How AI & Chat GPT Will Change Accounting with Randy Johnston

welcome to the xiaomfield 2. with us one of the legends of the accounting and fintech image Randy Johnson good data well good day rob it is such a pleasure to be back with you and your listeners again Randy we have had you on this show a number of times and you cover that wonderful gap between people and Technology you talk a lot about strategy to Big firms and big vendors and it encompasses the people in the technology I want to talk today about the new Revolution if you like artificial intelligence chat GPT it's really gathered momentum over the last few months you're an ideal person to talk about it take us back Randy how did all of this get started how did we get to where we are well Rob I appreciate uh I do like walks Down Memory Lanes you can just look at them on my columns and see like the history of cell phones and the history of Wireless because I write that stuff up and frankly I've been fortunate enough to be around many of the creators of these things through the years uh in the case of AI I haven't been around the Creator boy I'd like to have met him uh out of the UK guy by the name of Alan Turing a brilliant computer scientist which has lots of uh history of our listeners have not got background on all and Alan they should but he actually started working with AI in 1950 which I think Rob might predate both of us just a little bit so as it turns out uh over time I have been interested in artificial intelligence Rob since the earliest part of my careers most people don't know that I was actually writing computer codes starting about 1961. so I've been doing this a long time but the first computer vision came out in 1963. so you know I know with some of your history on vision and some of mine computer vision being artificially intelligent so supported is a big deal and frankly by the time we got to about 1976 as I recall there were some early voice recognition that was taking place so you know a lot of people are thinking this is all current Evolution well 76 I hate to say that's 50 years ago right and so but you know just kind of a a laughable thing that some of our listeners will remember Microsoft came out with an artificial intelligence-based operating system in 1995. it was called Bob Microsoft Bob I don't know if you ever ran Bob but it's a great operating system and it did some weird stuff it was almost as bad as clippy in those days but you know as it turns out Microsoft was trying to figure this out well if we continue through this journey neural networks started to get popular in about 2010 so about 15 years ago at this point and they were being worked on before that and the neural networks were some of the earliest of the machine learning and artificial intelligence now I won't get too far into computer science but machine learning and artificial intelligence are clearly separate fields and neural networks are one example of these but the neural networks were used to create the Watson product from IBM and it's hard to believe but it's been over 10 years since Watson wins on Jeopardy if our listeners remember that and it was also in another little neural network product that some of us use Siri from Apple was released so 2011. well so I just want to keep unfolding this because if you think about the voice recognition on our personal devices Google home Alexa Siri and so forth those are all machine learning types of approaches but within the last uh two weeks before recording this session for our listeners IBM updated Watson with Watson X and they actually extended this with generative Ai and this has really been the Breakthrough because in roughly 2015 the open AI company was formed as a open source artificial intelligence platform and then by the time we got to about 2019 it was taken private that was a little bit of a concern on some of that but Microsoft made Investments of you know a billion plus in this Consortium Elon Musk was there and of course elon's going to try to do a separate artificial intelligence but the open AI company then releases this generative AI product against their wishes by the way on November 30th of 2022 now what happened is there were several versions of chat GPT that one was 3.5 so there was a 2-0 and a 1-0 and all those beforehand but the 3.5 started to get in the wild it was not supposed to be released but people who had accessed the tools started publishing it so open AI decided okay we're just going to officially publish it and so that tool got updated very quick from November until roughly March of 2023 and on March 14th of 2023 chat GPT 4 became available now we'll talk with your listeners a little bit about what makes these models difference but here's the net for what all of the hype is around this these generative AI models use a technology known as a large language model or an llm and they basically have crawled the internet to accumulate pretty much everything published good or bad and then they use that to tokenize our searches and produce answers and it is just a big game of Statistics is really all that's going on but it's a very accurate game of statistics and because these large language models have so much content it's actually pretty hard to ask things that they don't know with this exception today the current releases of chat GPT 3.5 and 4.0 are cut off at September of 2021. so if something happened after September 2021 your search may not quite be accurate but you can go to Microsoft Bing which is now incorporating everything current and get all of the open AI chat GPT features up through that point so it's a long rant to get the history in there but that Arc makes a lot of difference because generative AI doesn't mean that AI has a conscience they're not patient if you will they're not complete like we would like to see some of these but they're a magnificent step in the last two to three years thank you Randy it's a phenomenal Journey you've just walked us through to how we get to where we are and it brings us right up to date I'll end by asking you the future of this and where it's going you're very good at things like that but let's make this relevant for accountants and bookkeepers and accounting professionals CPAs listening how is it affecting their world yeah and I appreciate the long rant on the history because that is exactly where I went go next um it is critical in my mind that accountants research and use artificial intelligence there are absolute risks there are absolute benefits and I want to pull the risk out first because the Privacy models of these generative AI tools and it doesn't matter whether we're talking Chad GPT three five or four or Bard or any of the other well let's see in in 2022 there were 29 commercial models developed in three University models so there's dozens and dozens of these tools out there not just chat that we're kind of focusing on but chat gpt4 takes what you provided and plugs it into the language model too so in effect if you put client confidential information in the generative AI models that becomes part of its indexed engine so you can't put company confidential information out there like we've seen done with Samsung where they had designs that were leaked through this methodology you can do beautiful things like cash flow statements and projections so you put in balance sheet and P L and you know cash flow and say you know project this out for two years boy it does a beautiful job of getting that work done even though Chad GPT isn't real good with numbers it will still do those projections but if it's labeled Rob browned world domination Inc uh you know it will identify that and everybody can look up Rob Brown's world domination Inc yeah financial statements that's on track as well I wouldn't want that to get out Randy so thank you for that because I thought you'd want to keep those things but you're right to bring up the privacy because accountant CPAs they're in the business of risk management aren't they Randy so they've got to be super careful they have so many data points so many points of information that their Guardians stewards over they've got to be so careful they really do and you know uh Italy responded on March 31st of 23 by Banning chat GPT and then they opened it again on May 5. once Chad had made some modifications but the EU and many other jurisdictions Canada and so forth are still quite concerned about this rightly so now there was an initiative in an open letter that was issued earlier this year about pausing AI development and you know that's like trying to close Pandora's Box in my mind I believe that what happened there is companies were caught flat-footed they wanted a chance to catch up and so they've taken the you know privacy and how do we harness this thing argument as opposed to saying look this is a big deal yes it probably does need some regulation around it but you know pausing it doesn't make sense and frankly I almost hesitate even as I'm explaining this to your listeners because AI is moving and changing so fast things that I say today could be non-factual in you know the next 24 hours so do check my facts carefully as we're talking about this but this risk management inside chat GPT is great and remember these platforms will make stuff up that's called hallucinating in the jargon but you know it will give you some great answers and then if it doesn't know anything more it'll just start kind of doing Babble and you you got to be able to filter it so we recognize the threats and the risks to accountants but there are clearly many benefits as well for accountants doing this before we get on to those benefits I want to ask you generally how Ai and chat GPT will adversely affect the profession we hear of accounts that's been put out of jobs the robots taking over I've had now that AI can complete the CPA qualification the childhood accountancy qualification it can pass the bar exam so are we under threat Randy yeah that is an interesting question because you are correct uh and it can also pass the medical exam as well so you know when you start thinking about all of the professions being at threat now it turns out uh closer to your home in Europe there is actually an AI legal piece that's been around for a couple years trying to to work this through however I think there'll be a difference Rob between accountants that use Ai and accountants that don't use Ai and I'm going to use myself as a bit of an example here because in my work so far in 2023 I do things that you know a lot of people say you shouldn't do I still keep track of my time with time sheets because I want to cost things but it turns out the work that I've done this year I'm doing in one quarter of the time of the work in the prior year now you know I'm not saying my work is great but I'm doing it you know four times faster and with probably higher quality as it comes out so here's the net I believe that if you're an accountant you can start with the generative AI tools to get guidance and then you can apply your expertise to tune it up so I've done very simple things Rob like generate the monthly closing steps for QuickBooks and it just builds them and they're perfect they're just and you couldn't do the research because it's a 15 second response and it's actually right um You can do things uh like uh explain the mileage rates and how it's applicable on the law in various U.S states or you can do privacy uh and Security in different jurisdictions so if you're entering an alternate jurisdiction you can actually get guidance and almost anything if I'm going to start with a blank piece of paper I now start with the chat GPT generated AI response and say okay that I'll kind of take it from there but the other thing that I'm going to recommend to all of our listeners if they is that they build a model I've been using a spreadsheet to do it where I tune the AI and treat it like an assistant and if you do that and you realize that you've got this assistant but you have to coach it uh I actually give it styles of my writing for example and tell it that I want a you know a 700 word or a thousand word piece and it'll write in my style and I have some idiosyncrasies in my language you can hear it my speech too that you know it just does it and I can't tell it's not me and the detection tools that say this is the AI generated they don't detect it either so just begin thinking about what you produce for a client and the manual work to get there and realize that it will all get done by the bloody AI tool so will it reduce the job release the job now it'll change the top job though and if you're an accountant trying to do this without AI you're just working hard instead of smart yeah and we've got you coming up on a future episode Randy to talk about casts these advisory Services I suppose if there's any some on the horizon for the accountants with the Advent of chat GPT and AI it's how it will free them up hopefully to be that trusted advisor and to get beyond the numbers and tell the stories behind the data using the analysis and research that chat GPT can provide you well if I got that about right you do have that right and the the results of the analysis are actually pretty stunning but remember every once in a while these tools will hallucinate right now so these uh characteristics by the way are known as emergent characteristics and uh let me think here for a minute because there is a uh Google researcher I believe it is uh actually I'm just going to see if I can recall that quickly sorry about stumbling around on this friends but the emergent technology or emergent pieces are very very important on this so so um about 137 I think are the current numbers on that at the risk of sounding like it's odd and it's actually open AI so it's attribute it correctly Jason Jason Wei who's a researcher at open AI has counted 137 these so-called emergent abilities across a variety of these large Which models so you start feeding it financial data and ask for prediction it can do that you feed it something in the language and ask it to translate it even languages that these platforms have not been trained on they can do and of course in the last week or so before our our time together today Facebook meta actually released a language translation piece uh you know so I I just look at the various GPT models that have been created too I expect a accountant GPT model just like there is oddly enough a Bloomberg GPT model already and there's already a Quran GPT model and a Bible GPT model and basically what's happening is they are feeding very specific things into these models and getting results so again for our accountant listeners begin thinking about the Corpus of knowledge that you have in your firm the way you do business the way you make your recommendations and feed that into the generative AI model so it thinks like Rob Baron world domination accounting firm you know and what happens is it will actually look at the world the way you do so on advisory the that's exactly what I expect to start happening if if you can say in this situation I would and can feed that in the model you teach your assistant here and it will produce those type of results too yeah this fascinating the outlook here what would you say to accountant CPA's busy bookkeepers that are already overloaded Randy they're overwhelmed or stressed out tax season just gets longer and longer and this is just one more thing for them to master one more thing for them to learn I've now got to be a chat GPT engineering prompt specialist I've got to know all about this this and this I'm too busy looking after my clients I'm overwhelmed what would you say to them yeah for those your promise is truly great mechanical things that you've been doing you're now going to have well we'll call it a robo assistant an AI assistant that'll actually do a lot of the things that you've been struggling to get done so you know let's take very simple non-ai things like OCR extraction so if you think about what happens with the likes of Dex prepare or Hub doctor or Sage auto text and so forth uh big Ai and and on and on these tools have been trying to automate the transactions and classify them just picture for a minute that your AI assistant classifies almost without error all of these things there's a tool that's not very popular around the world but it is more popular here in the US called uncat that can kind of do that but now if I fire this bloody thing with uh you know generative AI technology all of a sudden those types of work that you're doing not not such a big deal so you'll be judged yeah so if you begin to think that all of what you'd consider mundane work in tax preparation mundane work in account counting in bookkeeping and controllership type Services I really expect that to be very highly automated so here's the great great promise Maybe we're at the point that automation will finally help us with this Staffing shortage that is a global problem and just for the record I know of at least one CPA firm here in the US that I trust very much that I've been around a long time that is measuring as much as a 10 percent Improvement in productivity there are they're a big firm they've permitted it they have some restrictions they have some rails around it they're seeing a 10 percent increase in productivity it's a 600 person firm so they think they they are 60 people better off if you want to use the numbers like that and again you heard my numbers I'm doing things four times faster it's taken me 25 percent of the time so it's like you've got not quite but you've got about four Randy Johnston's doing the work now instead of one right it is local domination is on track for you Randy if you can clone yourself that that's astonishing it's great that how you stay so connected to what is going on and stay so relevant with all the things on your plate I know you consult to a lot of the top firms and vendors doctors about the marketing aspect of artificial intelligence and chat GPT you've touched on the talent aspect and we're all looking for more capacity how can we attract better people but we want to attract better clients the right kind of work that's going to benefit our firm how does AI plates roll their ending yes uh here if you know and have defined what you want your client to be and by the way you can use chat GPT to find opportunities for clients but that's a that's a separate conversation perhaps but as recently as yesterday Rob I took a firm's basic Niche marketing and Specialties that they had fed those items into chat GPT and ask it to generate a marketing plan and it did a beautiful 10-step plan and I had it right web pages then I had it right promotional literature then I had it right social media social marketing posts and as it did it you know I'd give it a little bit of coaching no not quite like that more like this and in less than 20 minutes I had built a plan that in historic times probably would take me the better part of two to three days okay two to three days 20 minutes now was it perfect no but was it good enough that I could have then apply my expertise to it yes so here's the deal if you know what you're trying to do and you can summarize it in a paragraph or two and you maybe already have it on your website take that text from your website feed it to chat GPT and then Coach it just like you would with an assistant no not that do this and you will be able to generate superb starting points for marketing now marketing experts clearly are going to uh you know do little nuances that we as non-marketing experts won't get but you know how it is when you've picked logos before and you know a graphic artist gives you like six choices and you say not that this and so forth it's the same type of thing yeah exactly and what happened with the plan yesterday I had a marketing expert online with me and I was generating in real time with that marketing expert and she said at the end of it she said you know I feel really threatened now because you've done what in in 20 minutes what typically would take me as much as a week to do and she said it's about as good as what I would do and that's where it takes to run it takes us to the very edge of human capability and Beyond and all that's left is the the people side of it the relationship side the emotional side the trust side the advocacy side all of these things were marketers can get behind the emotions and the personalities of a brand in a way that perhaps artificial intelligence can't an accountant can get behind the complexities of people who are very unpredictable creatures and relationships which chapter could never do and that's the Hope isn't it yes it is and and actually that's kind of the next Frontier because today the models were created dominantly by web crawls done between 2016 and 2019. and the intent is to put more of these emotion attributes in there so this General artificial intelligence you know I've said it was 20 30 20 40. I'm actually shorting shortening my predictions on that because of the breakthroughs here with generative AI but that is when it will be even harder to discern that you're talking with a real person or whether you're talking with an AI engine and over the last six months the ability to speak to these AI models is coming into place and they're able to speak back and you can actually pick voices and that's why you know what I would call uh deep fakes which you've heard about over the last two or three years both images and video but also voices uh are a real threat so again I I know that we need some rails we need some controls around this because all technology can be used for good and bad but as it turns out the Bad actors are using these for all sorts of bad things and we as good actors can use them for all sorts of of good and great things one of the big super bands in the world are Abba maybe you've heard of them Randy and there is now an abattoir which are Holograms of the original band members in their Heyday doing the original songs and they look like they are actually on the stage performing all of the hits as they were and they're charging good ticket prices to see this tour worldwide and do you think this augmented reality or virtual reality whatever this is It's A Whole New World coming up so let's finish by asking you if you cast your mind to the Future what is coming up particularly in the world of accounting and the fintech industry that supports them yeah and you know I think it's fascinating that you use the hobby example because I was thinking about the metaverse example out of the UK that I saw just yesterday okay we're teaching students and yes the metaverse was the really hot thing before generative AI came along and metaverse is still going to come along so which described metaverse for people that are not familiar with that or anything oh very good metaverse is a combination of typically 3D reality that you can use to build digital Twins A Digital model of reality and then you can apply immersive technique so in many cases today that includes goggles but that is going to also have projections and so forth you know Rob that is such a big topic we could spend two or three sessions on it but you're you're uh allusion to it there is actually where I think this is going a combination of immersiveness that is so real that you can't tell it's not the real deal combined with AI you literally will interact with closest friends and family from afar you know today we've got flat video conferencing if you want to think of it with the zoom and teams and all that stuff but now take it to 3D projected reality and like has been done here in the uh us and it's actually been done some globally if you just think about the Holocaust and some of the recordings that have been done with the the Holocaust Survivors you actually are applying Ai and they're talking about their experience but it's so real you're going back in time uh that way so I think AI will model the past and it will help you imagine the future and to me that's what AI has done for me in the last oh year or more even though I wrote AI back in 1976 personally uh because I was using lisp at the time so I've followed this dog on stuff a long time that here instead of me starting with a blank piece of paper I'm now saying what about and then I let AI become kind of my bouncing uh you know bounce the ideas off but and I get so much better results more rapidly now the one thing I worry about is that it Narrows my thinking too quickly because I've taught people for years you're still sometimes better off just sketching on a piece of paper or yellow pad white pad whatever and getting some ideas generated that way just like we've talked about business models in the past a while Rob some of those things these models won't do for you but if you have to produce a result man it's so much quicker does not start with a blank piece of paper so you're hopeful Randy you're excited we can see it and you're facing your voice you know I am so pleased about this thing finally coming of age I absolutely understand the threats uh many many more that we haven't talked about like hacking and you know social engineering and so forth but so many things that can be done for the good and you know I always want to see people use technology in the way that they want to have it benefit them the most and I can't even imagine all the benefits that are possible now with the new AI models for yourself and for your clients well Randy Johnson that's been excellent we will check in with you later in the year when the whole game will have changed and we'll be so much further on and you'll be able to report back then but for now thanks so much for your passion and your insights that's incredible thanks Rob great to be with you again and all your listeners talk to you again [Music]

Channel: Accounting Influencers

Views: 3,483

Keywords: #AccountingInfluencersPodcast, #ArtificialIntelligence, #AIevolution, #GenerativeAI, #DataPrivacy, #RiskManagement, #AIRegulations, #AccountingTechnology, #Confidentiality, #AIHistory, #AIUtilization, #PrivacyRisks, #AIOutputFiltering

Id: 6pk7ysoSePs

Length: 30min 36sec (1836 seconds)

Published: Mon Jun 12 2023