Tubetotext

    Thriving With AI

    Motivation to Write a Book on AI

    the motivation to write the book was then to say can we give agency an understanding to the lay person so that they can be uh you know better citizens uh in an age that's you know getting transformed by Ai and to finish that idea the we said that let's instead of writing stories about companies let's write stories about how AI is embedded in our everyday life

    02:00

    The New Revolution in AI

    so so yeah so so yeah so so yeah so the theoretical development around the ideas that are powering today's AI uh started in the ' 50s uh there was a conference at Dartmouth University in the ' 50s that essentially laid the groundwork uh for you know the models for example deep learning models that uh we are actually using today

    05:47

    The Combination of Human Intelligence and Artificial Intelligence

    I think in a in a nutshell uh what we need to figure out is how to combine human intelligence with artificial intelligence right um so the human intelligence AI is going to take is going to change the Baseline in a way that we can solve more and more interesting problems that we can't solve today right

    13:12

    The House of AI Framework

    uh it has a foundation of data engineering right so uh there is so much data out there to make it useful to do anything like to predict whether somebody will get cancer or not we have to spend a lot of time in energy in curating that data right

    15:42

    Anomaly Detection

    I tried to do something that I never do which is actually to try and go and buy jewelry for Sophia right so so I went there I I chose you know a bracelet that I really liked and I wanted to pay for it and I gave my MX and it got flagged right so I was like why is MX tagging this right the transaction amount is not you know I mean it's not small but it's not that big you know many many business class airline tickets are more than that it's not that you know uh I've never been to Rio before it's the third time I was there uh it was not you know so the interesting thing was actually it was the combination of the category of you know product that I was buying the location the amount and AMX did not unfortunately for Sophia AMX did not have a history of me buying jewelry for her right and and so so that was an anomaly uh that was flagged

    17:13

    Supervised Learning

    One of my favorite use cases out here is actually a clinic in Hungary that build a model to predict whether a particular cell that is seeing right now will be malignant uh you know in 3 months right so basically a cancer detection model out there right so this when you say this is this is the supervised learning part where you tell the model this is what I'm looking for versus the first one where it says this is find me any pattern mapping and then so this is the supervised learning part absolutely right

    18:47

    Deep Learning

    Deep learning comes into play when you move to uh you know Beyond Simple like Excel uh uh sheet based numeric data to images to audio to video to you know uh uh detecting for example uh whether you know uh when a tler is out fishing in the deep sea you know is the fish that they capturing actually legal or illegal right the endangered species that they should not be catching

    21:38

    Generative AI

    Generative AI has been it's built on the foundation of all of these uh the idea here is that I can generate text I can generate audio I can generate video uh and and we can get into some of the so generative actually came about with how should we respond because it's generating based on all these things causation prediction and prescriptive there's optimization involved in there it's it's a the foundational pillars are it is a basically a prediction model but it is tuned with lot of the other aspects as well

    22:25

    Ethical and Fair Use of AI

    The fact that when we develop these algorithms using historical data if our society is biased right then the algorithm that you develop will actually maybe even further those biases which is happening a lot now yeah and and and so that's a big big area of research in many universities we are we are designing algorithms that take into account it's not just about the most accurate algorithm it also has to be fair right

    23:10

    AI in Education

    What's happening with with especially with generative AI is that lot of time and energy is being spent in universities uh trying to uh basically quote unquote protect faculty's existing way of doing things from what gen can do right so that's the big conversation I hear a lot what I don't hear a lot about and which is what I think we people should be spending time and energy on is how can we use this technology to help students learn better

    25:56

    AI in Healthcare

    I was on a panel with the chief of AI at Mayo Clinic right uh so my friend puu is here where's puu right so he's he's got a loyalty card uh not for himself but you know uh for for somebody Clos in his family he you know I see p a lot in Minneapolis because he visits the Mayo Clinic is probably one of the leading hospitals in the world uh and most of the people who go there have severe complications right

    32:08

    Healthcare and AI

    so let's say you know a patient walks in I'm the doctor I have I have to you know examine this patient this patient has a 500 page medical file as a doctor I don't have the time to read that right so synthesis summarization this is a superpower of generative AI right and they have now embedded that into every doctor visit out there right so the patient chart is summarized for the doctor the doctor is better informed they can have a you know richer conversation with the patient out there right

    33:11

    Leapfrogging in India

    I mean one thing I've seen in India is that we tend to Leap Frog right uh so I think like the crappy electronic medical record systems that we have in the us we don't need them here so I think what we're going to get probably as you know hospitals get uh less fragmented and get you know perhaps more corporatized you're going to get you know good Superior uh you know electronic medical records

    34:04

    Biotech and AI

    I know time is short so we'll cover the last sector which you have mentioned first which is relationships and dating and and how AI is affecting that and and how what's going to happen there next yeah so I think you know my exposure to this uh came through a uh you know colleague of mine uh Professor ji ramar Prasad who's at Maryland now uh you know she went to school with the former CEO of OK Cupid

    38:01

    Online Dating and AI

    so so I think Christian's Vision was that you know can we create a platform where somebody who may be not that Adept at finding a date in the real world can because now we have you know large amounts of data you know that person has a shot right can can interact with somebody and then maybe go out for coffee and then you know

    39:33

    Robots and Job Displacement

    I think that's the world that we have to drive but I'll end by this you know really uh interesting piece of research that was done by a friend colleague at Wen Professor ly Wu she studied the adoption of robots in every Canadian company over the last 10 years okay and what she found was actually three interesting things

    43:28

    AI and Diversity

    so I think uh there is good work happening at I Madras and Indian of Science in Bangalore with the AI for Bharat project and the bashini project to try to create uh large language models that basically you know I think address 23 of the local Indian languages out there right

    45:53

    Authenticity of AI Results

    but is it I mean it's very convincing as well what do you see but is it I mean how authentic is it and that's something which and it's obviously unlike a Google search where it lead you to the sources right that there's a 20 things where you can look and then those are actual sources this is generating a response and maybe this is the most basic form of AI right

    48:22

    Generative AI Models

    these generative AI models that everybody is excited about like charp Gemini Claude you know you have models for music like generating music like sunno uh and others uh they all what their job is to predict probabilistically like the next token to be technical right but the token could be a word it could be a sound clip it could be a piece of video it could be a part of an image um so it's a probabilistic model and it can what we call hallucinate it can make up basically okay uh so that is the fundamental nature of these models

    49:05

    Hallucinations in AI Models

    now open AI would not have been the company that it is or any of these models would not have you know seen the light of the day if vast majority of what was happening was hallucination um so they have an incentive to basically you know put more guard rails to give references to you know validate the accuracy right and there's a lot of time and effort and you know investment going into that right

    49:40

    Guardrails and Validation

    so I think uh it in general is going to move towards a world where hallucinations are lower than what we have today and the other thing I should mention is that we can use this technology to create you know uh guard rail protected applications right so KH Migo is a good example it is not allowed to H because they've worked really hard to keep it within the boundaries of you know being a tutor right

    50:09

    Learning to Code

    my first answer is yes you should learn how to code because it's coding is actually a way of thinking it's a way to you know understand uh you know how to create logical structures so that you can solve problems right so it's like you know uh I mean it's a crazy analogy but it's like learning you know should you play tennis yeah you should play tennis because it you know improves uh your like you know high and ey coordination improves the seminar so you can think of coding in that same Spirit right

    53:17

    Human-in-the-Loop

    I don't think that look driverless cars are here probably know fastest decision making is required there so if you really want it that Tech is there I don't think Society wants that one two so you I do not see I'm working with Enterprises in the US I for now I definitely see a need for a human in the loop uh I don't think we'll want to eliminate it as well and uh when the time comes there will be so to answer the previous question as well I think just like Darwin we'll have to evolve to better than a and we will

    58:38

    Task Automation

    really lowend work like that is there anything that you can yeah robots are in China if you go to a warehouse in of of Alibaba you'll see there's no human beings at all everything is barcoded boom boom boom radio frequency RFID chips knows exactly where to send what it's pre-coded so there's no

    60:02