Episode 10
Artificial Intelligence (part 2)
About this episode
Welcome back to the second part of the two-part series of Untangling AI episodes, where hosts Alec and Jack continue their exploration of the intersection between Artificial Intelligence (AI) and Web3 technology. In this episode, they delve into the transformative world of AI and how Web3 can ensure and even empower its safe evolution.
We begin our discussion with some of the dangers associated with AI, as highlighted by Geoffrey Hinton (the Godfather of AI), exploring concerns regarding AI abuse, malicious use, "technological unemployment," and the potential risks of Artificial General Intelligence (AGI).
Moving on to the intersection of Web3 and AI, Alec and Jack highlight how AI can use cryptocurrency to purchase goods and services autonomously, thanks to the capabilities of Web3's decentralized financial systems. Our hosts emphasize the role of micropayments in funding AI services, providing incentives for ethical AI use, and revolutionizing the way AI is monetized and utilized.
In the context of AI explainability, Alec and Jack discuss how Web3 can ensure good or transparent data sets and models, preventing data poisoning and biased AI outcomes. They showcase the transformative potential of AI in content generation, underlining how bots belong to the Web2 era, whereas AI in Web3 can create useful and authentic content, empowered by decentralized identity (DID) and authenticity.
An exciting use case of AI is explored, where AI is trained on data that does not reveal any privacy-sensitive information for healthcare applications. Together, we dive deeper into the groundbreaking potential of AI in medical diagnoses, exemplified by IBM Watson's NLP technology, and emphasize the crucial role of Web3 in maintaining privacy, security, and data control in AI-powered healthcare systems.
The episode concludes with an optimistic vision of the future, where Web3 and AI collaborate to create a more inclusive and empowering technological landscape. Alec and Jack leave the audience with a sense of hope and excitement for the potential of Web3 and AI to shape a brighter and more promising future for society.
Tune in to this second part of the Untangling AI episode to discover the power of Web3 in securing AI, and the transformative possibilities that lie at the intersection of AI and Web3. Witness the birth of a revolutionary era, where AI's immense capabilities are harnessed responsibly and ethically within the decentralized Web3 environment, paving the way for a future of innovation and empowerment.
Got a question or topic suggestion? Send us an email at theuntanglingweb3podcast@gmail.com
Transcript
Auto-generated from the episode audio.
welcome to the untangling web 3 podcast your goto Hub to learn insights and the latest developments in the wild and wonderful world of web 3 I'm Alec Burns and I'm Jack Davis tune in each week as we navigate and explore the rapidly emerging landscape of the web 3 Technologies projects and ideas that are shaping the future of the internet we'll be talking to the best and brightest in the industry to keep on covering insights so that hopefully we can all learn together on our journey to untangle web welcome to the untangling web3 podcast this is the second part of our conversation on AI and web 3 in the first part we covered the history of AI some introductory Concepts and in this episode we talk about ai's overlap with web 3 some potential use cases and challenges of both we've we've talked about the content generation side explainability yeah you've got another one to talk about no no yeah it's kind of linked to it like it's we've talked about content generation and authenticity like
that that's a really important thing and you were talking about like micro payments and like I think that payments in general is going to be a really exciting way in which web 3 and AI interact in the future like I you know I don't want to excite the crypto Bros who who are potentially listening but I I do believe that AI right now is needs a payment rail right like have you you know you tried to sign up to a bank account how hard is that like the kyc involved is very difficult like there's no way we're going to have ai using traditional Fiat or you know traditional banking anytime soon but imagine like the possible use cases if AI could transact if AI could make payments on its behalf like it it makes so much sense for AI to have some kind of purchasing power like you imagine a world in which you just tell your AI model okay I want to create you know the next best podcast and the AI goes out
there and it you know it starts to you pay for a streamyard account it starts to pay for figma account it starts to generate content but for all this to happen it needs to have payment rails right and it's not just payment rails these payment rails equal execution rails right it needs this to interact on the web and start to procure services to do things and I think right now we have like a lot of human interaction you prompt chat TPT it gives you something back and then you can go action something I really believe that we're not too far away from a future where you tell AI to do something like go make a website and it just does everything it pays everyone it hires humans non-humans other AIS and everything all the interaction is covered and I think the only way to get that is through web 3 based payment rails yeah I mean I can see I can again see a can of wor being opened and some prob with that if
you start letting AI do all your subscriptions for you but I mean I mean there are examples of that I think where there AI tools that will look at your if you give them access to your your banking app and things they'll try and cut out unnecessary subscriptions things you you pay for look like budgeting apps type thing and also we mentioned you know uh I think the finance industry is already is it's probably way ahead in terms of giving AI money to do things with they'll give them they you know have I actually saw I saw someone I can't remember their name now I'll have a look after the episode but there is there is a fund manager who was had quite a good track record in America and he completely moved over to artificial intelligence about a year or two ago like he he completely lets his all his trade execution all his portfolio management is done purely by computers so he just gives it the money and lets it Go and that's
you know that that does need payment rails as you say it's an example of like letting AI just loose with money but I I agree like if you wanted to let that AI interact in the real world outside of the the internal kind of trading platform that they're using then yeah payment rails you'll need the other thing there will be because we just talked about micro payments right and your you're coming at it from the AI being able to pay people I think the other way around is also really interesting so using micro payments to pay for AI services like we saw what happened with a couple of months ago mid mid Journey was one of the big alongside chat GPT mid Journey completely took off uh for those who don't know mid journey is the is a is an AI model that it's a generative AI model that generates new images based on text prompts and that's the one you see with all you know when when we talk about those images being created
earlier that was mid Journey now they opened up a free trial for a good couple of months and it blew up they got so inundated that they had to completely stop the free trial so I did use it once just to play with it and then I I got locked out and then now they put it did you Jack I tried to generate a a logo for this podcast but it didn't go that well for me I'm not I'm not prompt engineer yet but um the the fact that they had to go and cancel the free subscription so you know close off this opportunity for many people and move to a subscription model for for however much a month I I just thought that was a real missed opportunity for for web 3 to be used because you could say well if instead of charging 20 pounds off the bat to use it each month let's say why can't I just pay 10 cents for generating one one of these images and that might have
been a much better model to do it with yeah I think we've talked about that in the previous episode right it's like so I pay for chat GB like I said I use it for all aspects of my life and I pay I think 20 pound a month or something like that and I don't think I actually get that much usage out of it have like unlimited prompts and things like that and it would make so you said the opposite this episode Alec I'm sure you can't go back on that now my mom's birthday is once a year so it's not like I'm using it all the time like I think yeah so you are right there's some people who are probably using it a thousand times a day I have no idea who those people are wouldn't want to interact with them but yeah there's like a there's a bandwidth there right and I think it would be a much fairer model if you were paying per prompt because that really is like what
the kind of the the burden is on on on open AI to maintain this rather than just a flat subscription for everyone and you reduce the barrier to entry there as well like you said yeah exactly well and it's also I mean at this point is worth mentioning probably and acknowledging you know why did you say earlier that AI uh is quite centralized at the minute it's these big companies it's because it's so expensive to run so you know chat gbt had I have I have the figure um I have the figure somewhere it's uh I think it's something like a trillion parameters in their model and it requires gigabytes of data to to uh to train but yeah it's um it's unbelievable the size of these these actual machines parameters my God yeah I I did have the figure somewhere I have to have to double check but it was it was something like a trillion parameters and to actually you know it's expensive there's big um uh GPU costs there's big uh server
cost required I mean even just to deal with the demand for the usage let alone because it's taken off so much I mean chat gbt got something like I think it was like four or five days it took to get to a million users and then had about 100 million within two months so the demand on the systems just to run what they're doing is insane and they have to fund it some way so you know if you're doing micr payments per call that might make a lot more sense yeah I completely get it I'm so for like I mean it works both ways right like humans like I I see in like the metaverse of the future we're seeing gaming applications coming out right now you know non-player characters in in big gaming environments that they're pretty crappy or they have been pretty crappy up to now and I think with these large language processes right we're going to have really exciting user experiences with these AI generated kind of MPCs and I think
you know we should pay for them even if it's a micro payment and even if you just send like an AI into this this world and its entire goal is to just make money and it says okay the way I make money is to go into this world and be the best damn NP DC in this world and everyone will tip me and monetize me that's its goal it's creating value if people want to give it you know micropayments want to give it money for the value that it's creating there it makes them feel happier it de-stresses them then it should be able to and I think these kind of you know these payment rails that we're talking about in web 3 are the perfect way to facilitate that you know what you're making me think of a really nice convergence of a lot of what we've talked about recently on the podcast with metaverse and things because I can now I can now envisage being in a game or like in a metaverse environment
where you have these PCS and you know you always have like resources in a game you'll have certain items that you pick up like normally they're just coins they're like gold coins and you pick them up and you can you know do what do whatever but what if you could use those coins to interact to pay for interacting with the AI right and and get them to do something maybe you could get it to generate a new quest for you you know and you have to pay for that and the cool thing there would be every time you talk to the to the to the uh to the AI it would you could prompt it so you could say generate me a quest that takes me to this land and I have to speak to certain people and do certain things and it would generate a fun story around that so it wouldn't just be like picking from a preset list of things it can do it could generate on the Fly task for you
and you could pay with the crypto it's going to be like that world really excites me like it gets my I get really like gaming excited about this like I was playing D and D the other day which you know Dungeons and Dragons um with some friends for those of us with a exactly um and I was like I had this kind of rough story that I was going to do and I was like I just plugged it into chat gbt I was like make this better like make it more fun make it more exciting make it darker and my God the story it came up with I know that it's just aggregating lots of stuff out there but it was the fact that within three seconds I had this fully immersive story like if I was playing the game and there was an AI in there that was capable of this yeah take my money like I would love to do that and I think everyone wins there I don't see what what yeah
like it it just make perfect sense right this culmination of kind of payment rails in games AI generating content and I think this is one of the the big things that we are going to see right now we think of bots in web to a spam and I think we're going to see I mean yeah there's a lot of potential negatives like the disinformation the Deep fakes that we've talked about about how AI could cause problems in the future but I think there's a lot of benefits that are going to come and I think web 3 is really gon to you know positively affect how we interact with AI going forward I think in a way it's absolutely essential for it as well yeah I I agree I mean there are there as we've said there's so many opportunities for AI and we want to realize those as safely and you know as ethically as possible there's a lot of talk about e the ethics of AI right now I mean remember there was that
letter going around that Elon and a bunch of people signed to say you know we need to stop the training now because we need to think about the ethics of this because of all the types of things we we've talked about and I I I I agree I really see web 3 being the the toolkit that help helps us to use AI productively in the future and minimize Bad actors I mean talking about Bad actors right that's a big that's a big worry I mean um I don't know if you saw but this is uh man called Jeffrey Hinton who is often called The Godfather of deep learning and so he he you know he invented so many of these things he won the touring prize uh a few years ago I think 2018 because of his contribution to the field and he was at Google he was very high up um in the AI department or or something he quit he quit Google recently because he he was like I need to you know
I know I've helped create this stuff he's got like the Oppenheimer story right he's like I helped create this stuff but now I need to leave Google so I can speak freely about about the risks to it and and he cites things like abuse of the system and and uh you know if you use them maliciously another things what kind of stuff Jack like what does it mean by abuse of the system can you explain that yeah I mean there there are a number of ways so as we've said one of them an obvious one would be just using it it's lower the barrier entry to create spam that's one example of abuse maybe and and maybe you know take things like cambage analytica further influencing elections Etc deep fakes all this misinformation stuff that's kind of thing exactly yeah but there are there are ways that there are things that AI is uniquely um uniquely susceptible to because of you know we described it's so dependent on the quality the data you put into
it right it's so dependent on that being how you train the model and I I have to I have to double check but I'm pretty sure I'd be very surprised if it's not the case but I'm pretty sure I read that it is that chat gbt and and similar large language models when you use them they're constantly being trained they're constantly improving so it's not like they just they just it's a onshot they just build the system they Let it Loose no every time you use it what you put in it's learning from and I'm 100% sure because you canate value exactly it's value up and down like every response is positive and negative it will improve for everyone based on that yeah so when you click you know was this a good response you're helping train the algorithm and the dark side of that is people could start poisoning that data set they could start putting in things you know telling the algorithm it's wrong when it's not you know starting to influence the
model itself now I don't know how practical all of that is but the fact that people like Jeffrey Hinton are worried about um about abusing AI in that way I think speaks volumes got you I think when I thought you were going to go towards then was like training bias and I think maybe you have kind of alluded to it it's like okay we were talking earlier about the the voice recognition right and how certain dialects in the UK probably weren't recognized because the data set you know didn't cover Wales the Black Country and this kind of stuff and I think you know realistically humans have bias as well and I think that the difference is that in AI it's po we have the potential to actually read in and provably say what is that bias I think the point is because it's not human we kind of think oh it's completely neutral there's no bias in that but we have to realize like it is trained by humans and there's always going to be
bias there and I think that the potential the beauty of this is that we do have auditability into that we can say okay you know this seems to be biased is it biased let me look at the data set let me read into it using web three using the principles of blockchain why is it biased okay that's why it's biased I think that's something we have in humans as well that we need to rec like there is the potential to be bias there but that's a lot more difficult to kind of Define right yeah and I 100% agree and that that I'm glad you mentioned the bias point that is that is that is important because you know we we go back to this idea that AI doesn't exist in in isolation it is created it is being trained there are people hand labeling the data set so they people will be looking at outputs and and you know not just you and you use it but people are paid to rate the responses of
these things or to uh to check if certain um outputs of other models are correct um and and it's worth noting that you know how far do these systems go they can go much much further do much more powerful things and we live in a global world to be very incredibly cliche and as much as we're talking about being ethical with AI and there's you know Europe in particular and America are talking about regulations to AI I think quite rightly another danger and another abuse Vector for it is that not everyone will necessarily subscribe to the same view people you know if if if a lot of the world self-regulates there will always be other countries or subsets of the world that don't and then can you know Go full full force and using it for malicious purposes which is I think one of the things they're worried about well I'm so glad you mentioned this because like we're saying um there's who sets the limits right like this is a big problem for chat
GPT you were talking about how people are poisoning data set you know putting in bad prompts yeah like asking for hate speech how do I you know I think what someone told me that they they were like there was this example on they did on Reddit they like how do I kill the most amount of people for $1 or something like that and in a previous version of chat GPT it would answer that question and one of the big developments moving from chat GPT 3.5 to four was around the safeguards and making sure that there's a lot more ethics involved but then the the problem there is who defines the ethics like this is going to become like a commodity service like a public service I I think in the future and who are open AI to Define what is ethical and what is non ethical right you think about the politics like I've seen a lot of people where they're saying there's a massive a massive bias um against the Republicans in Che 3.5
and one of the pushes for four was to kind of equal equalize that bias right but this is so tricky like ethics are very subjective they change from country to country from person to person there's no such thing as you know Universal ethics like the un's Human Rights Act probably as close as we come but that's not accepted by every country in the world like when these he's talking about global interoperable world how do we Define ethics how do we Safeguard how do we Define standards when there's so many people like we don't agree on them like no one agrees on what like the standards should be so how can open AI Define it for everyone to use it yeah and we're getting into the realm of the philosophical now right um web 3 Technologies won't I don't think solve all these problems even you know they won't solve everything lots of them they won't even solve to a satisfactory degree potentially we just don't know yet but I still see web 3 is the
best toolkit to addressing them in at least in the short to medium term um and and providing as much of the infrastructure for yeah identity authenticity um and and you know having evidence of what happened in these systems I I I fully believe that think I need to look more into um why Jeffrey Hinton left Google like I remember seeing that news article come out and being like Oh God this is big at the only thing that I heard I think this was Pub talk basically with one of my friends that's kind of in this space and I think his he told me his big fear was around the the war aspect he was like a lot of the funding for AI is coming from the military like the US military like it it's always the way right like that they push big into robotics they're big in Dara big into spot all these kind of AI Ena robotic systems and he was like he he didn't want to work for things that could potentially
be used to kill people and like be used for militaries no matter what military is and I think I I want to know more detail about what kind of fears he has right well I recall I I do recall what the kind of key ones were so the first one was around the misinformation piece that we've already talked about you know how it will just reduce the barrier to flooding misinformation everywhere so we've kind of covered that in detail the other one which I guess surprisingly hasn't hasn't really come up I mean maybe we talked about it with digital marketing and stuff but the the potential job replacement and and the automation of task because again we're saying we're seeing things like chat WT simulate human tasks pretty well now and that's I think that's the biggest consumer worry right now yeah actually the one that's in everyone's mind is is this going to replace my job we've seen I think was it BT I saw them saying they were laying off something like 20,000
staff in the next few years oh yeah replacing it with AI and companies are going to go this way for sure this this is a huge topic like so IBM said the same they were basically said they're going to put a complete freeze on any job um that could be replaced with AI until they have worked out whether it can be replaced by AI but I think like we kind of talked about it before this happens all the time right whenever there's a new technology that everyone thinks is going to be revolutionary like the computer like industrial robotics all these things the one of the first concerns is okay how is this going to affect us economically is this going to lead to mass unemployment and it hasn't so far people adapt I think one of the fears is that the rate of change means that it's going to we're gonna be C playing catchup which I suppose they tend to do but I think more and more as we're kind of pushing Humanity away
from laborious tasks we can push them more and more towards you know tasks that potentially add more value you know I guess in my mind if we think there's a pot of money for the NHS which is the healthcare system in the UK right and say you can replace 10% of jobs with AI and save you know 99% on cost or something like this right and it's the you know administrative tasks maybe it's like the filter Gates a big problem in the NHS right now is the filter people come through and they ask to say like you know they don't really actually have any serious elements they come through they ask lots of questions like to have like an AI filter initially initially say whether you need further further kind of um checks or need further tests or you're just fine you go back like that could save a lot of cost but like I was saying before there's a pot of money if you can save some money by removing some of these jobs
and then push more money to other jobs like more nurses more like the kind of the jobs that say maybe create more value in that sense like is that not a benefit overall yeah I 100% agree and I I I I see a lot of the talk about mass unemployment now from AI as similar to the Gary kaspero moment about the death of Chess you know it doesn't it doesn't bear truth in reality and there's a very heartening statistic I don't the number exactly but it's like it's something like more than 50% of the jobs we have now didn't exist 10 20 years ago whatever the figure is but it reflects the fact that employment changes over time you know the jobs now very very different to 100 years ago so just reducing some jobs and uh you know minimizing certain types of Labor is is going to be replac with other jobs and and it's interesting because I think what's different now is it's it's changing a lot of um yeah desk based work
that is written and the UK has you know you know tons of this this is a huge part of our employment the information uh knowledge workers sorry knowledge work it's really it's really having a big impact on that and I think that's partly why people are making such a fuss about it because you know the me the Industrial Revolution replaced a whole bunch of jobs we don't have people manually uh you know doing every part of the the car manufacturing anymore and we're doing just fine I think the same happening but people it's it's it's becoming it's more in people's minds because it's affecting a different type of people now in different people in different types of jobs um which is interesting but yeah I don't I don't see it as as really changing I don't think it's going to make everyone unemployed in the next in the near future it is very interesting like I follow a lot of American politics I remember seeing this Republican candidate basically say yeah AI is going to
change the world but like his fan base was effectively like working F people and he was like but it won't change the world we'll always need you know Farmers we'll always need laborers we'll always need welders construction workers AI will never affect those jobs so you're safe and I'll make sure you're safe and you're secure and you're completely right it's like a dynamic shift this is one of the first times that like a maybe a technological advancements really come for you know Des space jobs and kind of people that would think of themselves like middle class upper middle class and all these things but yeah again I think it's just that the emphasis is going to change like I see stats it's like you know we're going to lose 150,000 jobs and then like the subtext somewhere is like but it's going to create 150,000 jobs elsewhere right I think we're just going to see more productivity overall I think like a big example of this is coding like oh my God the way chat
GPT has advanced like software development and coding is crazy like one of the guys that um I'm really good friends with software developer he's like the I can do a task now that would take me a month in basically a day by prompting chat GPT and it's not like his work's going to going away he's just way more produc Ive and I think those kind of like areas have a lot to gain from this exactly yeah that's the right word it's it's it's increasing productivity essentially I mean what it can do to automate uh you know coding and stuff like people again a different sector that's being uh targeted by this advancement in AI would be people who are just experts in a in a in a coding language potentially um who can who would just write code and don't necessarily find and Design Systems because you can ask you know uh you can ask one of these machines to say okay write me the code for a website and yeah right now you're not
going to trust chat gbt to write all your HTML and CSS for your website but I definitely think in the future it will get good enough to do that and um yeah yeah I I I think that's just another interesting Dimension to the whole thing is that is all these different types of jobs that are under threat and I think it's going to below the barrier to entry like imagine if you're a startup and you're like okay I need to hire five software developers to do this thing like create a website with all this backend stuff that's going on and that's going to cost me like 250 Grand and then all of a sudden chat TBT comes out it's like 20 pound subscription you can do it and you're like oh my God finally my dream's been realized I can create a website make a startup and I get it I get the fear and I think it's kind of the fear of the unknown is my job at risk what does my life look
like but I think as with almost everything so far like this is the best time to be alive ever period and that's because of technological advancement right and I think there's always a fear with the unknown but Humanity so far based on you know the million or well 20,000 years of human civilization or whatever it's gotten better and I do think it will get better and I think your Society is definitely going of changeed to be hugely impacted like this is going to be AI is going to be the one of the most hugely impacting things since fire potentially but I'm excited from it overall I realize there going to be a lot of problems in the same way fire can burn a house it can cook food but I think overall it will be a massive net positive Humanity yeah yeah I 100% agree um I had I had a point there I was thinking about and it's just uh I've just lost it it will come back to me um is it more
of the Jeffrey problems no what what I was going to say actually I've just remembered is what I was going to say was what another way we could describe all of this like increasing productivity is uh increasing efficiency it's making it more efficient to do things which is an incredibly web through idea as we've discussed you know it's lowering the bar to entry it's just increasing the efficiency with which you can achieve task XY z um so yeah it's it's going to make it a lot easier to do certain things and there'll be big big winners from that you know there'll be there'll be companies that there are companies being built just just to solely leverage AI I mean that word leverage in itself is is why Everyone likes software and software as a service you get high leverage from it you build something once you can can deploy it many many times AI gives you leverage to your your brain you the people who are better at using AI will do the best from
it I mean I've seen I don't know if you've seen this I'm sure you have but as soon as chat GPT came out and and GPT 4 there's a whole new terminology around prompt engineering as a job and I've seen I've seen I've seen job adverts for hundreds of thousands of pounds for prompt Engineers people who are the best at wrangling the best responses out of the AI yeah it's crazy like I it is a skill and I think I saw someone say that it's like it's the same skill as human interaction in the same way that a salesperson will be a Wordsmith they'll know exactly what words to use to make people feel more comfortable in making the sale like it's just like that like it's about how do you use the right words to get the most out of chat gbt it's quite an interesting Paradigm Shift right but I love this idea of like a prompt Wordsmith it's so interesting that people are pushing this as a career choice now chat gbt
is like a Aladdin's cave and you've got to whisper the right words into it to get the gold you know oh my God that is the nerdiest reference we've had so far and I was talking about D and D earlier as well yeah well you know um one of the problems with this like you're talking about efficiency and I find it super exciting and but I also is there a part of me that thinks okay is this going to turn dystopian you know have you seen the film Wally where humanity is completely kind of supported by Robotics and Ai and we no longer have to engage is that a Humanity of the future and that's a potential right like that's scary as we kind of offset more and more of daily tasks and and thought processes does it mean we're freed up to do more of what Humanity wants to do which is like being creative artistic all these kind of things or do we actually need to work do we actually get pleasure from
you know the grind and graft of you know the work that we're doing now and this is I think there's you know there's two separate camps I think maybe more the technology people are think it's going to be a Utopia and then more people there's some people that think it might be a dystopia where it's a you know we we're offsetting too much yeah i' I've seen both definitely I mean people like uh I think I think Naval ravikant who is kind of on the money on a lot of things he he definitely sees it going in that creative Utopia but the only jobs left will be uh the creative Pursuits you know which actually goes against what we were saying about the risks to creative people actually when everything else is automated they'll just be production of new um inspiration works of inspiration so things from actual human thought and then the other people who are who are more worried but um yeah I don't know which way it's going to go because that
brings up a whole other conversation which I don't think we'll have today about things like Universal basic income right because if there's only creative work to do then you will need to subsist in some way um but your point actually takes me very nicely and neatly onto the third and final one of Jeffrey hinton's big concerns which is about um AGI or AGI the artificial general intelligence the big kind of Spectre that some people are worried about this is this is the Skynet you know this is the singularity right the singular ity so how far does how far do they go you know when it's not just a happy helpful chatbot and when it's when do the machines take over um and you know I I don't feel qualified to have a view on this you know I feel like um I feel like it seems implausible but then so did chat gbt until recently you know like so I've had a lot of maybe Pub conversations some sober some not so sober and I
feel like I've talked about this a lot like it's this is the first time in you know maybe a million years at which human brain power may not be the most intelligent driving force in the world right there is the possibility that this AGI could come out that is way beyond our comprehension that we can't keep up with and that bit kind of scares me is it possible that is a topic that is massively Up For Debate like I don't I'm want someone who subscribe and this is getting very philosophical I'm someone to Subs that does not subscribe to the fact that human humans are particular particularly special that we have this Inc conceive incomprehensible Soul thing that makes us different right I believe that you could track our Evolution from like a single- cell organism which could very easily I think be mapped by an AI to you know something as complicated as an ant which I think could be M by AI to something as complicated as a dog like which could probably
be tracked by AI why is there something that's all of a sudden super special it's just that were slightly more complicated slightly more difficult to map and I do subscribe to the fact that AI does have the potential to you know evolve to the singularity and that's you scary it is scary I think overall yeah I mean I think we're gonna have to re it in from the philosophical in just a moment but I I actually I think I broadly agree with you there um when you start thinking about things like free will you know that's what we're talking about we're talking about AGI like can something have free will and there's a kind of a growing movement around whether or not free will exists for humans you know there's a lot of people who will describe humans as wet robots now um and there's you know the science kind of supports that that interpretation even if like our experience in perception in terms of our own perception might um might feel different a lot
of the science actually supports that like I mean I won't repeat everything you can go and listen to Sam Harris he'll tell you all about it but I'm I'm I'm I think we're waiting too far into other topics so why don't we why super interesting yeah yeah I mean we should maybe we should just become Sam Harris's podcast one day um not that we would have any chance of that but let's just mybe close off this discussion by getting back to like the concrete use cases right we have we've littered them throughout but maybe why don't we just talk for a bit about our favorite ones like what what are you most excited about the minute favorite ones um I think the most tangible one for me is around like training AI basically and we talk when we talk about data ownership how we train Ai and what that means for privacy and I think to put it into like really tangible terms like imagine Healthcare like everyone wants the best health care available right
but when you look at the the statistics around you know when people get pulled about are you willing to give up your you know your personal private healthcare data to improve say the stand standards of health checks so a good example would be around you know cancer treatment okay so I talked earlier about how we're using AI already to scan to basically analyze scans to see okay does this person have a tumor do they have cancer or are they fine is it benign right but people don't want if they had the option most people in that instance wouldn't want to give up their personal information along with the scan even if it benefits the process overall so I think we're going to see a process where we can start to use some of the technology we've been talking about before where users own their data they own their scan right it's authenticated signed off by a doctor say yeah this is a real scan to make sure that you know the data set that we're
actually using isn't poisoned or toxic or bad as you were talking about earlier and they can disclose the details of the scan with the authentication without the personal identifying information in that okay so when we start to think about this on like a a wide scale then the AI has access to a lot of data a lot of okay this scan is good scan is bad that kind of data right the data actually cares about and none of the personal identifying information and quite importantly it's been authenticated and signed off by a trusted doctor so it knows the data is good overall right and then that allows us to create like a very you know potentially successful a model so we can start to actually get you know positive and negative outcomes from these scans in future and then when it comes to the actual testing you know this is my scan that I've now got all I have to reveal to that I model is my scan and it will tell me the probability
that this is good or bad I don't need to reveal that personal identifying information and I think in terms of training and outcomes I think this kind of like this kind of process we're describing here is going to be super beneficial it's going to make alleviate a lot of problems with you know the authenticity of data going into AI models it's going to help feed into a lot of the trust that is a big issue with how AI is done right now and it's going to lead to a lot of better overall yeah I mean I don't think I can beat Healthcare to be honest I have to completely agree with you right I I I'm fully on board with that I think is where it can have the most positive impact and benefit for society um especially with this move to I think preventative health care is like is really that's the key you know um I mean ESP people I think people are quite scared of the idea of Big Data especially with
parena of healthcare right like you don't want to be monitored your you don't want your data monitored all the time even if it means you're going to have a healthier outcome over your life if it means that you know big corporations have information about you know when you go into the toilet and these kind of things it's interesting they actually went to a conference like not too long ago where there was this guy who basically created a device where it monitors all of your urine stool samples constantly and he was like we can detect you know how healthy someone is throughout their life and the preventative outcomes are incredible there's like people don't want to sign up to it because the idea of dat like you know companies connecting that kind of data is super scary but I think we could see a paradigm shift that comes in as we say okay no person identifying information will be revealed it's purely about you know the actual you know biomedical stuff that's going in for the
outcomes and I think this could lead to like very successful Solutions yeah I I 100% agree I think that there are also just there such a diverse range of ways you can use AI in healthcare you know everything from like the triage process as you said I mean we're we're from the UK the the NHS is on its knees it's so understaffed it's not a case of replacing jobs to be honest at the minute in the NHS it's literally a case of filling a gap in in being able to help people and prevent yeah you know you know help people work out if they need to come to the hospital or go to the doctors or what you know what they need to do without having to to use these precious um preciously scarce human resources that we have and are losing the NHS and like call centers do this already right like I mean everyone is infuriated with how bad call centers are but if like why aren't healthcare services using it it's obviously
effective like it works and if you're saying I can save x amount and like you know get the right people to the right places by using this system there's obviously a benefit there yeah 100% And then you've got the whole you know things to do with um medical trials and and training data or training models with with data for for medical trials I mean what we saw with Co and how fast we had to respond to that and understand the virus um this this field of protein folding has been around for years right and using machine learning and is just being used even even even more every day to solve those kind of problems to detect you know cancer and things and I think you know where where web 3 and AI can intercept really nicely is in that I think as you said like permissioning the data to be used medical trials because if people know it's being used for a certain thing and know they can use it they can give that data
privately um people do do that right now but it's not on the scale that you need so you know you think about your your heart rate monitor your your Apple watch or something that's taking constant readings throughout the day that you're not necessarily sharing or you're not incentivized to share you can share it but you're not necessarily it's not go into the right places whereas web 3 and AI together will basically provide the incentive model to say oh no I I will happily pay I be paid to to give that data away it'll also help with things like you know then Insurance um insurance providers where then lower premiums for if if you provide data and things like that so I think yeah there are just so many applications in healthcare and I think that what you just said reminds me of that um there was that kind of flu outbreak app that Google worked on right whereas like they were trying to work out where there was flu outbreaks based on people Googling symptoms
and they were like the problem with that is the bad data they like there's a flu outbreak in every city basically every 10 minutes because there's constantly people that are saying am I ill am I ill so like it's it's kind of interesting that's come back up but it's all about the quality of the data right yeah exactly um and it's kind of even go back to the Apple watch thing right I think the the systems on there which are all AI supported now um things like fall detection so that was when I first saw Apple watch and I think it's pivoted to a health device from it pre being like oh this is a fun thing you can you know read your messages on and take calls on there was a moment where it really became like oh this is a proper Health device when they integrated the um fall detection so if if you fall over and you and it'll give you a prompt to say are you okay and if you're not
okay then it'll it'll call emergency services for you and I was like wow that's that's huge like that's that's something that everyone in the world can use especially you know older people who might fall in the home and not you know not not have anyone round that solves such a big problem for people and I know it's not the kind of AI like the the chat GPT stuff that we've been talking but um it's it's still going to be things that are supported by the underlying web three Technologies in the future I'm sure about you with blockchain how it's being used because fundamentally I think you can you can expand the pl the space of possib possibilities like the different places you're using the AI if you're using it with these balances and checks in place so as soon as you impl you know the blockchain auditing and things and and and the the evidence Trail then you can expand what you're using it for and solve you know different problems with it yeah that's
nice that's like a super nice note to end the use cases on so like what do you think the future of AI looks like with the with the web 3 aspect in mind like we've talked about a lot of things like what are the key points for you and what are the things excite you most do you think you know I think I think I've probably mentioned most I mean in AI whenever you talk about it you're out of date within 10 minutes so who knows by the time this comes out maybe the future would have been realized but I I think I'm most excited by the economic and the markets that it can open up so you know we talked about we talked about how it might replace certain jobs and I think web 3 is actually going to open up that's going to be the real key that opens up many many new jobs that we don't know about you know the micro payment economy all these new types of job types of
work types of business model that are facilitated by web 3 by micro payments I think that is going to be really interesting because it's going to create all these new types of jobs that will maybe being used with AI you might you might have micr payments to be labeling data and interacting with the AI that could be a job right to to to help improve AI I don't think that's the most exciting job um but yeah I just think the combin of AI improving efficiency in existing kind of Arcane jobs that don't that can be automated and then web three replacing those with new new markets that opened up by by what you can do by um by tokenization by data data sharing I think that's the it's going to be exciting to see them kind of uh pass the bat on from one another so to speak but yeah I know it's not very specific right but so General the thing that you're most excited about is efficiency improvements yeah so that's really okay
that is really boring fine I'm really excited I'm really excited for Netflix to get a better recommender algorithm because it recommends me rubbish um so you know if uh I mean that could be that could be a really good crossover as well so if Netflix started at the minute they have a recommender algorithm which is an example of AI it's some Tooling in AI is used to do that yeah and they try you know give you recommendations based on what you've watched before but there are all sorts of other things you could do the web through you could do recommendations that are waited by how much you were willing to pay how much of you're willing to pay per micro payment for you know how much you um they might be able to do it using other data sets that you permission and share from a different source so at the minute they have to do recommendations purely based on what you do in the app but you might be able to connect your activity
on IMDb or um on a different stream because there are like 20 different streaming providers now right so they don't know about each other they don't share data necessarily but if you could then share your data set with the all and say okay I'm GNA tell you all what I'm watching and then you can recommend me better stuff that would be quite quite good so you can spend you know more time at home watching Netflix and other streaming platforms less time scrolling through really bad recommendations efficiency improvements efficiency so I think there's two for me one with my like nerdy gamer hat on one is the the AI content generation like the idea of having in these kind of metaverse gaming worlds little AI Bots that like you said can like spin out incredible stories can interact real time can provide like really cool feedback and exciting kind of dialogue and exciting kind of actions and all this kind of stuff that stuff I I find really exciting as a gamer the other stuff is
more like the business aspect the payment rails when we start to get AIS that can you know Commerce an exchange and buy things and like subscribe and you can just imagine a world where you just go to an AI and you say make this business for me set it up for me or you know make this product for me and it just goes out there it actions it sends you know the funds wherever it needs to go and you just get these like incredible projects these multifaceted projects multifaceted projects sorry that are just like set up by this AI I think that is just an incredible idea yeah I've seen I've seen a super talented developer who basically implemented this exact idea right and he um he can basically he he built an app that you can speak to Siri on and you can say Hey Siri make me a I've just said Hey Siri so everyone listening on their speakers will now have uh just have just been paused by Siri they'll say yeah
yeah clickbait so they'll say hey Siri can you make me uh a new uh website for this type of product it should have a login it should have a database to manage users Etc and then in the background it gets all these other generative models to build to write the code deploy it automatically do all the all the all the cicd um it's insane like that's already exists you can just speak and it and it becomes so yeah I think that's definitely exciting but you know it's it's old news as well so it's exciting so like there's a utopian future and a potentially dystopian future but I think overall I see a utopian future and I think seems like you do too and I think we're excited about the potential of AI and also the potential of web 3 to support AI going forward and I think the overlap for me is one of the most exciting things yeah for sure so yeah that's that's a good place to end we've uh We've covered everything
from Ai and web 3 to the existence of Free Will in one episode and now it's time to sit in the dark corner and cry thank thanks for tuning in to another episode of untangling web 3 produced by Emma caleri don't forget to send us your thoughts questions and comments on social media and be sure to follow us on your favorite podcast provider to catch the next episode see you next time to untangle a little bit more of web the views we express here are our own and do not reflect the views of our
Topics: AI & Robotics · Blockchain & Crypto