Don’t Fear AI: But Watch Out for These Industries and Jobs | Primer on AI with Stephen McBride
you think about technology disruption they are the centerpiece of every single thing that's going on in the world artificial intelligence is an Unstoppable Trend that you me everyone needs to get more familiar with to get a primer on AI I've asked my friend Stephen McBride to join us from Dublin Ireland he's the chief analyst at research firm risk Edge [Music] Steve and I always love getting together with you and you recently wrote something that that really hit me you wrote the coders using a software tool tied to chat GPT we're completing tasks in 55 percent less time and that was that was kind of the start of the aha moment for me when it started to come together I started to realize this isn't just going to be about coders or or programmers as AI systems get released as more a bespoke tools are developed this is going to affect almost every industry and it's going to go fast do you agree with that assessment is that what your thinking is yeah I mean one of my all-time favorite American Dreams and quotas there there's going to be two kinds of people in the world people who tell computers what to do and people who are told by computers what to do um just to your point about their the 50 you know this this new uh it's called copilot the deal that you mentioned uh by Microsoft it helps coders complete tasks and 55 less time and I think Tesla's uh head of engineering um he says co-pilot now writes 80 of the code for him it's almost like kind of writers getting the force draft out of the way and then going in and editing it to your point about co-pilot for every single industry right accountants and lawyers and editors and so on I think that's what's coming you're already seeing it PWC signed a deal to give their 4 000 lawyers um access to a new show called Harvey that does exactly that it it automates a lot of the run for walk right and you didn't go to law school to to kind of write up legal drafts right you you did it to you know win cases and provide value for your clients I think that's an interesting way to think about what's going on since the the kind of Arc of technology is that it allows humans to move up the value stack and focus on more important problems more value-add problems I mean 70 of Americans I think walked on a farm in 1900 now it's less than I think two percent yeah America produces more food than ever before that's how I like to to think about what's going on here um and you know if you look at the the productivity numbers for America GDP is really just walking Edge population plus productivity right that's that's what what uh what goes into economic growth total Factor productivity I think um from the end of World War II truth of the 1970s grew about two percent a year since 2005 we're growing at point two percent a year okay so this is one of the reasons and I know we'll we'll just get into the risks and the dangers later that's one of the reasons why I think America and really every country should embrace AI if there's a possibility that gets that gets us back to two percent a year productivity growth that is your way out of all the problems that we think about today the day and the the demographics and all that stuff so it's almost more of a risk not to focus on it yeah I agreed I I really have come to believe that this is going to turn into just an absolute explosion in productivity you probably know who said this first there's a quote going around uh to the effect of AI won't take your job but someone using AI will yeah I don't know who said it but I agree with it and and I'm sure you've read Tyler Collins the age of average is over and you know the adhesive that book is you want to upscale and you want to kind of focus on really really valuable things because globalization has you know increased the competition pill um the way I think about that is AI is the next leg of this and average in the AI age is people who do not use AI so all the studies now it's really interesting what they're finding is people who you know we're not great coders or we're not great editors when they use AI they're they get a bigger boost than the 10x coder or the 10x editor so this is really the new age is you have to find out how to use these tools in your own business whether you're a lawyer an accountant whatever these are really general purpose tools it's kind of like you know you as as a firm you wanted to set up a website back in the day or have an app or whatever this is really the next version of this so can you help people understand what other Industries you think could be on on the near Horizon that are going to be impacted by AI we talked about law accounting uh programming for sure what else is medicine going to be impacted yeah so I think this really gets at the fundamental misunderstanding a lot of people have about AI people talk about it in the singular right AI line whereas in reality there is many AOS um this and they all do different things so you know the AI that is going to drive your car is not the AI that detects cancer which is not the AI that can help you write articles my mental model is kind of its software that automates a process that could previously only be done by humans when I think about we talked about the productivity boom and how this can really move the needles for America specifically because I think America is going to be the country that Embraces this the most I always keep coming back to healthcare and education um maybe we can we control the charts later for for viewers but I love that I think it's from the AEI Institute uh it's the you know the the price increases and the price decreases of about 15 Goods since 2000 and the cost of tallies and all these consumer goods has gone way down on the cost of Health Care and education has gone way way up and you could argue that the quality has in some cases improved but certainly under education has it really changed in 50 years I'd argue no I think AI can help us finally break into these industries if you look at Healthcare specifically um I think it was researchers from Harvard Medical skill and the University of Copenhagen they recently built an AI tool which I successfully identified people at risk of pancreatic cancer three years before they were diagnosed with it again pancreatic cancer um 95 of people who get it die because it's detected so lame so that is one area where you could see massive breakthroughs and there's other medical image um medical medical image AI companies that are really doing this for every vertical another one is in drug Discovery cell I think the average uh price of researching and developing a new drug today is like a billion or two billion dollars something crazy and it can take up to 15 years well the thing about AI is that you can ingest mind bending amounts of data and you know kind of simulate molecules adding together you can speed up the time it takes and the cost it takes to develop these drugs so I think there's 19 AI drugs in the pipeline today up from zero a couple of years ago again this can really move the needle if it's allowed to and the other one is um what I call the paperwork problem so the amount of Administrators working at American hospitals is up 35 x since I think 1970 and some studies are found for every uh hour doctors spend with patients they do two hours of paperwork I saw this myself visiting my grandmother in hospital last year you know it's like it's like uh being time warped back into the 1980s or something like that there's just paper everywhere a lot of the systems aren't even on computers um AI can help with this right there's already companies that are basically have Alexa like devices that sit in doctor's office they annotate all the conversations that are taking place and they send you know whether it's prescribed to medicine or whatever true to the pharmacy just automating that process so the doctor then does not have to sit there for two hours writing out exactly what happened the other big one as I said education and there was a study in the 1980s uh it was called Bloom's two Sigma problem and basically the researcher looked and said how can we best improve children's outcomes okay and he found one-on-one tutoring was by far the best uh by far the best predictor of better outcomes basically it took your child from the 50th percentile average to the 98th percentile which is exceptional um but of course we don't all have money to get one-on-one tutors for our kids um just wasn't possible before with AI you make that possible okay you can have an AI that has interacted with your child from Bert and knows maybe they're a little slow on math but they're great on English and now Khan Academy Duolingo is doing something similar they're building super tutors you can think about it that you can sit in a classroom and you can be on something like an iPad or another device and it can interact with you one-on-one and it knows exactly what you're good at where you're lagging hey you know let's do two extra hours of math this week so this is a real game changer I mean you just think about I don't think anyone is Happy completely happy with the education system today this could completely change the game especially for lower uh income households um now the big question in my mind when I talk about this ad and just back to the the chart I referenced earlier is technology going to be allowed to flourish in these industries right because I've been reading about health Tech and AD tech for 10 years yet nothing really changes and when you look at these industries it's not for a want of trying it it's because technology has essentially been outlawed in these industries um Joel Lonsdale pilanteer co-founder who I think you interviewed a Strategic investment conference recently he had a great story about this that his friend was a um it was a doctor in the San Francisco Bay area and he tried to introduce some technology into the business to automate processes make the business better you would have been able to see supposedly 10x more clients he wasn't allowed to make those changes because of the American Medical Association so that's the real question here is technology going to be allowed flourish into these industries I hope it is and if it is we have a real shot of breaking down America's two most broken Industries that's really interesting as you've been talking I've started to think about just who might be at risk I mean you mentioned Administration in in hospitals education is is Rife with any if you know any teacher they'll tell you that their outnumbered by administrators in the U.S education system at every level from from elementary to high school straight through to higher education um and and then I started thinking about boy what could the impact be on government I mean if you start to think about applying AI to routine tasks and paperwork and I mean the head count in in government could drop dramatically it uh it the impacts are going to be I think a lot bigger than than most people are realizing at this point do you see AI as a job taker um because I'm starting to to come around to Kathy Woods argument of AI actually being deflationary hard to imagine in this inflationary environment that we're in but I'm starting to see the other side of that argument yeah look I mean I have studied this deeply I've thought about it deeply and um I keep coming back to the fact that no technology in history has caused mass unemployment now that is that's a that's a numbers game there's going to be some jobs that go away like you said there's going to be a lot more jobs that are created um Mit paper I think I found 60 of employment as in people you know people working in 2018 UH 60 of those jobs did not exist in 1940 so that kind of gets at the fact that this technology will create more jobs than it displaces another one that I love to say is um 270 occupations in the uh 1950 US Census just one has been completely automated away which is the elevator operator um and just think about how much the economy has changed since then yeah technology has continued to provide jobs one thing I you know it's it's it's always hard to see what jobs that we're going to create it's kind of as you said you look at the administrators in college you look at maybe bloated government and you say man those jobs they're going to go away it's much harder to think about the jobs to imagine the jobs that are going to be created because you have to imagine those jobs one thing I would cite very early on is that you might have heard of prompt Engineers they're basically people who get the most out of AI they know how to ask AI the right questions and get the exact right answers those guys they're in six figure salaries um and that job did not exist two or three years ago like I mean imagine telling someone hey I'm earning 200 000 uh talking to talking to a chat box they would have called you crazy yet that's gonna happen and there will be many more of those along the way one thing I would say is that you really want to you know having studied the history of this stuff again I'll go back to the Industrial Revolution before the Industrial Revolution skilled Artisans they made everything by hand and then of course machines and factories and all this stuff came along and low-paid labor people unskilled were doing their jobs um 20 or 30 years later yeah employment and that industry went way way up simply because demand for clothing and other Goods went way way up similarly um during the during the you know the Heyday of Detroit when they when Henry Ford introduced the assembly line Auto employment 10x I think over the next 40 years and again you would have said the assembly line that should put every one of these out of work because it made it way more productive I think it reduced the time to make a car from 12 hours to 90 minutes and again just because demand for cars grew so much faster the employment went up so when you're thinking about jobs that will be displaced and maybe jobs that will Triumph you really want to focus on those in areas where you can say productivity is going to go up but demand is going to go up way way faster so maybe if it becomes cheaper and faster to write software the world's demand for software will just go that much more but those jobs are fine so that's the that's the dividing line I think about when you know when you think about which jobs will will be safe and and those will be displaced okay great point so there's another side of the coin so one of the things that you've written about is that everybody needs to get some exposure to AI they need to start understanding how it works and how to incorporate it into their professional lives and maybe even their personal lives so give me an idea of how somebody watching today could do that how can they start to experiment with AI sure so two two-man two-man kind of avenues right now you have church EPT which is built by open Ai and you have Microsoft Bing which is essentially the same technology but integrated into a search at Microsoft search engine so if you go to if you open Microsoft's browser Microsoft Edge you can access Microsoft Bing through that I think you can you can it kind of pops up on on a sidebar and then any website you're reading say you're you know you want to go on and you want to read a scientific paper on whatever that is you can post the link to that paper in the sidebar and say Hey you summarize this you know 100 word dense academic paper for me into five Key summary points so I've I've spent uh I've spent endless amounts of time reading white papers over the years and all that stuff and it's just it's mind-numbing right so this this has I found it does 90 of the work and 99 less time another thing that you can do especially with the latest version of chat GPT which is gpt4 can handle up to 25 000 words at one point again you're reading a dance article or you just even even a small book um you can post the whole text into chat GPT and again hey summarize it and five sentences for me please and it will spit that out and if you don't like the answer you can ask a follow-up questions you know what breakdown Point number one for me more explain it like I'm five explain it in there the voice of an expert so there's many many things you can do um I know a lot of investors uh you know are followers of yours at one thing that we've started experimenting with at risk Edge is using it to analyze earnings calls okay so as I said you can post the transcript of an earnings call I'm the Q a into chat GPT and ask her to summarize the main points what stood out what should I ignore all that type of stuff um and this is just it's an amazing productivity tool that allows you to do more with a lot less how comfortable are you turfing off sort of some of that type of of work uh because you know there's there there's money on the line depending on what your conclusions are how much do you trust it do you find that you can teach it to to illuminate what you specifically are looking for how does that look you you know we've we've started playing around with it uh in with stuff we've already written about our earnings calls that we've already listened to Simply to see you you have to check its work right it's really like it's a junior analyst right just like a junior analyst comes to you and he says hey I read the earnings crime shift and here's the five million bullet points you kind of want to check his work so um that's that's really how we're thinking about today it does hallucinate still hallucination like if you ask um Microsoft Bing sometimes like why is Nvidia stock down here today it's over 170 or something it will give you an answer of why it's down year to date so there's still it's rough around the edges the key ad is that again it does 80 or 90 of the work in 99 less time um and just to put a fine point on this you can criticize it today but just remember this is the worst AI is ever going to be it's only going to get better so you know we're talking about using it here now is the time to use it right now is the time it's like using the internet when it was clunky and it was janky back in the the early 90s sure get the grips with it it's only going to improve any tips or tricks that you that you have because uh I I know several people just yesterday A friend of mine I actually uh share an office with him he was mentioning that he tried out uh chat GPT uh asked it a few questions wasn't really that impressed with with the answers and sort of wrote it off and I you know I thought to myself you probably didn't ask the right questions so what what is important to know when you start to uh play around with with it yeah look it's like any other technology right you really have to you have to learn you have to spend time with it two things that I would um I would I would encourage people to do is well first of all you need to the key is learning how to prompt so if nobody or you know if anyone watching hasn't played around with it it's essentially a a search bar that you type queries into and it's and it spits the answer back out that's called a prompt and the key to interacting with this stuff is learning how to prompt so you know one example I'd like to give is that if you're a high skilled student and you're tasked with um writing a essay about the Vietnam War you don't write a generic um question like hey write me an essay write me an essay about the Vietnam War because it's going to spit out a generic answer that's kind of not good and just to spend a little moment on explaining how this stuff works GPT and being another uh large language models they're essentially scanning the whole internet they're taking in this example everything that's ever been written or spoken about with the Vietnam War and coming up with a a kind of an average answer to deal with that okay well instead of just asking something very generic you want to ask it be specific so write a 500 word essay about the Vietnam war in the style of an American Soldier that's fighting uh on the battlefield and the mikon Delta for the first time ever and he's pinned down be specific with your answers whether that can be for anything it's really drill down and be specific and I'd also mentioned you can ask your follow-up questions if you don't get what you want ask it to simplify something ask it to elaborate on something it's it's a it's a conversation right and over time uh you will figure out how to use this stuff you're the one that I really found that has helped me is ask it to almost role play or play a Persona so again if you want to know what causes inflation don't ask hey GPT what causes inflation maybe play uh maybe play the role of an expert so hey GPT in the style of Milton Friedman or Keynes or whatever tell me what what the card underlying causes of inflation are in 1000 words and again I've found asking it in the style of an expert really helps improve the quality of the answers interesting okay so so far we've talked about kind of all the positives and then and the fun side of AI but there's a dark side too I mean you've got people heavily involved in AI uh Sam Altman Elon Musk you've got you've got people who really are are deep into it and have been for for years talking about risks and saying it needs to be regulated how do you regulate something like this can you regulate something like this so I think it's too early to regulate um I think Sam Altman is very smart talking about regulation because if you think about the Battle of the last maybe five years with big tech companies they kind of shied away and ignored the government um Sam Altman is the only complete offset asking for regulation okay so he's kind of coming to the table first if you will it's too early to regulate it simply because you think about how could you have regulated the internet in 1990 we didn't know how Amazon was going to flourish or Google or Facebook and similarly how could you have regulated the iPhone or the app market in 2007 or 2008 when Uber didn't exist or Airbnb didn't exist and all those type of things I think the best thing to do from a policy perspective is to take a hands-off approach give it a couple of years see what see how the chips fall see what type of apps are developed here that's that's the key because you can't know ahead of time um what's going to come out of this stuff I really think policy makers should do exactly what they did with the internet um which is to to kind of give it a Safe Harbor and of course the community Communications decency act in 1996 are kind of I think section 2 Authority was was in there and allowed Google and Facebook and and these companies to eventually flourish and it's no it's not a coincidence that America warned the internet uh vis-a-vis the rest of the world simply because they were the ones that took a hands-off approach you look at the amount of well good paying jobs of course there's downsides for the good paying jobs the amount of shareholder value that's been created there so I think it's too early um to regulate and I'll be honest with you do we expect um good regulation smart regulation from you know the people that have brought us the student loan prices and 30 trillion dollars and in debt and kind of botched the code response um I think you want to be very smart about this and not steiny future Innovation I I appreciate the opinion and I I'm sympathetic to it I there's not much agreement on this topic uh I was listening to the uh the all-in podcast a couple of weeks ago and uh those guys had just kind of a knock down drag out brawl over over uh how to regulate it should it be regulated each one taking a a different side from from full-on regulation right through to let's let it play out so it's uh it's gonna be a real challenge I think but but let's move on I I before we get to the end of this conversation I want to congratulate you uh because at the end of the day you know you're an equity analyst you're you're looking for companies with the potential to disrupt Industries and you have held Nvidia which is so far the number one winner of AI you've held that for three years so I want to talk a little bit about how you got into Nvidia like what what was the thesis that brought you to it was it was it AI way back then what did you find disruptive about it um tell me a little bit about about your process we originally started exploring a new year five years ago I actually wrote an article saying if I could buy one stock for the next five years Nvidia would be a I'm a long time uh semiconductor fan I found the most fascinating things in the world and they're really at you think about technology disruption they're at the set they are the centerpiece of every single thing that's going on in the world um and and that's what really you know back then we talked about it in the sense of machine learning um uh video games and and self-driving cars which is is really a subset of AI and it was really all running on it on Nvidia gpus we added the portfolio three years ago and you know and people watching this might think I'm I'm super bullish on AI and I am but I'm coming from it from a place of AI skepticism because for a long time people talked about AI I mean AI the term was invented in 1954 and direct myths and back then there was predictions about in three to five years we'll have a computer and that can do everything a human can do you know in the 1950s we envisioned humanoid robots um and and that didn't come to pass but really we saw back in 2012 uh early AI researchers they started to play around with Nvidia gpus and they made a series of breakthroughs forced on imagenet which essentially let a computer um recognize pictures of cats and other images and that was really in Nvidia fell into this industry those chips were created to to make better graphics on your PlayStation and now they're being used to power every um every AI system in the world so yeah the original thesis were was Ai and we looked at this and we said there's kind of like you know people know Moore's Law or metcalf's law and there's a new AI law that's emerging and it's essentially when you 10x the amount of data that you put into these models you 2x the quality so what has happened as a result of that is that you get larger and larger and larger models so Moore's Law I think it doubled every 24 months the amount of uh AR the amount of data going into these AI systems is doubling every six months I think it's up wow one million x over the past decade so this whole thing it just requires more and more and more gpus and which is which is what exactly what Nvidia makes when you look at its business it's really been transformed in the last five years or so so I think when you know a couple of years ago 30 of the sales were from data data center AI related and now it's 60 60 or so so I think obviously it hit the trillion dollar Mark um it's it's incredibly impressive it was a lot easier to own it back then than it is now I still think it has room to run I it almost feels contrarian to buy it here um and when you think about from doing research into if it's open AI if it's mid-journey if it's stability everyone is just clamoring far as many gpus as they can get their hands on and you know when you think about the chip exports it's almost like gpus that are the new weapons of mass destruction right wmd is they're trying to limit this stuff um it's incredible how important they have come but bringing it back to to how investors should think about this um I think it really I think it has still has room to run one thing that I'm looking for is almost a story signal of maybe when to sell I'm struck by how many times a hot IPO marks the top of a specific market so AOL Time Warner merger in 2000 I think it was that kind of marked the top uh glencore or Blackstone going public in 2007 marked the top of real estate uh Glenn Corgan Republican 2011 marked the Commodities top it looks like coinbase APL marked the top for crypto so far maybe open AI maybe anthropic and one of those hot uh private AI companies maybe that's the time that's the thing you should look for uh when to sell Nvidia but as of right now all that money chasing all that that those dollars institutional retail chasing AI wanting the allocated AI it's really only one stock that you can buy right now so you know I I can't resist one last I always have to bring some geopolitics or macro in into the conversation you said one thing so the Nvidia makes the products what I I don't think that's the case right in Nvidia designs and develops they're an engineering company but where are their chips actually fabricated yeah tsnc and Taiwan okay look um I've listened to to Louis garv and and Michael poppich and geopolitical experts people that know a lot more about this stuff than me they put the risks of a Chinese invasion of Taiwan very low and there's many reasons that they that they talk about this stuff but one is to just on a five or ten year time scale look at what they the Chinese did in Hong Kong and it was essentially to move Chinese Nationals in and over time you know take over the country it's not an all-out invasion um and uh you know tsnc seems to be it seems to be kind of building fads in America it's looking to do so in Europe now so maybe that um that diversifies the risk a little bit but yeah as of right now Taiwan is a is a very important um centerpiece in this operation which you know investors of course do your own due diligence but uh tsmc is an interesting um stock in this all as well because it's the only company in the world that can make these Cutting Edge chips if you want to know why I brought up where nvidia's chips are manufactured check out this video that I recorded with Washington DC lobbyist Bruce Melman last week we cover the United States addiction to All Things China and why that is going to be a hard addiction to break
2023-06-16 07:38
