Keeping Up With Generative Intelligence Near Term with Ben Taylor

Keeping Up With Generative Intelligence Near Term with Ben Taylor

In an era where technology is transforming every facet of our lives, the healthcare sector stands to gain immensely. The presentation addresses intelligence and the ability to acquire knowledge, through experience, and apply that to future decisions, data, beating human benchmarks, and cost & time. In addition, the speaker explores the role of Artificial Intelligence (AI) in revolutionizing healthcare over the next three to five years using images to illustrate transformation and engagement with AI, accelerating the human creative, comfort and confidence with iteration, and the benefit of AI invented algorithms. We stand at the brink of a new age where every timestamp, notation, and piece of ‘digital exhaust’ could potentially contribute to our health assessment and longevity. The advent of ‘in perpetuum’ tools promises a future where AI will shape a new health economy.

hello everyone I’m G to share my screen and get started I’m Jetson tlor and today I’m going to be talking to you about generative intelligence keeping up with pace of Applied intelligence in the near- term um I am data iq’s Chief a AI strategist and one of the things I’d like to do before just jumping into AI is I’d like to start with intelligence I think before we can talk about artificial intelligence let’s just discuss intelligence because I think it’s a good foundation so there’s different definitions the definition that I like the most is intelligence is defined as the ability to acquire knowledge through experience and apply that to future decisions and I’m going to have a little bit of fun we’re going to go through um two animals in the animal kingdom I’m going to ask you if they’re intelligent then we’re going to settle on humans so take this caterpillar for instance is this intelligent so according to our previous definition is it going to learn if I do something today if I poke it in the eye is it going to pull away tomorrow this is it’s a terrible experiment but it’s something that we could test this might be a simple enough life form we might be able to show that it has a difficult time anticipating um something like that but maybe maybe it can learn take something like a cat something that’s more more advanced if I poke this cat in the eye it will avoid me not only for tomorrow but it might avoid me for the rest of its life but a key difference here when you think of cats versus Humans is I could poke every cat in the eye at least once humans are unique they are no other animal comes close to us when it comes to our ability to transfer experience so a negative experience and if we keep the the analogy or the story going if I poke a human in the eye and if I do that frequently not only will all the humans in society know that and avoid me but even if I left and came back hundreds or thousands of years in the future they might still remember that through stories and so our experience is compounding and for those of you that are listening that have medical backgrounds I there was one point in my life I was I took the mcad I was studying to go to medical school my dad’s a physician I the amount of material that I learn in chemical engineering to try to achieve a certain level of knowledge that many of you have is pretty overwhelming but if we if we take a step back and think about it I can’t take none of us can take credit for the amount of knowledge that we have we have to thank all the people around us we have to thank our teachers our peers our mentors we have to thank the people that came before us that had those lessons um that did that initial research if you think if you think about all the mistakes that were made in what is considered Western medicine to get to to where we are now it’s pretty significant so blood leting used to be considered Best in Class it it used if you think about um holistic or alternative medicines versus Western medicine things like bloodletting and some of the ideas and cures they had to prevent scurvy were pretty nonsensical and silly but they were it was all they knew at the time so so going back to this concept of intelligence uh two questions I’d like to ask the audience are if you think about your own intelligence and where you are now compared to when you were younger when you were in school how much smarter are you today and it’s pretty hard to quantify are you twice as smart 100 times smarter it feels feels nearly infinite but one of the important things to think about is time is not a guarantee of increased intelligence neither is experience did you know that experience can bias us we uh we have unconscious bias we’re creatures of habit the longer you work in a process the longer you’re stuck in a process the less likely you are to innovate that process you’ve probably worked um for organizations where there is a culture or an attitude of this is how we’ve always done it and I’m hoping to show you as we go through this talk track that this this is all changing drastically um especially this year and it’s only going to change faster and faster going forward so the nice thing about having the foundation intelligence is it now sets the table for AI because what’s the intermediate between human intelligence and AI well the intermediate is data so just like you have life experience opportunities challenges that you’ve been able to overcome that have made you wiser to make bigger more complicated decisions you also have your data experience that is flowing through a business through an organization through a hospital and un un fortunately there are fundamental problems that every every organization deals with and so one of one of those key problems is processes are stale so if I came in and audited your systems and said when’s the last time you changed this process for a lot of companies and a lot of hospitals and organizations we might find out that it hasn’t been changed for five years 10 years or never it’s never changed we’ve never changed this process the second issue we run into is knowledge workers leave so if you think about yourself when you’re junior you come into a role initially you feel insecure you kind of need to Shadow someone and come up through the ranks but then over time you become the principled gray-haired Gray beard individual in the room but then if you leave that knowledge can never be completely transferred and then the final issue that organizations all deal with or the final problem that a lot of organizations deal with is they do not leverage their data for better experience so I’m going to tell um a few healthc care stories that are related to this so during Co I was tasked with calling Hospital networks to get access to Patient level data and I remembered having a very frustrating conversation with someone on the phone where they were a senior health informationist they had access to Patient level data and they told me over the phone that not enough people have died to understand a disease and um it it’s very upsetting to hear that because I know where they’re going and I saidwhat do you what do you mean and they said we’ve only had six people die in our Hospital Network and so this isn’t this is a huge tragedy for for Americans or for anyone who’s dealing with a healthcare data set that has not been centralized and anonymized so the thousands of deaths in New York were completely meaningless for building models and understanding the disease for Hospital networks that weren’t sharing data the other the other issue that is more personal is um I I have relatives and family members that have different ill illnesses and my wife has chronic kidney disease so she has gone 17 18 19 years now where Western medicine has failed to diagnose her she said a kidney biopsy they looked at all the blood work all the urine analysis and they don’t know why she permanently lost 20% of her kidney function every time she had a child um fortunately she’s not having more kids and the kidney function is not dropping but this is an example where whatever issue she’s having um whatever problem is causing this it will most likely not be named after her and there are many many people that are having this issue but that data is not leveraged uh for better decisions but if you can make the data centralized and if you can respect privacy and if you can have a proper data Lake that’s ready to go to work where you’re not having to do heroics to get from data to value you that now opens up the conversation for AI and I I’ve been in this industry for a long time I didn’t really go do a more formal or deeper introduction but I have been I’ve worked in semiconductor for five years I’ve worked as a Quant at a Hedon I was the chief data scientist at an HR tech company predicting video interview outcome and I built and sold a deep Learning Company so I’ve so I’ve been in this space for a long time through different Industries but I’ve seen this um I’ve seen this industry grow up and one of the things that’s been happening as this industry grows up is it gets into we’re able to get into more higher consequence decisions that require more and more trust uh decisions that can make or break a company decisions that could impact a life for good good or bad we’re we’re finally playing at that level where we can do that responsibly the other thing that’s been happening too is our ability to consume different data types has also increased so before it was just structure data so think of data that is in a database and that then evolved into unstructured Data text images audio video um can turn off the notification really quick apologize for that um so by bringing AI into our data and by allowing us to collaborate this this last point is very important AI is not success uccessful in a silo it is not successful as a blackbox effort it has to collaborate with the domain experts the domain experts will always have something that AI does not it will edge cases intuition domain knowledge but also the safeties that we want in healthcare you don’t want an AI physician that is um writing prescriptions or or doing things that don’t have a human in the loop because AI can do things that are unexpected sometimes for various reasons so I’m I’m going to show you a few examples where I want to contract where we are with AI and how far we’ve gone so this was a marketing piece that I I I worked on a project seven or eight years ago teaching AI to play Call of Duty on the Xbox um it had a very powerful server from Intel 12 terabytes RAM and we and I was using reinforced learning to do this and for the marketing piece I sat down at a coffee shop and sketched this out on a napkin and the question for the audience and I I chuckle a little bit if this is outside your willhouse your normal willhouse but guess what was the cost and what was the timeline so basic so I had a sketch on an napkin I handed it to a designer and then they came back certain amount you know weeks later and I paid them a number for this so if you have the time and the number in your head a lot of people guess $5,000 two weeks it was $1,000 in one week very I this is a great designer that I work with and now for the next image some of you know where I’m going with this so this is another one that I created look at the detail look at the depth of field look at the reflection on the shoulder this um and there’s also some Optical effects that are happening here some heat effects as well if I this was made with mid Journey so it was done for pennies and uh in a few minutes and I’m using gp4 to help me write these detailed U descriptions or prompts that go into imid Journey if this was done before this I don’t know if I’d be able to deliver a piece of this quality for less than $50,000 and sooner than two months because because of the complexity that is happening here that this is a much more complicated piece and so the world is changing uh and I’m going to show you I’m going to watch walk you through some mid Journeys and the the the thing I want you to take away in these mid Journeys is the amount of creative comprehension that demonstrate that’s it demonstrates that it’s really understanding the underlying data and the analogy is also true in healthcare data so when it comes to EKG heartbeat waveforms um medical images patient level records the amount of comprehension that we’re we’re at now is pretty remarkable so this image here the input was I gave GPT 4 the article about worm GPT that’s that large language model that showed up on the dark web to do malicious activity and I asked it after reading after it read that doc that article I asked it to come up with an image that describes the Creature if there if such a thing existed that is worm GPT and it came up with this and if you and this is this is a profound image especially for AI to come up with here’s another one that’s a little bit more fun um this is me I enjoy mid Journey just playing around coming up with new Concepts testing them a little less depressing here’s another the the amount of detail is remarkable the other thing that might impact some of you in the home is it can do much more than just coming up with pretty images you can actually ask it to design acoustic wall wall paneling so if you have a grand piano in a room and you want to improve you want to break up flat walls not only can It produce this design it can also go take it into um into a depth map and that would then go Downstream into a cad system so you can imagine something on the left actually being designed by AI approved by you or even optimized and then sent through Cod and sent into a fabrication um provider and then shipped to your home and installed something that would be very labor intensive before can now be done um much much sooner I’m going to show you a few more examples of comprehension and then I’m going to get into a new technology category so for this example I asked GPT Ford mid journey to describe Microsoft as a creature let me show you another one SpaceX in in the through line going through this the thing I really want you to appreciate is the level of comprehension so the the fact fact that it’s designing all aspects of this alien type insect to represent the company SpaceX is it’s it’s jaw-dropping that it has that level of comprehension uh here’s McDonald’s the other thing to uh think about too is there aren’t just a few variations of these I I have an unlimited number of creatures that represent McDonald’s and if I want to try different variations um I love the french fries in the in the hair um it’ss everything about this image is coming from the data snowflake if you’re familiar with snowflake they’re a a Lakehouse company they compete with data bricks and Ice Wolf so the theme that we started with with the images what’s the time and the timeline to deliver um the artwork there’s also something that’s starting to happen now in the algorithm space so it’s not that important that you understand all of the code that you’re looking at but if you were asked to estimate what’s the time uh and the cost to deliver the code that you see the the cost is there is no price tag before GT4 showed up so what you’re seeing is this is a synthetic algorithm that’s been written by gp4 so it’s a generative algorithm and it’s beating human benchmarks so this is a complicated plot but it’s an important plot to understand so I’m going to try to walk you through it every bar is an algorithm every red bar is an algorithm that was invented by a human and these are algorithms that are actively used in healthcare in biopharmaceuticals and oil and gas these are optimizers and they’re used to optimize um expensive processes or potentially systems there’s a lot of sensors within an MRI machine other things like that that are optimized for signal to noise ratio and to improve the quality of the data that they’re able to capture so if every bar that is read is an algorithm that’s written by humans every bar that is not read is a synthetic algorithm that was created by gp4 gp4 is powerful enough that it can create synthetic algorithms the shorter the bar the more competitive the algorithm is and so what that means is we’re moving into a new era where I believe in the next three years every algorithm that matters will be completely Rewritten by Ai and this would also be true for healthcare so if you think about algorithms that process an EKG signal like heart rate variability uh things like that they will all be replaced very very quickly in the next three years because there will be um they’ll be faster more performant higher accuracies more trusted more reliable Etc so the problem we have with generative AI is it doesn’t innovate it just generates and some of those Generations can be hallucinations it can actually produce garbage uh as we Ru into this next category uh called generative algorithms we are able to innovate we’re able to drive to a goal so now the goal is producing an algorithm an app or even an operating system that achieves the goal that is defined by the human user this is another way to visualize it that I like um so you have artificial intelligence within that you have the subset machine learning deep learning generative AI and then with gener of algorithms that is really the final frontier there’s not anything else because gener of algorithms that is the category that allows for self-improvement so everything under the hood is all on the table for self-improvement and the reason I have a DOT for that final circle is That’s The Singularity so you’re essentially chasing the infinite as we enter this world of generative algorithms can we can we rep Place more algorithms more algorithms more algorithms can we be more efficient can we bring on quantum computers there’s not really another big AI breakthrough that’s needed uh something that I find interesting I I hope this gets people’s attention uh I have a 13-year-old daughter she’s interested in making money with gp4 and she was realizing that there is an arbitr intelligence Arbitrage opportunity on fivr so you have a lot of people that are asking for resume coaching they’re asking for code um copyright a lot of the jobs and tasks that they’re asking for generative AI can actually fulfill and so it’s it’s a wild world to live in to think that a 13-year-old kid could potentially begin automating a lot of these fiver requests away when people that were there before didn’t realize that um open AI was available to do many of these things and the the two slides I’m going to end with i I was talking to someone a couple days ago and my Uber driver and he was asking if all of this is a good thing for society to have um AI this cap AI of this level of capability coming online and I I thought this was a great thing because if you think about what you do for work you for many of us we are overqualified for the work we’re going to do tomorrow think about a surgeon that’s doing rounds um checking on patients posttop in the ICU a lot of those patients they don’t need their time and attention um a nurse can have good bedside manner they don’t they don’t need the cardi thres surgeon to walk in on checking on them if their data shows otherwise there are other people that need all the attention and so what you’re going to see in the future is you’re going to have experts and Physicians and clinicians that are focusing a lot more attention on the people that need it and then the other thing that’s going to happen and I’m getting myself so the the image you’re looking at this is a giraffe in in the zoo and this is us we are the giraffe in the zoo it is there are brief moments in our career where we’re able to work on things where we are over we are underqualified we have to learn we have to collaborate we have to work with a team we have to do research and Innovation and those moments are they can define a career they can give us Purpose with the future generation they will be able to do that every day uh definitely every week if not every day if they choose to they can work on Impossible problems with smart people they can be taught knowledge much faster than we were able to in a way that is also much more engaging and because of that you are going to see healthc care begin to transform when it comes to leveraging more data for better decisions um augmenting uh processes better bringing more sensors online because AI today has no issues with sensors a human is not we are not designed to assess an EKG signal very well at all but an AI system can do that much better than we can because it can monitor every heartbeat over long periods of time at resolutions we can’t see and it can comprehend what this means the other important thing in healthcare that we desperately need and we are not bottleneck with this at all from a technology perspective is this um it’s allowing for search so if I have a national or ideally Global patient level lake house where all the data has been anonymized but it’s in a format that is uh usable I can build machine learning models to predict these out disease outcome but more importantly I can do search so now let’s imagine a common a scenario that is all too common going back to my spouse house if they go into a clinician or if they meet a physician and if they are undiagnosed or this happens all the time you you can’t figure out what’s wrong with someone if there is a centralized vectorized database any physici can go there and with that patient level record that’s been vectorized and encoded they can stack rint the millions or potentially billions of patients that are the most similar to their patient and this doesn’t mean they are violating Hippa doesn’t mean that they’re violating data privacy they don’t even know who these patients are but they do know where they are they know which hospital networks they’re a part of they know which Physicians worked on them and so now you you can have a process where if I’m a physician with a patient that has an unknown illness I can reach out to five different hospitals similar patients I can talk to those Physicians potentially collaborate and then mask them to understand what was the outcome and so the one in a million disease that is celebrated today with the human um expert that will become the one in 10 million the one in 100 million the one in a billion where we can quickly find it and the tragedies of uh disasters like the viox um the viox disasters causing over 20,000 heart attacks in the wild that will not happen and so that it’s a very exciting future that we can look forward to the the other point I’ll leave you all with is as we refine and optimize these leveraging data with better Ai and algorithms in the health care space that will allow us to escape the hospital so the best type of Health Care is preventative where you can I’d much rather anticipate that I’m going to have a heart attack then have to react to it when I’m out with my family and so you’re going to see a lot of these algorithms coming into the home coming into wearables where we can uh really get ahead ahead of our Healthcare needs well thank you everyone I’m um this is a if there’s any questions that have come up during this presentation I do my best to be approachable you can reach me at json. Taylor dat IQ you can also connect with me on LinkedIn and I am a resource to all of you when it comes to thinking about Ai and taking on some of these bigger Healthcare initiatives and I thank you for your time and I hope you enjoy the rest of your conference