Sustainable Supply Chain

The Supply Chain Game Changer: ChatGPT Explored

March 06, 2023 Tom Raftery / Doug Marinaro Season 1 Episode 298
Sustainable Supply Chain
The Supply Chain Game Changer: ChatGPT Explored
Digital Supply Chain +
Become a supporter of the show!
Starting at $3/month
Support
Show Notes Transcript

On today's episode of the Digital Supply Chain podcast I asked Doug Marinaro to come back to talk to us about the impact of ChatGPT on Supply Chains. It was a fantastic conversation that covered a wide range of topics, from the impact of ChatGPT on the supply chain to the future of human-computer interaction. And for the first time, a video version of this podcast is available at https://youtu.be/UB8HQ-ZfjYo

Doug and I discussed how ChatGPT has changed the game making it easier and more efficient to get the answers you need. We also talked about the different versions of ChatGPT, including the free version and the premium ChatGPT Plus (which I highly recommend).

We covered was the impact of ChatGPT on the supply chain. Doug believes that this technology will be incredibly transformative, fundamentally changing the way humans interact with computers and each other in the supply chain. He also shared his excitement about the future of ChatGPT and what it might hold for the industry.

Finally, we discussed the importance of trying ChatGPT for yourself. Doug encourages everyone to dive in and come to their own conclusions about the technology, as the 99% of the amazing awesomeness it has to offer will be what you experience.

Btw, at Doug's suggestion I asked ChatGPT if there was any question that I didn't ask, and should have, or any aspect that we didn't touch on that you think it is important for people to think about?

ChatGPT replied:
From the podcast transcription, it seems like the conversation between you and Doug Marinaro covered a lot of aspects of ChatGPT and its potential uses in the Supply Chain industry. However, there are a few questions that could have been explored further:

  1. How is ChatGPT different from other AI-based systems in the Supply Chain industry?
  2. Can you give specific examples of how ChatGPT has been used in the Supply Chain industry to improve effici

Elevate your brand with the ‘Sustainable Supply Chain’ podcast, the voice of supply chain sustainability.

Last year, this podcast's episodes were downloaded over 113,000 times by senior supply chain executives around the world.

Become a sponsor. Lead the conversation.

Contact me for sponsorship opportunities and turn downloads into dialogues.

Act today. Influence the future.



Support the show


Podcast supporters
I'd like to sincerely thank this podcast's generous supporters:

  • Lorcan Sheehan
  • Krishna Kumar
  • Olivier Brusle
  • Alicia Farag
  • Joël VANDI
  • Luis Olavarria
  • Alvaro Aguilar

And remember you too can Support the Podcast - it is really easy and hugely important as it will enable me to continue to create more excellent Digital Supply Chain episodes like this one.

Podcast Sponsorship Opportunities:
If you/your organisation is interested in sponsoring this podcast - I have several options available. Let's talk!

Finally
If you have any comments/suggestions or questions for the podcast - feel free to just send me a direct message on Twitter/LinkedIn.

If you liked this show, please don't forget to rate and/or review it. It makes a big difference to help new people discover it.

Thanks for listening.

Doug Marinaro:

the ways that some organizations are going to adapt to it and are gonna be able to take advantage of it faster than others, which will give them the ability, for example, to communicate, effectively with far more customers in far more scenarios than they have before to give them a step change, improvement in customer satisfaction

Tom Raftery:

Hi everyone. And welcome to episode 298 of the digital supply chain podcast. My name is Tom Raftery, and I'm thrilled to be here with you today. Sharing this very special episode of the digital supply chain podcast. Why is it special? Well, this is the first episode that is being published, both on YouTube. As well, as on the normal digital supply chain channel. Not only that. But we're talking today about the latest hotness ChatGPT. But before we dive into today's show, I want to take a moment to express my sincere gratitude to all of our amazing supporters. You're supporting this podcast has been instrumental in keeping it going. And I'm really grateful, sincerely grateful for each and every one of you. If you're not already a supporter, I'd like to encourage you to consider joining our community of like-minded individuals who are passionate about supply chain. Supporting this podcast is easy and affordable with options starting as low as just three euros. That's less than the cost of a cup of coffee and your support will make a huge difference in keeping the show going strong. To become a supporter, simply click on the support link in the show notes of this episode or any episode. Or visit tiny url.com/d S C pod. Now without further ado. I'd like to introduce my special guest today. Doug. Doug welcome back to the podcast. Would you like to introduce yourself?

Doug Marinaro:

Sure Tom, glad to be back. I'm doug Marinaro. I'm CEO and founder, co-founder of Riptide three-Way text messaging platform for service and delivery.

Tom Raftery:

Okay. Superb. So tell me, Doug, we had this conversation before all about Riptide and it was a fascinating one on the, the three-way messaging platform that you have, but we've since had a bit of a change in the, , in the world , if we wanna put it that way. As, as grandiose as that, the world has changed. We've had the the launch of ChatGPT, and we've seen how that has changed the world utterly in all kinds of ways. And so I thought it would be interesting because yours is a very text-based, service. So I thought it'd be interesting to have you back on to talk about how the likes of ChatGPT and G P T AI models in general are likely to change supply chain. So it's a bit of a broad one, but, how is it gonna change supply chain Doug?

Doug Marinaro:

Oh, in, in, in so many different ways, Tom, in so many different ways. You know I think the supply chain is, the means by which people manufacture and distribute goods and services. But I think in, in, in many ways the, the supply chain is a, is uniquely human activity. At every step of the way, people are involved except for maybe in a lights out warehouse. You have people communicating with each other, about the design and manufacture of the things that they're gonna be making about the, the distribution of those goods, about the, uh, availability and distribution of those goods to, and services to the individuals, to the consumers in the end. And every step of the way, you have communication that's taking place over both, you know, time, in different time zones and, and, and space, you know, geographies, you know, happening from place to place and, and, and cultures, you know, different languages, and, and all the systems that exist out support the supply chain, whether you're talking about SAP and Oracle or a, or transportation management system, or a last mile delivery or a customer support, um, at every step of the way, there's human computer interaction or human human interaction, Mm-hmm. and the ChatGPT, tool and discovery, has, revolutionized human computer interaction. So I think it's gonna impact the supply chain at, in, in, in ways that we cannot even imagine at this time.

Tom Raftery:

Yeah, we should mention as well that it's not just ChatGPT, Microsoft launched their new Bing, uh, recently. I'm not sure if it's generally available yet, but it is, is certainly available you can apply to, to get onto it. And they've had some teething issues with it. Google had serious teething issues, 170 billion dollars worth of teething issues their Bard launch in Paris a couple of weeks ago. But in general it speaks to a direction of travel. It means, I, I mean, I did say initially that the world has changed and that may sound a little bit over the top, but I don't think it is. I don't think it is. I think the way we interact with computers is fundamentally gonna change off the back of this. Uh, am I exaggerating or, or do you agree?

Doug Marinaro:

I agree. And, and you know, in our, earlier conversation, Tom, you talked to me about how you were one of the early users of, ChatGPT is of these large language models and, uh, I, we gave some pretty good examples of how you were able to, to use it and tap into it right from the get go.

Tom Raftery:

Yeah, yeah, So, one of the first times I tried it, I, as you say, I was one of the first on, I was there at the, it, it, it, it launched on November 30th, and I, if wasn't on it November 30th, I was on it certainly December 1st or thereabouts. And I remember. going to it and being amazed by it being blown away. And I tried to explain it to my younger son. He's 16 and he wasn't quite getting it, so I pulled it up and I said to him, okay, tell me what you're studying in maths at the moment. And it was, some kind of functions. I can't remember. So I just typed into G p T ChatGPT explain these functions, in the voice of Hagrid talking to Harry Potter and out came this fabulous explanation. You know, It read literally just like something out of a Harry Potter book with Hagrid talking to Harry Potter. It was, and my, my kids' head just exploded. You know, it was just. Now that's not, that's not a particularly, useful use case. I, I, I gotta think for the likes of you and I, and for people who are in supply chain listening to this, but it, it still, it speaks to how imaginative, ChatGPT can can be, or, or large language models in, in general. I'm using it all the time, and using it all the time on these podcasts. And how, how do I use it on these podcasts? Well, we'll, we'll come to some pre-stuff pre-stuff in a minute, but I use it on post all the time. And what do I meant by that? I mean the, once I've done the editing, what do I do after that? Well, I edit using a program called Descript, which is a, you, you throw the audio file in and the descript app converts the audio to text. So it, it does a full transcription and then you can edit the text and it'll output edited audio. So so that that's how I do the editing of these podcasts. When I'm finished with the editing, I can then export the audio and publish it. But what I can also do now is I can export the transcription and give it to ChatGPT and say, Okay, if I were to write a LinkedIn post to promote this, what would it look like? And boom, it'll, it'll give me a sample LinkedIn post for a promotion of the podcast, you know, or, write 15 tweets for me to, promote this podcast and write them in a, uh, first person, engaging tone, and be sure to use the word supply chain in the tweet, you know, or something like that. And leave room for a link and an audiogram, and bang, out come, 15, uh, potential tweets. You know, that kind of thing. It makes generating social copy. And it's all based on the transcription of the podcast, because the transcription is text. You can feed it to it. You can say, there you go, that's the podcast. Go for it. Help me generate social copy to help promote this. And boom, you know, the output that I've been able to create as a result of that since ChatGPT came out has been incredible.

Doug Marinaro:

First off, it's a great, great example. And I, I think, it, goes to a number of, of, I mean, there's a number of topics that come off of everything that you just talked about right there. First of all, for, for your listeners who want to try it out, uh, yes. I think Microsoft, um, Bing may be expanding its use, but they can also go to openai.com and at the top it'll say, try it. And you can go ahead and click on it and you can just start typing in questions. And so, you know, kind of maybe taking a step back and helping explain to your listeners, when they start to use it, what, what's it, what is, what's happening? You know, what is this AI right doing here? Yeah. So, you know, when you're interacting with ChatGPT, you are interacting with a, um, uh, what, what's, what, what g p t stands for is, I mean, I think I wrote this down, is, uh, generative, pre-trained transformer. And, what these are is, these are very, very large math models and the purpose of the model is simply to predict as accurately as possible the next word that it should say. Yeah. So if you think about when you when you, when you ask it to go ahead and generate a tweet for you based upon reading the transcript, uh, what it's done is it's, it's taken, the, the knowledge that has already been in inserted in it has been embedded into the ChatGPT model. And, and that knowledge in, in the ChatGPT 3.5 was, was roughly the, the corpus of all human knowledge, everything that had been put up on the web you know, all the books that had been put up there, all Shakespeare's, plays and so forth. uh, and, and in many different languages, um, as well. And processed, you know, exabytes of information processed with petaflops of, of computation. As, as I've said to a couple people Moore's Law finally has something to do again. As opposed to giving us 25 megapixel cameras, which, you know, we don't know what to do with anyhow. But, you know, so it's, it's processed all that information and, and the output of all that, of, of this is a 175 billion parameter model. Uh, just a side note, um, the human brain has about a hundred billion, neurons and, and or neuro neuronal connections, I believe, uh, or neurons, not, not many connections. Uh, and you know, so there's, there's there, there's a slight analogy there. And with 175 billion parameter model based upon all this text that people have written, what it does, it takes the conversation that you've started and it, the, the, the prompt that you've given it and all that information, you've given it up front and it tries to predict what's the next word it should say back to you? Yep. Mm-hmm.. And it, and it is doing that like one word at a time or one sentence at a time. One set of phrases at a time to go ahead and try to predict that. And so, in, the end, what the discovery of that, that ChatGPT represents is that with a model of about 175 billion parameters, trained with a neural network, with some, human feedback at the very end, after they ran through this, all this training, they actually went through and ran the model a couple times and had an answer some prompts, had some humans going through and said, no, that's not the right thing to say. And they, they, they improve that, that with that you can actually create something that, does a, a really remarkably good job of representing and, communicating in human language. Yeah. And because human language is at the basis of, you know, all of our societies, and, uh, and all of our interactions, this is a pretty amazing discovery. And it's an engineered tool. I mean, this a feat of engineering, you know, and, and the same vein as, you know, Crispr as a feat of genomic engineering. And, and, you know, this latest laser fusion capability was a feat of engineering. this is another feat of human engineering to be able to give us this tool that we can use now to generate responses to things that people input in human language, in a way that people, to a large degree won't be able to differentiate between, you know, whether it's a computer saying it or a human saying it. And that's, the remarkable thing. And, and I think that, what is, what is always amazing about, uh, people is, these new kinds of technological revolutions occur and they've been occurring at a faster, faster rate. But is, is how fast we all adapt to them. My next door neighbor is , is a high school English teacher. And of course the ones who adapted this thing immediately in December, were his students who were, who were turning in essays generated by ChatGPT in response to the, the questions he had posed. And so of course, you know, he figured out cuz there was something created ChatGPT on how to determine whether the essay was generated by ChatGPT, and he started using that. And then of course, he turned around and like a day or two or later started using it to generate his lesson plans. You know, when he was, you know, gimme five topics to go ahead and write about, One Flew Over The Cuckoo's Nest or something like that. Or like yourself, you, you immediately adapted this new tool and figured out new ways that you could go ahead and, and use it. I mean, that's, the, I think the brilliance of, human intellect here. And, and, and I think it calls to, the rapidity with which you picked this up, and think about how much more effective and productive it made you. Now imagine in the context of supply chain, you know, all the different, unforeseen applications of being able to have rapid human computer interaction at a level that is, has never been struck before. And the ways that some organizations are going to adapt to it and are gonna be able to take advantage of it faster than others, which will give them the ability, for example, to communicate, effectively with far more customers in far more scenarios than they have before to give them a step change, improvement in customer satisfaction, or even in the context of what riptide does. One of the reasons I was so excited to talk to you about it again and is, you know, when we, we have met, we just introduced the fact that Riptide is about bringing together everybody who needs to be together to resolve an issue in a conversation. Think of a delivery of a barbecue. You got the dispatcher, you got the driver, you got the customer, and they're trying to deal with the fact the customer's not gonna be home. And the dispatcher can be communicating with the customer, but the driver needs to know, the driver needs to communicate with customer as well. The dispatcher needs to know, well now we put everybody together in the same conversation so they can communicate. But imagine that done for the literally billions of, or hundreds of billions of service and and, and other transactions that are happening all the time. The, the scale of labor that would be required to maintain and manage all those kinds of communications would be staggering. Well, now along comes ChatGPT and we can insert ChatGPT the middle of those conversations and now organizations can be carrying them on all the time. Uh, you know, one of our biggest clients is one of the largest, automotive roadside service clubs in the United States. They do, you know, tens of millions, 30 million, service calls, every year. And, and now they could be using this to allow, a smaller number of dispatchers and drivers to be communicating constantly with their customers to go resolve issues. And, and, and this is again, is where we see applications happening, all end down the supply chain.

Tom Raftery:

So just run that by me, in a little more, in a little more detail doug. In your scenario, would ChatGPT fit into things?

Doug Marinaro:

Yeah, yeah. I, I actually wrote a little blog article and I gave I gave an example of, how this, might happen in the future. It's a ways off yet, but, you know, mom taps her smartwatch and, and says, you know, Hey, Molly and I are going to the park now, and it won't be home when that new barbecue arrives. Please set it up in the backyard and take all the packaging. And then the smartwatch responds you know, no problem. I, I alerted the store and they confirmed that the extra work could be included in your price. I also checked with the dispatcher and delivery team, and they confirmed as well. If they have any questions, I'll handle what I can and connect them directly with you, for anything else, right? imagine an an, an agent that could go through and do now we're, we're, we're quite a a ways from being able to do that and what we're applying first is we're, we're using these language models. let, lemme take a step back. The amazing things about ChatGPT three, three things that, that they change. Okay. First off is, is what I would call fluency. If you think about all the chat bots you've integrated, you've interacted with, just think about your, your Alexa in your home. You know, you start to have a conversation with Alexa. She can understand much of what you say, sort of mm-hmm., but it wrong. And she doesn't respond intelligently. As you've experienced, ChatGPT is whole different level. Spooky um, and, and spooky and, and G P T four is, uh, which has been experienced by some people in the context of the Microsoft Bing Is, is spookier yet. um mm-hmm. And, and I think there's an expectation that within the next. By the end of the year, we should be at the point where, pretty much won't be able to differentiate between, you won't be able to tell that the computer is talking to you. It'll be the point where you won't able to tell. So that fluency, that really changes the nature of how humans are interacting with this, you know, and what they expect out of this interaction that they're having with these, um, with this things. That's first the fluency. The second is, understanding. So again, I'll go to the Alexa example. You go through and it's, it's very easy to go through and, pose a question to Alexa that she just doesn't understand or comprehend in any way. And, what you find with ChatGPT, because it's been trained on the breadth of all human knowledge that ChatGPT pretty much understands everything you're asking it. Now what's interesting is, you know, ChatGPT remember is designed to predict the next word to say. It's not designed to be right. It's not designed to be accurate. It doesn't exist in the real world. One of the things that, you know, the experience with Bing has come out with is, has highlighted is the potential for these large language models to do what's called to hallucinate. Um, which is, the scientific word that, people have come up with to describe when it makes up stuff that it doesn't know. So, so you gotta be careful and, and it all comes down to how you construct the prompts and how constrained you can keep it. So if if the conversation goes on too long, it may just start to kind of make up stuff because it's trying to finish, come up with the next word. But, I think that, you know, those people who can understand the limitations and understand where these models can apply, will be able take advantage of the understanding that is embedded in here. And then I think it's not just the understanding that's embedded in the model at time it was trained, which is, you know, as a G B T 3.5 was 2021, but the understanding, of the other information that you feed to it at the beginning of the conversation. So like you did when you went through and you fed the transcript and you said, give me a summary of this Well, you, you gave it more information that it was able to go ahead and act upon. And I think so fluency number one, its ability to understand. And it uses a whole bunch of technology called embeddings and so forth to be able. Understand what people are saying. Um, and I'm not the one to go through and describe how that works and go read a lot of articles on that, on the web. The, the third area that I think is probably, um, most, amazing about it is it's, it's its trainability. And, um, anybody who has been doing, AI for a a while has understood that, the trainability of these systems can be really complex. Even trying to train up new skill on Alexa is, is a non-trivial task. To go, to get Alexa to go ahead and be able to do something different. Well that's changed because with chatGPT as you find, you don't have to train it to understand your transcript. You just give it the transcript and now it has that additional information. What we're doing is we're actually taking some other Lang large language models, and have had 6 million conversations that have been conducted in Riptide to go do issue resolution. And we're pumping those conversations into an existing large language model, um, that we're doing. We're doing this together with aws, uh, with Amazon, and, and we're gonna use that to, to create basically, a model that's trained on the specifics of dispatch, communication. Communication between a dispatcher and drivers and customers and so forth. And so it'll be expert on, on how people can respond on that, and we can do that. I mean, it involves a lot of computation, but it doesn't involve a huge amount of human effort, to be able to do that. Gotcha. So these three things, fluency, understanding, and trainability really are, game changers for, for, for this technology, that'll make it much more applicable in a wider range. We are are gonna use it first given those capabilities now, to understand number one, when, when should an agent engage? So imagine those a hundred billion conversations coming back at you. and, uh, you're an agent trying to figure out, well, you know, which conversation should I pay attention to first? Because I got all these people, messaging me. Let's say it's a surge in the morning of your del of your last mile deliveries that are supposed to occur and they're all on demand, and you got, you got drivers calling you and you got customers calling you. Well, if I could have something that could look at the the conversations and look at all the messages that are coming in from people and say, oh, well this person just said thank you, you can ignore that. Uh, and this person said, Hey, my gate code is this, uh, and this person said, I don't need my order anymore. And this person is really pissed off because of their sentiment. And you can analyze those conversations quickly. You now know where to pay attention, um, and that you now know when, when do I need to engage. And there's been a lot of work with sentiment analysis that's been done in the past. What we've found is can actually go in and say, how close is what conversation is so far to the statement. the customer is satisfied or the customer has new information ChatGPT will come back and immediately tell us, you know, rank order these conversations on the basis of that statement. You know, for example, so when to engage first. The, the next is, is, is, is how to engage, you know, or, or, or rather who to engage, who should be involved in involved in this conversation. Because remember, we're doing three-way. Oh, should I bring in the driver? Should I bring in the merchant? Should I bring in, uh, uh, the warehouse? shall, who else should be involved in this conversation based upon what's going on? And that goes to this understanding. And then the third is like to engage. And, and one of the easiest ones that we're gonna be implementing first is just language translation. All of these large language models were developed first to support really accurate language translation. If you think about language translation, you know, it's, you know, people can, can go through and say, oh, okay, well the, you know, the, the French word for I am is suis But, know, you're gonna say, I am going to the store, well, you're not gonna say, you know, je suis going to store. There's a whole a whole different translation for that. Right. You know, um, and my, my French is so rusty, I can't, I can't remember how to say that. Something about Jve, I think believe, whatever. Yeah, yeah. But, but it, it's predicated on understanding the context of the conversation and understanding that much more accurately. And these large language models were done, were designed to go ahead and do that. So one of the things we can use it for is, one of the big problems you have right now is if want to go involve a driver in the conversation, but the driver, you know, english is a second language for them. and then you're gonna be getting kind of, you know, broken communication. And what if the customer is speaking primarily in Mandarin? Well, we could support a three-way conversation with a customer speaking in Mandarin, the dispatchers in English, and the drivers in Spanish, and they're all talking about the same thing and communicating to go resolve the delivery issue.

Tom Raftery:

Wow, wow, amazing. Yeah, I, one of the other first tests I did with p ChatGPT t was I started speaking to it in Irish, and of course Irish is a very, niche language, not a lot of people speak it. So I asked it was it able to speak Irish? And it replied to me in Irish saying, yes, it could speak Irish, but with the proviso that it wasn't as good at Irish as it was at other languages, so it might make some mistakes. So I thought that was fascinating, but I was amazed, I could actually understand when I just addressed it in Irish and it, it re responded in Irish, in in perfect Irish. Uh, so that

Doug Marinaro:

I was at the, uh, I, I was at the manifest conference in Las Vegas, just two weeks ago and we're having lunch and I was showing this, uh, this woman who works for a, a, a large technology company, um, but was not aware ChatGPT I was just showing it to her on, on, on my phone. Uhhuh . And, uh, you talk about speaking different languages. Well ChatGPT not only speaks different human languages ChatGPT different computer programming languages. Oh, you can ask ChatGPT to create computer code. Um, yep. And I don't know if anybody's act on, on your, one of your listeners have actually, you know, worked with any of these large, uh, supply chain systems. I, I, I, one of the ones we work with right now is Salesforce. And, and, Salesforce is is an enormous system. There's just so much complexity to it. And if you're new to it, you know, you know where to start, know, um, but it's incredibly powerful for that reason. But, you know, you could go in and you could, you can ask ChatGPT. To go write, you know, a, a lightning component for you and you describe what want and it will actually write that component. And, and you can say, you know, what, what language you want it write it, and it'll write it in the right language. So I was her how to go through it and, and, create computer code. Uh, and I, I said, I said, you know what, what's your, what's your native language? And she says, oh, it's, know, it's, it's Mandarin. and I, I I, basically said, I you know, rewrite this in. we first wrote the, wrote the code in C and then I said, uh, rewrite it as if you're speaking to a Mandarin user. And not only were all the, not, not only was all the commentary done in mandarin, but all the, all the error codes were done in, in Mandarin, And still in C Sharp, it just boom. Just like that. Amazing. Amazing. It, it, it's, it is a a powerful tool.

Tom Raftery:

So just to prep for, today's episode, cause I, I, I do occasionally do some prep. I, I, I. Went on to ChatGPT and what are some potential use cases for chatbots such as yourself. In supply chain and chatGPT responded. There are several potential use cases for chatbots in supply chain management. Number one, customer service chatbots could be used to provide customers with real-time information on their orders, delivery status, and other inquiries. They can handle repetitive and basic customer queries. Freeing up customer service representatives to focus on more complex issues. Number two, inventory management, chatbots can, da, da, da da, da da. Number three, supply chain planning, blah, blah, blah, blah, blah, blah. Number four, supplier management. Number five, logistics. And then conclusion. Overall, chatbots can help to automate and streamline various processes in supply chain management, improving efficiency, accuracy, and customer satisfaction. So then I went on and I, I asked a follow up question because microsoft have said that they're they're going to allow organizations to upload their own internal data ChatGPT and to train bot on your own organization's data. So I said if a chat bot like yourself could be given access to the entirety of an organization's data, how then could it help that organization better manage its supply chain? And ChatGPT replied, if a chat bot like myself were given access to an organization's data, it could help the organization better manage its supply chain in several ways. And it gave me five ways. Predictive analytics was the first time, realtime monitoring, process automation, improved communication, and the last one was risk management, and then it gave a summary and so on. So, I mean, boom. I mean, what's really interesting about this, I, I, I didn't try it, but I've tried it with other as sim similar scenarios where it gives me a, a numbered list of things, you can then go back to ChatGPT and say, okay, give me more information on point number one. And you know, it'll come up with a much longer, uh, use case for number one or explanation of number one or whatever it is. And you can do that for each of the points. And then if it breaks those into lower points, you can go in and say, okay, give me more details on this. You can dig right down and keep going down into it. The amount of information that it can surface and put out there in, you know, a, as we said earlier in spookily human-like, uh, verbiage is just insane.

Doug Marinaro:

Uh, yeah, yeah, yeah., you know, I mean. We're not gonna need to have these podcasts anymore, Tom, because people can just go to to ChatGPT and know, know summarize what Doug and. and. Summarize what Doug and Tom might say about out of ChatGPT as Doug and Tom, you know, what do you you think? Um, yeah. I, I, I, I think what that goes to is, uh, you know, first of all, I think everybody should be interacting with it just to understand what it can do. Uh, and I think that those who understand what it do will be the ones who just like with any new tool, you know, there were, know, kids who were using computers and kids who weren't using computers, and those who had, using computers had an an advantage. Kids who, kids who could ride, kids who could ride a bicycle, could go faster than kids who, who only run. Right? Um, correct this, this is a new bicycle for mind Steve Jobs used to say. Right? And, uh, we, we need to understand it, to understand how we can use it so we can help us be more effective and better. Um,

Tom Raftery:

there's gonna be a a

Doug Marinaro:

it will,

Tom Raftery:

There's gonna be a new role or job come up, prompt engineer because I, I had this conversation, uh, very early on, uh, when talking to someone about ChatGPT and was a university professor, and, uh, she was saying her university is thinking of, you know, banning access to it. And I said, that's exactly the wrong thing to do because I said, yes, your, your your students, when they go out in the real world, are going to be hitting a real world, which using ChatGPT all the time. And if you have denied your students access to this tool, they won't be able to use it in their work. What you should be doing is having like an open book, approach to it, where in exams all the students have access to ChatGPT during the exam. And then the ones who can get the, the, the most creative responses, or the best responses are the ones who get the highest marks. Hence the, the, the term prompt engineer. Because the prompt what you give ChatGPT to get out the answer. So the people who can best use this going forward or, or similar tools, are going to be the ones who are going to succeed. You know, to your point about organizations, it's down to individuals as well. The people who can use tools like this, the best will be the ones who can get the, the most advantage from it.

Doug Marinaro:

Absolutely, absolutely. I, as a matter of fact, I think some of the things we're looking at is, you know, how can you actually first off, in, in response to your point, I agree 100% and I I think there's, gonna be an initial reaction of let's, let's, let's ban this, let's prevent people from using this in particular situations. and that's, that's a lose lose. it's gonna happen. It's gonna be used. It's already being used. The Kids are, the kids have already figured it it out,

Tom Raftery:

Yeah.

Doug Marinaro:

and, and they're, and they're, they're using it. And the idea of, of understanding the, the prompts to get you the right information is, is really gonna be the new, the new art form. It's, you know, I mean, look, I, when I first started coding, I was coding in assembly language and, and then then there was this whole new thing called higher level languages. And I was working in basic and, and in Pascal and, and, you know, in Fortran and so forth. And then, and now the, and then language has got higher. And you're write how to write in web, well now, now people are gonna figuring out how to write ChatGPT as you say they're gonna be figuring out what are, the right prompts to be able to get the information that I need right now to me, help me be more effective? And, um, couple points on that, if you think about it, what does that mean for, for websites? Mm-hmm., um, you know, and this is, this has been the point about, you know, the use of ChatGPT in the context of a search engine. You know, w if you think about the classic search engine, you go and you type in something through Google and you get a list of all these websites you gotta go visit to go find out the information you're really looking. Yeah. Well, what you could just ask for the information you're looking for, it just told you. I think our, our, our children will at some of our grandchildren will be sitting there going, I, I understand you guys to have to visit websites and read things to figure out stuff. Why do you just ask the computer? Right. Um, and I I think that's gonna be the for a whole variety of kinds of of human computer interaction that are gonna occur Yeah. There. there is an interesting, um, um, next level of this is where you have chat ChatGPTs interacting with other ChatGPTs. Different chat agents that are working. And one of the things that we're looking at is, uh, you know, especially in the context of, because the way ChatGPT is working, it's trying to predict the next word in, in the conversation, what should it be saying next? Um, all based upon the input it's already received and it's, let's say just responding to you, Tom, as a customer. Uh, but now how does it get new information? How does that get inserted into the conversation in a way that doesn't interrupt the conversation I'm having with Tom? Like, you know, oh, it's a new ETA update, you know, how does that get inserted there? Right? And so one of the things we're looking at is, well, what if, what if you could kind of have like a multi-level conversation happening on where the, there's a ChatGPT that's getting all this new information from different sources. and and then it's, also now predicting, okay, what should now be said to the customer. And then that's actually what gets pulled out and then presented to the customer at another level for that conversation. So I I think there's gonna be, a lot of lot of different ways in which people are gonna begin to use this. One of my colleagues, uh, in, in the venture industry said, you know, when, when, when Steve Jobs walked up on stage and introduced the iPhone, it blew us all away. But if you had asked me at that time, what impact would that have on the taxi industry? I, I don't know, , you know, and then, and then, and then a year later, Uber appears, right. You You Yeah. and, and, and the taxi industry's gone. And, and I think there, there's gonna be like, uh, and you and I talked about the other day. You talked about how, how you felt like the trend for ChatGPT was going in the wrong direction because it was no longer able to accept as much input as you wanted to give it up front. And it would seem to be operated a little slower. Well that's cuz everyone was piling on. Right. Exactly. And they were just throttling its usage cuz they couldn't, they couldn't keep up. But, you know, this technology is only gonna get better and better and better. so the, the underlying you know, computations and and engineering that are occurring, we're just at the the start of this process and it's, it's gonna be, it's gonna be this exponentially improving, you know, glide slope that we're all gonna be on for this. The second thing that's gonna be happening is gonna be all those smart prompt, engineers, you know, using their prompt engineering expertise together with their human creativity to come up with a whole host of of new applications and new ways in which we can take this new tool, which is, you know, underpins all of human society, our ability to communicate with each other yeah. And, and begin to apply. Yeah. Now, so of course you, you can't, you know, you can't, you can't say that without actually having to touch upon, you know, the scary parts of this know, with any new powerful technology, from, from the atom to CRISPR to to the discovery of of fire. Right. You know, it's, it's available for misuse and, and I think we've, we've, experienced certainly in the realm of, you know, of facebook based misinformation and, everything QAnon you know, whatever, you know, the ability for people to go use communication and to use language to cause other people to un to, do bad things, right? Um, now, now we've just placed this amazing tool in, in everybody's hands, or that could be used, you know, maliciously as well as positively. Yeah. And so that's, that's gonna be an interesting thing as well.

Tom Raftery:

Yeah. It's like any of these things, I, I often say, you know, a knife is great. You can use it for spreading butter on your, on your bread or you can use it for attacking somebody. It's how you use the tool. But you, you, you raised a number of points. You talked about search engines and, you know, I've had access to New Bing now for a number of weeks, and I gotta say I am using it all the time. I haven't gone to Google in a long time now since I started, since I got it. And it's, it's phenomenal. It really, really is. Uh, so, you know, if you can get access to it, I, I would recommend that is significantly better. Now. There have been issues with Bing in the last, particularly in the last week, and Microsoft have scaled back some of its abilities because, people, some of the prompt engineers were too good and getting it to say things it shouldn't be saying. So that, that's been a lot of fun to watch. It now will only give you like five follow up answers to a question, and after that it says, okay, time to start a new chat. And, uh, you have, you have to refresh and start over. But apart from that, uh, it's, it's really, really good. Another point you made about ChatGPT itself degrading over time, uh, because of its use. Uh, so there, there've been several releases of ChatGPT. That was the first one on the 30th of November. The, then there was an update to it, on the 9th of January and another on the 30th of January. And with each of the updates, it's functionality degraded. And that was my guess on was it degraded because to your point, too many people were using it. They had assigned too many resources to each user, and so they scaled that back, and so it became slightly less useful. So it's not as useful now as it was in late November, early December. Having said that, they have as well now a ChatGPT Plus option. So ChatGPT is free, and that's the, the basic one, which has had its utility curtailed slightly. The The ChatGPT Plus is $20 per month, 20 euros a month here in, in, in Europe. It's amazing. I went and tried it out and i, I said, you can cancel any time. So I said, for 20 euros, I'll give it a shot. If I don't like it, I just won't renew. So tried it and went, oh yeah. Oh, oh boy, I'm keeping this. Oh, absolutely. ChatGPT Plus is significantly better. If anyone tries ChatGPT and thinks, yeah, this is good, great. If you try, you ah, this is a bit limiting, maybe spend the 20 euros, try out the Plus version and see if that's better. Because for me it was orders of magnitude better. And the other thing I would recommend to people to try out as well is Excel, Microsoft Excel. If you are a Microsoft Excel user, you can go ChatGPT and say, create a formula for doing boom, boom, boom, boom, boom, whatever it is, you know? I want to do a pivot table for these kind of things. How would I do that?? Boom. Up it comes. It knows Excel upside down inside out. So it's not just the programming languages you mentioned the, like Fortrans, C Sharps and all this kinda stuff. It can do, it can do that upside down inside out, no problem. But it can also do the more basic stuff. If you are someone who gets into Excel and you know you're a heavy programmer of Excel, fantastic. Even better, you can start making use of this as well for that and PowerPoint and, you know, any, any kind of thing like that. It's, it's just it.. mean, try everything on it if you're gonna try it, because it it, it will make your day so much better. Even email. I've seen people who, you know, were snowed under with emails and they would just go to ChatGPT and go, I've received an email which says, boom, I wanna answer with something professional sounding, which makes these points, uh, what would a response to that look like? And boom, outcomes, a lovely formatted email, copy, paste, send, boom, gone,, you know, just matter of minutes. And you, you get really good responses to your email out. So it's just, yeah, the use cases are just, uh, they're, they're only limited by by your imagination to, to use that expression,

Doug Marinaro:

Uh, it's, I'm, I'm looking forward to to having this conversation with you again in a year. as, as as we, as we look back on, on all the things that have, come about as a result of this. I, I, I don't think people understand just how impactful gonna be. I think there's a, I, I think you'll find a lot of commentary about what it can't do, a lot of of commentary about odd things it can do. Uh, I, I concur with you, Tom. I, I would just suggest to everybody, dive in, come to your own conclusions. The, the, the, the 99% of amazing awesomeness that it has, um, at this point will be what you'll experience. Uh, and the 1% of the weirdness is about whatever all the news media is gonna be writing about. Um, and, uh, and, and I'm, I'm, I'm eager to see, I'm eager to see what, um, what, what comes out, out of, uh, usage in the supply chain. I, I think it's gonna be, um, like I said at the beginning, I think it's gonna be pretty transformative. I think the supply chain is fundamentally, uh, a human experience, uh, even though we're talking about goods and services and it's gonna be humans that are interacting all the way up and down throughout that. And, this changes human computer interaction just fundamentally.

Tom Raftery:

Fundamentally. Yeah. Agreed. Okay. Doug, we're coming towards the end of the podcast. We've gone way over, in fact, but it's just, it's been such an interesting conversation. Anything that I haven't asked that you wish I had or any aspect of this that you, we haven't touched on, you think it's important for people to think about?

Doug Marinaro:

Uh, no, I, I think we've, we, we, we've covered all I'd. I'd love to be able to come up with pithy at this point. I, I probably could ask ChatGPT to summarize this conversation for us, and I'm sure you'll , uh, you may, you can maybe, you can maybe ask actually, uh, you, you should, uh, here, here's a fun exercise I will suggest, uh, when, you're, when you're done with this, uh, transcript. uh, take that same question and pose back ChatGPT and, and, ask ChatGPT to go through and, uh, and add anything it thinks it ought to add to conversation, making it the third party. How about

Tom Raftery:

Nice. Nice, nice, nice. Yes. I'll try that and see what Happens. Interesting. Good, Good, good, Okay, Doug, that's been great. If people would like to know more about yourself or any of the things we talked about today, where would you have me direct them?

Doug Marinaro:

They can go to our website, uh, riptide hq.com. Uh, check out us on LinkedIn as well. We're, we're constantly posting not only on ChatGPT and what it's doing, and what that might mean, but also on three-way conversations and just conversations in the supply chain in general.

Tom Raftery:

Doug, that's been great. Thanks a million for coming on the podcast today.

Doug Marinaro:

You bet, Tom. Have a good day.

Podcasts we love