Apple Podcasts
Spotify
Castro
RSS Feed
Breaker
PocketCasts
YouTube Channel
The player is loading ...
June 17, 2024
The Evolution and Impact of Generative AI with Martin Musiol | Episode #85
In Episode 85 of Great Things with Great Tech, Anthony Spiteri interviews Martin Musiol, founder of GenerativeAI.net. They discuss Martin's journey in AI, the growth of GenerativeAI.net, and the current and future state of generative AI. Topics include GANs, transformer models, and the practical applications of AI in various industries. Martin shares insights on challenge…
In Episode 85 of Great Things with Great Tech, Anthony Spiteri interviews Martin Musiol, founder of GenerativeAI.net. They discuss Martin's journey in AI, the growth of GenerativeAI.net, and the current and future state of generative AI. Topics include GANs, transformer models, and the practical applications of AI in various industries. Martin shares insights on challenges and opportunities in the AI landscape and offers a glimpse into the future of artificial general intelligence.
Generative AI in 2016? Get an early look and how the AI/ML industry was back when data science was evolving into what we see today! Future directions in Generative #AI. This episode features Martin Musiol, founder of GenerativeAI.net, as he discusses the evolution and impact of generative AI on various industries. The conversation spans Martin's career journey, from his early days in data science to the founding of GenerativeAI.net and its mission to educate and implement generative AI solutions.
Key topics include the differences between traditional AI/ML and generative AI, practical applications, challenges and opportunities in the field, and the future of artificial general intelligence (AGI). GenerativeAI.net was founded in 2018 and focuses on educating and consulting on generative AI technologies. The company offers online courses, a popular newsletter, and consultancy services to help businesses implement AI solutions effectively.
Technology and Technology Partners Mentioned: GenerativeAI.net, GANs, Transformer Models, IBM, Infosys Consulting, AI, ML, Data Science, NLP, LSTM, Bird Model, OpenAI, GPT, AGI, Cognitive Services, AI Consultancy, Fleeky AI, RAG System ☑️ Web: https://www.generativeai.net
☑️ Support the Channel: https://ko-fi.com/gtwgt ☑️ Be on #GTwGT: Contact via Twitter @GTwGTPodcast or visit https://www.gtwgt.com
☑️ Subscribe to YouTube: https://www.youtube.com/@GTwGTPodcast?sub_confirmation=1
☑️ Subscribe to Spotify: https://open.spotify.com/show/5Y1Fgl4DgGpFd5Z4dHulVX • https://gtwgt.com Web • https://twitter.com/GTwGTPodcast Twitter • https://podcasts.apple.com/us/podcast/id1519439787?mt=2&ls=1 Apple Podcasts ☑️ Music: https://www.bensound.com
Transcript
RAW
imagine a world where machines can create art write stories and even design products this isn't Science Fiction it's the realm of generative AI but hey we know that but were you talking about generative AI back in 2016 my next guest certainly was welcome to episode 85 of great things with great Tech I'm Anthony spery and today we're joined by Martin moel founder of generative a.
net Martin has been at the Forefront of AI Innovation helping companies navigate and harness the power of this technology in this episode we'll explore how generative AI is changing the landscape the Journey of generative a.net and Martin's insights into the current and future states of AI hey it's great to have you here Martin um you know from your early passion for AI to becoming a leading figure in generative AI ipace you've had an inspiring Journey so maybe just give a bit of a background about yourself um
you know I believe you've had a really interesting background in data science and machine learning and AI but give a bit of background about yourself before you know we talk about generative a.net and everything that you're doing there yeah first of all thank you so much for the uh for having me on on the show um yeah I'm Martin musol uh I am um in 2016 um I was aad a scientist at frog design and there was the um a basically sort of a design agency um and I started my career and I stumbled upon a paper
that said generative adversarial networks it was actually from 2014 by Ian goodfella and um that inspired me to like think a bit further like where is the all of this teag potentially going and so I decided to um actually have a conference uh presentation a conference speech about that and so I presented about generative on on a yeah on a conference in Milan Italy and the title I say remember was generative AI how the new milestones in AI improve the products and services we built yeah um so that was that was quite interesting
for the audience that was there but the that sort of like little buzz faded away quite quickly um anyways I then continued my journey as a data scientist I was a senior data scientist at IBM Consulting um then I became a team lead at IBM Consulting yeah manager data science manager basically um on the side I was uh continuing with generative AI but uh there was no real business so I was focusing like I had to work as a data center which was also great I learned a lot and we also implemented lots of AI Solutions but not generative AI
Solutions uh in um uh 2022 I then switched to infosis Consulting um or actually to be more consistent with the timeline 2018 I founded generative a.net uh four and a half years before cgbd came out we can talk about that in a moment uh 2022 I uh became the generative AI lead for infosis Consulting for Europe that was uh fantastic for me because uh there was a you know lots of uh there was right after chbt and lots of projects it was uh it was very busy very busy times um and I was uh not not only Hands-On but also steering teams into delivering
these projects um and I decided then October 2022 uh 2023 to um quit my job and go full-time into my own projects met generative a.net all right that's a that that's that's a really good Roundup um let's dive into it though right because I think people want to kind of hear about you know where you came from first cuz what I'm hearing there is that you coined you probably didn't coin the phrase but you were certainly using the generative AI phrase back in 2016 which if most people are honest about
themselves no one was really talking about generative generative AI until maybe when chat gbt did its thing and released um you know towards the end of 2022 so you know what just talk about your early years in terms of you becoming a data scientist how did that start what what triggered you to sort of get into that field initially so go go right back to the start there yeah okay thank you um so your your first point that you mentioned coining gener AI I didn't coin I wish I would back in time I actually did at some point because I
wasn't so sure anymore years later uh so I did actually the in-depth research and then you know I found generative models which is the technical term for the yeah or like generative adversary networks by Ian gfeller is also a generative model but generative AI I actually found it then someone mentioned it 2015 somewhere so I thought okay but it's anyway it's not so important um and yeah I so when I saw what G yeah short what they are capable of it was not it was not language uh you know language models was
not not even in the picture yet really um language models came like 2017 yeah there was with bird by Google um that became a topic but before that it was image generation so we saw first versions of um images being generated uh uh very low low Fidelity images yeah but um look looking at the progress of technology and the exponential um developments that we have experienced with also the Computing needs and and and all of other all of the other things with Mo lore and so forth um I thought okay the these images might get photo
realistic at some point uh I didn't know that that they become so early photo realistic um but anyways with that with that in mind and also working um because at that time I was working at a design agency um I knew that this is very relevant for them uh and uh yeah so I I I it was a big topic uh that I talked about and yeah so that triggered everything yeah and language models 2017 that started and at IBM Consulting we had a project where we implemented two years later um The Bird model um but very like in
comparison to what uh large language models are capable now or even small language models um bird was not even closer okay bird is also much smaller than a currently considered small language model but yeah but still yeah so so in terms of your own passion for data science and AI in general how how did that start like where where did you get your start when you were were you were kid was it in high school like where where did you start that passion because obviously you're very passionate about it right yeah I am I am passionate
about it and I was always passionate about science fiction Star Wars RTD2 and uh robots Androids all of these things uh very imaginative uh mind and I also loved reading about it um but uh when I look at uh my so I did my bachelors in Northern Germany in a com in a university called Lana University and there I I I studied um indust industrial engineering what electrical engineer electrical engineering but in German it has industrial head um and there I I I took a course that was called self-learning systems and there the the
professor was talking about perceptrons and in this course I actually understood it was half a year I understood um that uh how how programs can learn along training data and that that there was a like a light bulb went went up immediately and I knew immediately that I want to go into into self- Learning Systems AI um uh at that time and I uh did because I did quite well on my bachelor so I I could choose why where I wanted to go and I wanted and I chose in my opinion the a good uh German Technical University the Technical
University of Munich um I did the computational science and engineering course which has lots of AI components machine learning uh reinforcement learning track a couple of other things and I and I and I focused our deep learning and I focused on that uh yeah and uh and from that on I continued yeah because I I think you know obviously the generative AI today that we talk about from a consumer sense is a lot far removed from you know data science from you talked about you know machine learning um natural language all that kind of stuff
right so you know what what is the difference between you know traditional a ML and what you learned um reinforcement learning and all that kind of jazz versus what we see today I mean there's a big difference right because I mean generative um generative AI is new um but it's a subset of what's come before so maybe just explain a little bit about that because I think people are always interested to sort of say okay well what is AI ml versus generative AI yeah H good good point so um the in 2014 um um or I think even
already earlier um the uh people found out that you can use gpus not only for gaming but also for training models and with these models uh uh initially um um are part of the or the ones that first had an application in the industry that really brought value were um traditional machine learning algorithms such as uh or in the technical term I would call that discriminative AI discriminative because that's the that's a contrast to generative because discriminative models they um what they do they basically have
to discriminate or select choose from options so it could be a classifier yeah choosing from different options can be a recommendation engines that basically clusters us so it has to identify where are we clustered in in which which classes so decide yeah discriminate um so the technical term is the discriminate colloquially discriminate might be understood a bit differently um then also regression is some kind of like a decision in a continuous space reinforcement learn traditional reinforcement learning which is also
part of discriminative AI is um deciding the the robot the warehouse robot which is the the shortest path to take to this yeah decide between these different uh options and then we have also dimensionality reduction but that becomes a bit more theor technical deeper um but it's abstracted I would say uh it's deciding between options but the options can also be um uh continuous yeah so uh computer vision models where it comes detecting um if a certain uh edge of an industrial process uh is U is the right Edge has the right angle or or
not yeah this be identified with these cases is it maybe our face unlocking the phone yeah yes I was going I was going to say I was going to ask where does you know because the typical use case for AI and ml before you know chat gbt and open AI did what they did to bring it to the rest of the world and like they've done was really about okay can you have an image and if with that image can you determine whether that image is something that you're searching for versus not like that to me was was the the Crux of AI and ml um you know then
that was applied to to data to to to Big Data it was applied to science it was applied to engineering to Industry um I've got a friend who was um he works with for Grain Company and he's trying to create a a machine that basically checks and photographs all the all the the grains of of um the weight and works out which ones are good or viable versus not so and is using machine learning for that so traditionally if if I'm not mistaken that was really where that technology was being applied right yeah correct and there was a the
return of invest made sense also that it really had a an industry application there yeah so there it was not like a big research for you could implement it sort of in a lean fashion and provide real value I remember we had a projects with an oil and gas company where we were uh uh with a knowledge CFT or scanning documents uh first and then in the documents there were like different images and we were classifying the images with with uh with convolutional neural networks back then uh the text uh we we actually went through the text
also and um had traditional NLP um so classifying the text is that um a report a dwelling drilling report or is it some other kind of survey and classifying the texts like this this yeah nowadays language model would easily do that yes in depth it's happened very quickly hasn't it because even if you think about it a few years ago we really didn't a lot of people didn't have a notion of what was a natural language model natural language interface really what that was but yeah when did that sort of come into into play with regards
to AI ml natural language llm so the official paper is by Google by head author is vasvani um 2017 um was it was called um attention is all you need was the the paper's name and that was the the the first um architecture of a of a okay a Transformer language model because Transformer yeah Transformer language model that was the first Transformer language it was called BT and actually fun fact yeah if Google would have um doubled down on that paper at that time 2017 they could have had a chat GPT you know two years later maybe
or three years like like years before open AI would have done that H but that's a that's just a side info but there was if we are technically um if if you want to call lstn long short term memory uh um architectures also language models then that was the state-ofthe-art before Transformers entered the the stage yeah we would do things with lstms so what's a Transformer so just explain explain that for people that might not know what a what a Transformer is in this context um uh a Transformer um is a is a yeah is
a is a language uh model it's a it's an architecture that has a context window it takes a certain uh size of text into account that it scans it the original one was a bird B directional it was like scanning it from both sides and based on that it could identify in or let's let's put simple way what language models are able to do very well um along billions trillions of token that they have ingested in the training phase they are able to uh train the to to um predict the next word and this simple
one-word next prediction um we have we have seen that over a certain size of these of these Transformer models and over a certain size of training data um cap capabilities are coming out so so-called Emer capabilities and these just with predicting the next word but if we prompt it right through the through the through the context window if we prompted right then um they are able to emulate personas they are able to um do some kind of forecasting classification they are super versatile in in in the terms of tasks that they can take and
U on a level of a Almost Human intelligence level yeah which later we'll on the AGI convers that's I know you gots on that okay so you've gone through your background and and how you got into this world so you know your passion started around 2016 for generative Ai and you talked about that um what what what conference was it I'm I'm intrigued as to what type of conference it was it was called Data driven Innovation Summit I'm not sure about the data driven innovation maybe only datadriven Innovation I I can
I can find the link yeah that's interesting so so so it was all about data in data basically yeah data driven Innovation these three wordss for sure yeah for sure you see and you see that I say with my with my accent I say data um so data data it's all all the it's all the same um okay so then you know then in 2018 you're still working you decided to launch generative a.
net so tell us a little bit about the launch of that and what you were looking to achieve by launching generative a.net and by the way that's a obviously a very very highly sort after domain as well I would have thought right uh so because so first um to first point uh I launched it because um I um I had one one goal in mind because I I realized many like that was not the only conference I was talking about then there were many other conferences coming and I realized that uh this is a topic where um other uh like-minded people are super interested
in and so I thought I I'll make an online course about this which I did and uh that was 2018 together with generative i.net and uh to my knowledge it's the world's first uh online course on on generative AI um now there are many many yeah and many very good ones also um and um so that was and regarding to the to the domain yeah generative i.
net was good I actually had even the chance to get generative fi.com but uh at that time like I was I know yeah the com's always more worth more than the net isn't it yeah yeah and then I thought it's too expensive I take the net and and I actually also got already uh um offers to to sell the the domain but uh I was not satisfied with the offer and I and I'm not seeking to to sell any anyways yeah no it's good it's a good one I think it yeah it shows your intent there and it's it's obviously going to rank very well um and that's
part of you know where where this the positivity comes out like four years later when this thing really hits um so you know tell us a littleit bit about the courses and then you've also got a newsletter so you know take us through a little overview of the courses what they what they allow you to learn and then obviously after that talk about the newsletter but start with the courses yeah so so the courses uh we have currently two courses uh one so not that much yeah one course is the general uh generative AI course all things
generative a um and um here uh so I have iterated it multiple times yeah and I guess at some point it's yeah it has it has to be maintained because things are just changing so much um so it's not only about language models it's also about image Generation video generation 3D object generation uh basically the whole spectrum and I'm also reasoning about uh like uh just very indicatively what's what's the path to to further higher intelligent in artificial intelligent um and the other course is a
marketing course because um there is a roughly roughly nine months ago there was a survey um about from McKenzie where they we mapping business uh functions versus industry on where generative AI has sees the low hanging fruits and and you know Revenue increase can be achieved and across all Industries uh marketing is very thick thick blue uh so with the highest potential of yeah of of of disruption gen AI disruption oh and I also got invited to many marketing conferences it makes sense yeah when I I me I work I
work with I mean I'm I'm a effectively a tech marketer um as part of my job as well and I work with our marketing people and I know that they're very worried about their jobs moving forward so it's um it's one of those areas which it's it's it's a natural sort of fit because it's all about creating content right and generating content from nothing or from from prompting as well so yeah it's no surprise that that's there you know just today I I I lost roughly four hours with
one tool and that is a fleeky AI not advertising anything but I I I got to know that tool and you just paste in an Amazon URL and it makes a review video from that like it's three clicks and you can post it on YouTube and have a review that was that was what's what what's it called this just out of Interest flicky ai flicky ai oh yeah great there's so much um there's so much in innov there's so much innovation in this in that space and there's so much abil uh potential for smart people to
create really you know tremendous accelerated Technologies based on this so flicky that's one that I'll I'll look at for sure that's very interesting afterwards um and then you know then when did you start the newsletter because starting newsletters obviously in Tech is is something that people do um but you were mentioning before what it's got about 40,000 subscribers now 43,000 43,000 yeah yeah yeah so so how so what was the idea to start that yeah so um I had I started actually with a
waiting list for my uh um online course uh I I posted that on Hacker News that that went like mediocre viral there then on Reddit uh that was that was for number one for a long time so it got really lots of uh subscribers and then I created the online course and then I had already sort of like a base of of of emails and I used that to start with I told them hey I I write now about tech because I'm like I'm anyways daily I'm thinking about it um also because of my job and so forth and I'm interested so
I'm I'm I'm writing weekly about it or I started actually bi-weekly so every second week and uh I asked them hey if you want to uh know about this stay otherwise feel free to unsubscribe um and I did it through Genera net so so this Squarespace hosted um and uh yeah then I basically uh start wrote about it then I had a little break I must say because it's not always so easy to to follow up but now since roughly a year I'm doing it very consistently and um the the the subscriber base grew and
grew uh we I I didn't advertise it a single time yeah it was all organic good that's great yeah 43,000 again not a single dollar spent on on ads um and uh and they appreciate it a lot yeah I'm very each episode triggers lots of uh questions uh or conversations yeah yeah and so and so much so that you've I mean you mentioned that you've worked for IBM and then your other consultant of your role so are you have you stepped away from that now and you're doing this fulltime is this is this your your gig
now no actually this is still a s work that I'm doing so it amounts to 10 to 10 hours in in if there are many updates then maybe 12 hours 14 hours a week still a lot yeah okay still a lot yeah but I also have like lean weeks where it's uh maybe 5 hours for two episodes two and a half and two and a half uh roughly um I actually like the the Tuesday episode is more like a deep dive into a topic that I'm that I think is relevant at that time and then the Friday episode is a review of of my top findings what I think is worth to look
at uh yeah if there like some a new tool or uh or a video that's yeah or a new model yeah oh yeah I get you and then and then you also you've you've co-authored a book as well so what's the what's the book about um and what's it called yeah the book is um it's called generative AI navigating to the artificial general intelligence future AGI future um and it's sort of all things generi um a little uh history uh part to that because uh for instance 1965 there was a already the first
chatbot called Eliza by Dr Joseph yeah yeah you could literally chat with it and as soon as you said something like uh family or mention family then it asks you a question about your family and it get a conversation not really with of course not with a language model but yeah was interesting First Take on this kind of applications um then um I talk about the current state what exists yeah I'm also discussing different topics but what exists like what I said uh 3D object generation text generation code generation in science uh data
synthesis video generation and also the untapped Market of that I also discuss uh topics such as open source versus closed Source um and large language models and smaller ones and then the last third is towards like where is all of this headed and I'm looking also at even at robotics um continuing the exponential uh movement exponential behavior of computing power data availability talent in the space um yeah uh investment in the space and uh make an educated guess of where this this could go yeah that's
really interesting oh go ahead yeah one more thing I didn't mention um so I'm I'm having the newsletter I wrote the book but uh my main job is I have my own consultancy and we Implement generative AI solutions for for companies yeah there you go okay right so that that that's the main gig which is important and and obviously practicing what you preach um in terms of that so okay so a question for you so what are the what are the current challenges and opportunities in in this field of generative
AI so opportunities there are there are plethora of opportunities um I think um when it comes if you have a large knowledge uh base and you you have frequently some various kinds of questions against this knowledge base then you could use um yeah basically uh generative a i through a rack system retrieval augmented Generation Um to to to meaningfully ex answer questions uh automate on that end uh you can automate processes uh that you might have in an organization you can accelerate uh uh Innovation yeah because they use it
maybe as a sparing partner or or just getting prototyping things very quickly off the ground with uh yeah the various uh there are thousands of tools uh to choose from um I must say 90% of the cases are automatic uh answering yeah of questions that are coming in I've I've done that for an international uh uh uh container Shipping Company where literally millions of emails are coming in and there was each time there was a human answering these emails and they had to check uh on their documentation which what answers their resp to write
it down and cross check it and send it and now we have we have lowered that work by 80% manual yeah yeah so that that leads itself to a good question around um and I heard Sam ultman talk about this with Lex Freeman actually on that podcast where they were talking about in that case you know the shipping container example you know that that was someone's job it was a monotonous job it was wasn't the greatest job but that was someone's job that people got paid for but now you know the company hires you
to come in consult and try and work out how they can automate that and make it more efficient right and that's all about efficiency but in that efficiency someone is not working as much as they used to so where where do you stand on that in terms of where this you know is it augmenting and making us more efficient or is it replacing jobs I believe in uh Evolution instead of Revolution so um how we have built now with this specific case how we have done this is that the representative the client representative is still uh
working in between and working with a chatbot in the initial phase and basically copy and paste the the answer in gets an answer from the chatboard cross checks this maybe has to interact again with the chatboard and then sends it out and first we do that um so that's like a heavy argumentation I would say um but we do that also because of liability issues yeah we we want a human in the middle is the the liability is quite easy to answer that's then the responsible for it yeah um then then after uh this augmentation part comes um
Automation and then at some point like where the human is in supervision only doesn't have to maybe interact so much and then at some point um it has to be deprecation also of the human that it's fully automated and things are because they just become so reliable these systems yeah so at the moment they're not at a point just yet do you think where they are completely reliable and we we will TR it's it's an element of trust right in these models it's a question that I think I got asked
yesterday I was I was asked a question around how much do you trust the answer um and the response based on the inputs and the prompting so from your point of view it's still not quite there where you know it's it's it's it's it's fully trustworthy so there's still an element that needs to be a humanized element right yes U it I think it depends a little bit on the on the use case uh probably there are use cases where we can fully even automate it already but uh the like simp simple things um and
also where one error is maybe not a big such big of of a deal but if you work with clients directly or you write as some sensitive information to a database you want to have this this check in between uh as of yet because there's also this uh I'm not uh I just recently stumbled upon that that there is um a hall hallucination leaderboard so it it shows which language yeah has is How likely to hallucinate oh who's who's leading GPT 40 is leading with 2 2.5% of hallucination 2.
5 then there is 2.6 a smaller model and I always thought the smaller the model the less likely that they are to hallucinate because uh there's just less learned and you can maybe more focus on things but that's not necessarily true there's ation point of view just so is hallucination where it basically creates an a fictitious answer that's not even correct or is it getting the answer wrong because what I've noticed with small models when I download Al as an example and use their smaller parameter model it's got less
context or less parameters to answer the question in so if I ask it about um myself it might say that I'm born in New Zealand versus Australia or something like that is is that a hallucination or is that just an error what's a differentiation there um that is a good question I think um I we talked about it I was last week on on a conference we talked about that um uh so I I think uh if if it gets something wrong um if it has the context information or not I would I would just actually the no on the on the
leaderboard they differentiate between incorrect answer and hallucination like really making something up um like incorrect is basically having a a factual knowledge just wrong and Hallucination is crafting it into like a its context crafting some story into a context because hallucination sounds more dangerous yeah yeah it's confident about something that is just not true yeah exactly like interesting that that's that's really interesting so we haven't got a lot lot of time left time has flown by I want to kind of lead into
that and I think it does lead into it very nicely around um gen gen General artificial intelligence and where we're heading there you know so you mentioned that we don't trust we don't trust these things just yet fully um you also mentioned your interest in cyborgs and Robotics and that sort of thing so this is all kind of coming together right so the first step is to get a large language model that is sentient and that is you know generally generatively intelligent where do you see this going what's what's your take on it yeah so uh
first of all I I think when we talk about an AGI it's about an understanding of the world and just going on on language level can in my opinion and also in the opinion of uh of of yan leun not um not not a language can not be right because uh it's just an approximation of the world for instance if um if someone uh tells you how it was uh to live in the rainforest for like 3 months sounds interesting you you you have understood maybe what the person said and you can like yeah sympathize with that but is it the same as really
living for three months in the rainforest it's not um so there is an approximation sort of of of language and how do we like get closer to the truth to the reality of the world um and we get this first with multimodal yeah we have to have visual um then also audio and and and and and multimodal aspect then multimodal in my opinion is also the ability to interact with the physical world um which can happen with robots yeah and we see already open AI collaborating together with figure for instance where I'm not
sure if you have seen that video but where the the robot is like really taking cleaning the kitchen and then talking and yes I saying that yeah yeah yeah there we in this video we actually also observe the another piece which is multitasking our brains are always multitasking at any point in time if we do it consciously or subconsciously um this is another important piece um yeah and and and and and and many more MAA has also a good approach yeah they have this japa Jaa architecture where they draw conclusions uh between
different modes um of of of of topics and and basically abstract Concepts so basically they they try to predict Concepts instead of words that's super powerful stuff but not right now in the top in the top models um present um yeah there are there are many things so multimodal multi uh asking and multi- sensory is also uh important there is here's in Munich actually a company that does artificial skin so each square cimer of of that artificial silicon skin has hundreds of sensors and um right and uh yeah that uh that's and of course
bigger models um perhaps a change in in the architecture that we beyond Transformers we see also Mamba the Mamba architecture coming up um so there are a couple of things and even if we do all of this right and more data larger models all of the mention before even then we might still miss something that that will bring us to this uh yeah so yeah so if I was to ask you what is um General artificial intelligence or artificial general intelligence what what how would you describe it when once we get there what would be the the base
explanation if you were telling someone about it yeah so um looking at language models right now they obviously they are not they they they are lacking some uh problem solving skills or some uh some human creativity element rather than just you know they are better in creativity in the sense of like generating images uh but they're not as creative as as humans in in another dimension so um having this capabilities and to a degree they also right now much more intelligent than humans because they have just so much
more you can ask yeah so much more knowledge so having this knowledge on steroids including the the the the skills of the the lacking skills that humans have right now this is for me an AGI system interesting and in in in a in a minute or less than 30 seconds does the future excite you or worry you or scare you where we're going with this technology it highly highly excites me um but I'm also a techno Optimist um yeah and 100% understand if someone is um also uh afraid of the future or has like high highly respect of where where it might
go because it's just uncertain I understand that yeah awesome good answer hey hey Martin this has been a really good conversation I thank you for your insights your time U please check out the newsletter we'll link it in the show notes it's been a great episode um I really appreciate you being on the show so thanks a lot for that for being on great things with great T hey just as a reminder thanks for listening to this episode 85 of great things with great Tech stay tuned for more episodes where
we continue to highlight companies and Technologies shaping our world don't forget to follow us on social media at jtw JT podcast and visit jtwj t.com for more great content and all past episodes if you enjoyed this episode make sure to subscribe on your favorite podcast platform and on YouTube please spread the word and if you feel like it drop a review thanks for joining us and we'll see you next time on great things with great take [Music]
Join the Email List
Listen On
Support On
Top Episodes
Here are some great episodes to start with. Or, check out episodes by topic
