush me to
death it's only as good as its constituent Parts I guess right well you two are ban from training it or building it or as people always saying data rubbish info in you get rubbish out I can't believe they can self-drive like that is AI then I don't think it is AI yet I think it's just a bunch of sensors no I think it is how do they know what the speed limit is that's a sense of reading a thing yeah cuz they also like drive and they can like adjust to speed and like know how to flow blow
the traffic and the like emergency break thing cuz my friend's car has that if you get too close and you're not breaking it breaks for you but is that AI or is that just a sensor of like distance maybe I'm just like assuming it takes a similar decision making process that I would and say oh what what distance do I think we are and then you know but it knows all the science and maths behind it whereas I just go I know from experience if I start breaking now
I will be okay whereas it probably knows how much force to apply or is it doing the mass really quickly and working it out based on velocity because if there's a bunch of calculations that's just someone's programming yeah rather than it's learn it itself yeah but AI is still by definition just like a computer doing something that human would nor would normally do based on that definition like Ops maybe tweet this out yeah self-driving cars AI or not maybe there are some Ai and some that
aren't yet yeah there are ones that definitely do self drive cuz in San Francisco they're trialing self-driving taxis so they're like driverless taxis oh I watched a video of that and the lady was like like ordering an Uber but you get there and you don't have an Uber driver you just have the car and then you get get in creepy quite creepy great if you hate talking to your taxi good for the socially anxious yes but then I just feel like it would go wrong because you have no control can
you override it can you then like get in the front seat and be like I don't like this I want to dve yeah you should be able to that's the thing with a lot of like the robotic ones is that you should be able there should be like a human override what if you get trapped right you get to your destination but it doesn't let you out that would be terrifying guess you'd have to break out the window I hope there's a man override cuz I feel like you remember when saavs
first became really big and people were like not really concentrating and just following the satinav blindly and then driving into lakes and rivers yeah like I feel like there's going to be a teething process with self-driving cars do people trust a self-driving car not the moment no but maybe in the future and will people trust chat Bots more who knows probably not but we're probably going to see more and more of them in the future but is that just going to be
like annoying customer service like instead of getting stuck on a long call to someone going through all the buttons it's just going to be the chatbot version of that yeah but sometimes it's better to just be on hold for a while and get a human to do it cuz sometimes those chat Bots are annoying cuz they're like yeah solved your issue bye I'm like no you didn't solve that issue but yeah talking about trust I saw a really interesting well a couple of Articles where they survey people who work in
machine learning and artificial intelligence and they asked them do you think they will kill kill Humanity this is another good reason not to give AI Corporal bodies yeah let's not do that for a while because at some point it becomes more intelligent than its creator isn't that the idea that like all these Sci-Fi movies are based on like it becomes too yeah that's like the premises in most of the Sci-Fi movies is that the computer is as or is more intelligent or thinks it's more
intelligent than the human but then the human somehow manages to win so what does that where does that leave us as species or as a society if we do have machines smarter than us does that mean we are we in a are we going to be in a workless Utopia like in Wally I love that film that there's an author who just writes all about this but it's obviously not all a Utopia I don't know that one enm Banks yeah so I M Banks has written sort of like just books and books about a
utopian world where people don't have to work I mean there's loads of different places and scenarios in one of them they they just play games all day but to the point that they started making deaf games like Hunger Games because they got so bored oh that took a turn yeah I suppose if the idea is that eventually AI will take off a lot of those like data processing manual jobs then there would be more free time right more or or are those people then going to become
very poor if they you know they've been replaced what they're going to do next so I think for like some jobs still be would be humans but I think for others like there's definitely going to be a phasing out of humans in some sectors like for take for example now like podcasting like this wasn't really a job I say that like we get paid for this like podcasting and YouTube and infl waser they were not jobs 10 years ago so I think we're going to see see an evolution as the technology progresses
but like what it could be I mean anything could happen Twitter is now called X and that's that's your measure of program yes yes I would like to see uh an AI go up against like 30 school kids like if an AI can be a teacher against like 30 teenagers and like get them to do what they're supposed to be doing or will it go so far the other way that we won't even have classrooms and everyone will just be learning online and there'll be no point in a teacher in that sense or
do we even need to learn if the information is available to us we just ask them with a we just ask the AI the question and then we have the answer in which case wouldn't that be making humans dummer because we're just lazier yeah that's what I reckon if you just always rely on you know having yeah cuz I was reading something that's said that like humans nowadays in comparison to how we were like 100 years ago or even longer like our memory is way worse cuz like we don't need to remember as
much stuff cuz we can just like set reminders or put stuff on our phone or actually write it things down but like people 100 years ago that wasn't an option they had to remember stuff but then that doesn't that enable us to do other things if we're not just trying to remember but but but yeah but how do we know it's actually useful stuff because it's just like playing games or making games I wouldn't say that was useful you freed up part of your brain right you
that knowledge that you would have to remember people's birthdays or remember to phone the post office you've you know outsourced that to an AI now what's that part of your brain doing is it just sitting on the sofa watching Netflix or is it you know working on something else maybe it'll be like Star Trek where we have so much time that we will just work if we want to or not work and we will be looked after anyway so I think we've explored what we can with AI as we
know it today in October 2023 Maybe will check in and just see how much further things have gone you know maybe we will trust web chat to sort out two pair a duplicate pair of trousers yes so thanks for listening the views expressed in this podcast belong entirely to the person that said them they do not represent any industry or organization if you enjoyed listening to these views it would really help us out if you could rate US leave a review and tell a friend this podcast
was sponsored by no one but if you're interested in funding us to continue to have Frank discussions about science and engineering please get in [Music] touch