251 GPT-4, Microsoft Copilot, Google Bard, OpenAI CEO Sam Altman’s Concerns about AI’s Disruptive Impact and the Latest in Organizational Applications of AI, Tom Taulli and Mahan Tavakoli | Partnering Leadership AI Conversation

***DISCLAIMER: Please note that the following AI-generated transcript may not be 100% accurate and could contain misspellings or errors.***

[00:00:00] Mahan Tavakoli: Tom Tulley, welcome back to Partnering Leadership. I am thrilled to have you with me. Thanks very much. Glad to be here. Tom, you are the first person that is a repeat guest, in 250 episodes in part because. , I truly see this as a transformative moment for humanity, including for our organizations, and you do an outstanding job in both explaining artificial intelligence and its impact for businesses as well as keeping up to date.

So that's why I am thrilled as we get to have these conversations because the field is moving so fast.

[00:06:22] Tom Taulli: Yeah, I've been in the technology business for a lot of my life and I've never seen things move as fast before, I can't imagine if someone is not into this business, I don't know how they could keep track of everything.

I can barely keep track of what's going on, .

[00:06:35] Mahan Tavakoli: I feel the same way, Tom, I feel overwhelmed every morning. Yeah. Yeah. When I see the tools that use artificial intelligence and some of 'em are magnificent, and I'm like, oh my God, you need full-time just to focus on all the different tools that are coming out every day.

[00:06:51] Tom Taulli: Yeah. We need more hours in the day. That's what we need. And technology hasn't done that yet.

[00:06:56] Mahan Tavakoli: Tom, when we talked last, we were talking about how incredible G P T 3.5 was and its potential open AI introduced G P T four. When I saw the video, it was mind boggling. I'm sure a lot of the audience members have seen the video.

 From the image of a napkin G P T four produced a website, so there was no need for coding. So would love to get your take and understanding of what is different with G P T four as what some were exposed to in G P T three point. ?

[00:07:34] Tom Taulli: The original G p T three was basically a large language model, just language only.

So you would type in a question or some type of prompt and it'll generate content. This is multi-modal and that's a fancy word for saying it's more than just writing. You can also use images. Now there was talk of video. , we haven't seen the video part, but at some point probably get video.

So like you said, you can take an. And go to text or create a website or use text and create an image or a video and so it's more powerful, more generalized system for generative ai. But they've also added other things to it as well that, in terms of things that you may not see,

improve the accuracy of the models and the power for reasoning as well. So there was a lot to this new version. And we'll see how it plays out, but so far it's pretty impressive. .

[00:08:24] Mahan Tavakoli: So what do you see as potential uses that now this G P T four will have

[00:08:31] Tom Taulli: Yeah. There's been so much happening. It can get overwhelming with all the new developments and so forth. So businesses will start using it, get creative with it. But it'll probably be something similar to the way Excel is.

I remember years ago, Microsoft did a survey with its customers and they asked them, what features do you want in office and Excel and Word and a lot of the features people. That they wanted already existed in the software . Probably what's going to happen is people will use 5% of what's there and the capabilities, unless the company is gonna go and do some pretty intensive training, which would probably be a good idea what's gonna happen is business.

We'll hit on a few features and functions or those that they've seen or they talk to other people. Microsoft Co-pilot I do think shows some insight into how businesses would use this. It could be used to summarize emails it could be used to generate emails based on what you had to prioritize.

A lot of it seems to be about emails , how much of our day do we spend. reading, responding to emails, keeping track of emails if companies could find a way to, streamline that there's some big improvements for a business.

So I do think, something as mundane is emails, could be the killer applic. For the G P T four there could be other, like you said, you could create a website with someone writing it on a napkin. That's cool. I don't know how often people are gonna do that.

But every day we write emails or we create documents. The other thing too is Presentations. We do so many presentations , in our daily work, and it's not intuitive. To create animations, a lot of people just don't know how to create animations or, to create some compelling images and things like that into the presentation.

So I just think actually some of the simpler functions will be Make a big difference. And you'll always have the power users out there, but for most people it'll be fairly simple, but it'll will be transformative to their work.

[00:10:20] Mahan Tavakoli: So you mentioned the Microsoft Co-pilot and Google is also incorporating AI technology into their office tools.

You mentioned email, you mentioned presentations. How else do you see artificial intelligence impact work and collaboration using these tools? ?

[00:10:38] Tom Taulli: , part of what it where organizations are not as agile is just the amounts.

Busy work that needs to be done. Things that are repetitive, things that are tedious and those are the things computers are just really good at. So Microsoft has something called Power Automate. Now there's something called robotic process automation, which is what you can automate certain processes in an organization.

It could be as simple as cut and paste. We cut and paste this into our CRM and into our ERP. There's all these types of things that just really tedious activities, but suck up so much time. But creating a bot to automate that is actually of complicated and requires some programming, but with generative AI and the chat GPT systems it could be more of just you type what you want the system to do, or create a bot that cut and paste this type of information over to this and voila, it does it. You don't have to do any coding you don't need to know JavaScript or anything. And then it does it. You can test it. If it works, great and then, I wanna do it this time every day.

Okay. Boom. And so I do think automation or these next generation RPA tools will go a long way in helping that, and it will be so much easier because the interface is just natural language. , so you don't have to learn really anything. If you could speak English or speak whatever language you speak, cuz it'll understand any language, then it'll create these automations in your system and improve the business.

Because the other thing is, you could take software out of the box, but those types of solutions only go so far. Every organization is different , they have different processes and policies and so forth, so you need some customization.

So that's where, creating these bots are important. And generative AI is gonna make that a lot easier. And we're gonna have a lot more automation in our workplace, and that leaves people more time to spend on things that are more important than these tedious and mundane activities that do suck up a lot of time.

[00:12:31] Mahan Tavakoli: When we want to travel at this point still, it's helpful to learn the language of the country we are traveling to. And in the past when I wanted to use PowerPoint or Excel, or the c r m in the company. I needed to learn the language of PowerPoint, the language of Excel, telling Excel what I wanted to do, but in Excel's language, not in my language.

Right now. It sounds like with generative ai, we can have conversations with these tools and get them to do what we want them to do in our language, not in their language.

[00:13:13] Tom Taulli: That's. , that's powerful. Cuz today there are these power users that are great at creating macros and scripts but there's only so many of them out there.

And they just may not have time to do what you need to be done. So what happens is people just rely on, these processes whereas they could create their own automations and save a lot of time for themselves without having to rely on someone, they don't have to go and learn this, or they don't have to rely on someone within the organization who usually doesn't have the time to create a bot for that.

[00:13:42] Mahan Tavakoli: That's interesting. So in certain respects, the way I see this is that it will benefit people that have the creativity of thinking and not hamper them if they don't understand the interactions with technology. For. Couple of decades, people who have been able to best understand the technology have had a huge advantage.

 Because they're the ones that have been able to leverage the technology. Now, it seems like the field has been equalized. Therefore, the advantage goes to the people who can think more creatively and ask more creative questions of whatever the technology tool.

[00:14:24] Tom Taulli: Yeah, it's a democratization of technology which is a huge development very important.

And it's really important for companies. Again people need to know that these functions exist .

Once you do it it's pretty natural. So companies, will need to do, it's a little bit more training co-pilot, Microsoft will have its own videos and so forth, and how to use this type of system. But you need to reinforce that within the company because old habits die hard. The other lesson too is change management is critical. Technology falls apart and when you don't have effective change management.

[00:14:55] Mahan Tavakoli: There's a lot of anxiety, with many of the CEOs and executive teams that I've conversations with, , lack of clarity with what's going to happen now. One of the things that . Would love for you to share some of your thoughts about is the significance of plugins. . So it sounds based on what some people have been able to find out openAI hasn't talked much about it yet and hasn't released it, is that they're already. 80 plus plug-ins that have been developed to work off of G P T four and many more will be what are plug-ins and how can they help us and organizations tap into the power of G P T four? .

[00:15:40] Tom Taulli: . So the plugins are integrations so it's an existing piece of software.

So if I have a CRM or whatever type of application, a lot of those applications have integrations with other systems so they can communicate, with those systems and work smoothly with those systems. It makes it easier for me if there's an existing plugin so if I use a certain type of software, I'm familiar with that certain type of software, I could use that plugin and then start getting some of the benefits of this type of generative ai.

 It is really beneficial to the software vendors because it makes it easy for them to extend. That capability to make it available to their customers.

And what will happen is that, get ready because we're just gonna see a lot of natural language, interfaces starting to emerge. So it's a way for open AI to make it as easy as possible to spread this technology and to educate the market about how to use this type of approach of using prompts

to get things that you want, whether it's to create a blog, whether it's to create a automation or whatever it is you want to do, it can all be done with just writing natural language. But for an application to support that, using this plugin can make it possible. So I

[00:16:51] Mahan Tavakoli: was reading Tom about plugins, specifically enabling.

Organizations to have their people interact with their own internal data. Built off of the conversational format and power of chat G P T, mm-hmm. , which in my mind be transformative. One of the things that has been really hard for companies for as long as I've been involved in business is knowledge management.

, there are different, bits and pieces of information in different parts of the organization or different parts of the IT infrastructure. , How do you visualize that? Working plugins. Enabling teams and organizations to have a conversation with the information from inside their company.

[00:17:38] Tom Taulli: . Yeah, you raised a good point. So I talked to Morgan Stanley. They're one of the, Companies that have been working with OpenAI and G P T four for some of the applications that they're developing and what they've developed or it's still in the process of being developed right now is making it for their financial advisors to get access to information.

Generally it's all there again, like you said, people just don't know how to use it or get value out of it. So what they've done is taken all that information. It's not client information, it's research report, it's reports analysis it's a lot of those documents. And then they leveraged the open AI models to make it easier.

To work with.

So , knowledge management is not just about finding a piece of information. There are search systems that can do that. Okay. The next level of that is that you have this information and how can it help answer my questions? How can it help solve my problems? So that's where the power of this model is.

It's not just about querying the database and getting some info from it. It's getting insight out of that based on complicated scenario.

[00:18:48] Mahan Tavakoli: That's very interesting this sticks to internal organizational information. Making sure that the data that is provided and the answers that are provided are.

[00:19:00] Tom Taulli: Not to say it's always perfect, but at least it's your information and it's gone through your own process.

You mentioned the open AI presentation for the demo that they had for GPT 4. The president. Rockman. He did one situation with taxes, and with the prompt, he put in a document from the IRS.

IRS regulations. Most people don't know how to read a IRS regulation, much less understand how it applies and even tax preparers may not even know how this works. But he put it in there, put in this long document of complicated information and asked it some, tough questions and got the correct answers out of it. So that's the difference here. It's getting value out of that information and interpretation out of that information that is usually generally complicated, so just imagine doing that for medical documents or research, just think of all the medical research that's out there.

There's just so much of it. Even experts in the field have a hard time keeping track of it, but generative AI could do that. .

[00:19:59] Mahan Tavakoli: So it's interesting, Tom, with specialization, we have added complexity to both the words that are used and the way concepts are communicated. So whatever field you think about, as you mentioned, whether in the medical field or technology, aerospace people that have spent decades in the field are only the ones that understand it. Read the paper and understand it this way. It sounds like pretty much any information. Can be translated in a way that the normal person can understand the intent or what to do with it, rather than just someone that has spent decades studying in that field.

[00:20:43] Tom Taulli: Again, it's about democratization.

[00:20:45] Mahan Tavakoli: I had a conversation with Louie Rosenberg and he mentioned augmented reality and how our future will be augmented.

. And just today I was reading about three students from Stanford University, which have combined G P T. With Whisper and augmented reality glasses to create what they call Riz, G P T, which listens in real time to conversations and then guides the person who's wearing them on what to say next to be more charismatic.

Oh, wow. Which is, wow. So it's it's marrying that augmentation. The conversational capabilities that exist in at G P T four, I've founded Mindboggling. .

[00:21:30] Tom Taulli: Yeah. There was an example in the pre G P T four era, which is, not that long ago there's a company called Do Not Pay.

 It's online service that tries to, get rid of your subscriptions, you don't want or handle. A traffic ticket or something like that. So what they had was for a traffic ticket in traffic court. And they were going to have someone use G p T three system to guide that person in court to deal with the judge and to basically, Say, this is what you should say.

 This is how you should, respond to this question and so forth. And it caused so much controversy that they had to pull back cuz the local bar, wanted to sue them or put 'em in prison or something like that. It's going to come to the point where, yeah, this will be augmented intelligence that, you hook it up

to some type of glasses and the earphone and it could make you sound a lot smarter than you really are. Yeah.

[00:22:21] Mahan Tavakoli: It does make it interesting because at the same time I've seen some tools that allow people to automate conversations. For example, they have in dating apps . , it adjusted a profile of the person that you're having a conversation with. And in essence, the chat bot.

Determines based on this person's personality, what would appeal to them most. And has a conversation with them. So it gets into gray areas with respect to , when is it us, and yeah. When are we augmented and when are we being replaced to allow AI to make the decision making for.

[00:23:03] Tom Taulli: Yeah, it's kinda like that movie her we're pretty much there. , figuring out what's human and what's not will be a big question going forward

[00:23:10] Mahan Tavakoli: to that end, Tom Sam Altman had a great, very long conversation with Lex Friedman. And he is positive about the. Of ai. He mentioned that he believes we're closer to artificial general intelligence than many assume

but one of the things that he mentioned is that we as a society, are not having the conversations we need to have that this thing, even at the G P T four level, will be truly transformative and impact significantly. Most of our jobs, I believe 80% plus of our jobs will be impacted and almost half of the jobs.

Will be significantly impacted as a result, and we haven't had those conversations. Would love to get your thoughts. Yogi Barra said making predictions is hard, especially about the future. Yeah, that's right. But would love to know your thoughts at this point with respect to the disruption that this level of

generative AI is going to cause for our organizations and in the community.

[00:24:20] Tom Taulli: Yeah. Yogi, I don't know what it was, but he had something for everything. And whatever he said could be applied to generative ai. What happens is it's the Crying Wolf syndrome about lost jobs that this fear has been around since the fifties.

IBM actually avoided getting into AI research because they feared that they would be associated with killing jobs. And that was the case for a lot of other software vendors or computer vendors at the time. There's always this talk about lost jobs and it just never happens.

But past is not a prediction of the future is, your financial advisor would tell you or should tell you. And because it hasn't happened we're clearly not prepared for it. It's like one of those things.

Yeah. It'll happen. It probably won't happen to me, is how it usually is. But it could happen to you. And what's different is this is about knowledge work . The automation in the past has been more about physical work. Robots on a assembly lines, one of the first major examples of this in the auto industry, and it did have devastating impacts globalization and all these types of trends with automation have had huge impacts, mostly on physical labor.

This time it's knowledge labor. So when G P T four came out, and if. Thinking about the applications about jobs. The one area I would've looked at in that report that OpenAI published about the G P T four was all the exams they aced. The bar exam was top 10%, biology exam for ap, which, these are not easy exams.

And by the way, the first computer software company I started was an exam preparation. So oh, , I, never thought this would happen, but it's happening. Now, passing exam does not make you a lawyer or a doctor but it does tell you that the reasoning capabilities and the knowledge capabilities of these systems are advancing at a very extreme pace.

And if you're an employer, and this has always been the case, again, with physical labor, if a robot costs me $50,000, but an employee costs me $70,000, but it's actually a hundred thousand, when you add benefits, I'm gonna go with the robot. , if chat g p t cost me 20,000 and knowledge worker cost me 70,000, we know what's gonna happen.

And there's a lot of jobs that are about knowledge that to me can be automated. And it will take, some fine tuning, but it'd probably be worth automat. So while long-term, maybe it creates more jobs than not, but the short-term disruption could be severe because people have a tough time transitioning.

We've always seen this re-skilling and transitioning to new careers or, new ways of doing things has always been very challenging, and in least in the United States, we're really not good at having programs to do that. So if this technology advances the way it is.

, the impact in society can be very substantial and I don't think in necessarily positive way for a lot of people. And what that could create, is resentment against technology and those who are selling this.

Sam Aman says, it's beneficial. Yeah, if I owned 10% of open AI , I would find it very beneficial but if I'm that customer support person, And I have a family to feed. not sure if I'm finding that beneficial, so I don't wanna be overly dystopian. These types of predictions have always been wrong in the past when it's come to AI and knowledge work, but I do feel this time we're in different territory and I do think we need to take this much more seriously than we.

[00:27:34] Mahan Tavakoli: I really appreciate the point that you make, Tom.

I'm optimistic with respect to some of the potential of the technology. However, just saying that because past technological. Revolutions have not resulted in job losses. Therefore, this won't, it's not the right approach to it. We have to have the conversation around what it will impact, who it will impact, and try to address that as a society, not if 50 years from now we look back, we see that new jobs were created because, Possible for a lot of people to be left behind. And I also agree with you that this is different up to this point. We've had transitions that have taken human labor and force.

This is impacting taking out human intellect. . And I don't think there is much beyond that, that I know . That's right. Human force. I know. Human intellect. Yeah. Maybe we'll figure out something that's above that. . Yeah. But. That can be really significant for us as a society. Before we started recording, I mentioned to you that I've been involved and very familiar with organizations and groups that are focused on retraining and training people who are underserved to access jobs.

And one of the things that has been funded, and they have pushed for the most, has been. Basic coding skills, and the one thing that Sam Altman said in that conversation is second to customer service roles that he thought would most likely get eliminated in vast majority of organizations. He said a lot of coding jobs will disappear except for the people at the very top of the field.

[00:29:18] Tom Taulli: Yeah, part of this is that , it's the nature of coding because a lot of coding is cut and paste. The little dirty secret about coding is cut and paste. They go to Stack Overflow and, they have these repositories and, all these frameworks and stuff, and it's kinda piecing together Legos.

Obviously you still need understanding, how to use the coding, but that's something that can be automated and is being. . For the very, high powered coding, that's probably not yet in the realm of computers. It probably will get there, but how many of those types of coders do you need?

 That's a big question mark. I don't know the answer to that question. It could actually be that you need more of those because, c. We'll have solved this problem, but how do you find those people? Co coding is not the easiest thing in the world, , and you have people that have these fancy computer science degrees and they're not necessarily the best coders.

Back in the eighties, bill Gates, had this 10 x principle, it was. The difference between a good programmer and a really good programmer is off the charts. And if he found one of those, he would do whatever he could to hire him. He'd give them anything they wanted.

It's really fascinating category of employment. But it's probably just like anything, the difference between a, baseball player versus someone in the minor leagues is usually pretty significant versus just a, regular amateur.

But I've looked at these coding tools. They're absolutely amazing and they're getting better and better. Seems like every month or every two months they get more capabilities. And it shouldn't be surprising either because we've had this huge shortage of qualified, programmers.

So what's gonna happen? Companies are gonna look for ways to automate and find alternatives. And that's what's happening.

[00:30:51] Mahan Tavakoli: So that's one of the areas that will be impacted. But would love to get your thoughts, Tom, as we mentioned, even in the couple of months that have passed since we had the conversation, it's our initial conversation is almost like a lifetime has passed in ai.

There isn't a day when I don't see news. Related to the applications, that just boggles my mind. The exponential rate of growth and its impact. . So how would you recommend for professionals for leaders of teams and leaders of organizations? To try to keep up enough with this so they don't become irrelevant.

Yeah. Have their teams not be able to use the technology and the potential. Or their organizations. And their business models to become irrelevant.

[00:31:45] Tom Taulli: You still need. Foundation to work off of, you don't have to understand how a deep learning model works, at its intricate details, but you should have a basic idea of how they operate. What is machine learning, some basic statistics and probability.

 It reminds me when I was in high school I got to. Bill Gates several times, and I asked him that question, what should I do? This is back in the early eighties. What should I do? What should I learn? Basically, this is what he told me. It really hasn't changed since then, basic math and statistics and, we should all know that stuff because it's important.

So you have a good basis to work off of and it's easier to evaluate what's happening in the real world. Just even Google News and having a filter for generative ai going there every morning and if a headline sounds interesting, maybe read that article.

So read a few articles a day. That'll probably keep you in. As to what's happening. But I do think having some basic re-skilling of the workforce for some of those basic subjects could be really helpful. Because you still need to interpret the information, evaluate it Apply it, understand where there might be problems with this information.

So that's where the basics can go a long way in helping a company or organization, get through this. The other thing is not to create your own. G P T for, project. It's actually a pretty involved process. No problem experimenting with it a little and checking it out and, playing around with it.

But, if you wanna do a G P T four project, you need data scientists, you need some serious resources to do that. So if you wanna start using this technology in your workplace, go check out copilot.

If you have the office system, look at some of the features. If you have power automate or thinking about it, check it out and see how that can work for you. So I would rely at this point on vendor. Tools and not do the custom projects. Probably more of a waste of resources and time and money.

[00:33:40] Mahan Tavakoli: And there are lots of great tools that are being built off of what's already out there with their g PT four with the plugins that are being added. So I wholeheartedly agree with you learn the basics behind it, and with.

experiments, but experiment not building your own. Experiment using the many tools that are out there. Some of them will thrive, some of 'em won't. Again, literally every day, Tom, I'm seeing dozens if not hundreds of different AI tools. Many of them. Doing identical things, .

[00:34:17] Tom Taulli: That's right, that's true.

[00:34:19] Mahan Tavakoli: Some will survive, many won't. You don't need to recreate that wheel within your organization. You just need to know how to use it for your organization. .

[00:34:30] Tom Taulli: Yeah. There's a lot of hype. So a lot of companies are suddenly now generative AI companies for chat, G p T companies, when they're really not.

And just a few months ago Sequoia had a chart of generative AI companies and you could fit it on one page. Now, you need pages, , and that's another problem evaluating the product. Now a lot of 'em do have trials for each trials.

That's one way of, testing it out. But it's, it is getting to the point where it's hard to figure out what's real and what's

[00:34:58] Mahan Tavakoli: I appreciate you helping the audience. Mm-hmm. , Find out what's real, what's not , you write on tom Share your insights on LinkedIn, on Twitter. I will link to those in the show notes as well, and really appreciate it because it's mind boggling how fast this pace of change is. The other recommendation I have for the audience is you don't need to read everything to try to understand it.

Find a few good, credible sources, and you, Tom Tali are a good credible source. There are lots of people running around. Saying, this is how you can make money off of Chad g p t become a millionaire. They could become millionaires. They would've become millionaires themselves. They wouldn't been promoting it

So find a few good people that understand the technology and then can explain what the potential of it is, which is why I am thrilled to continue having these conversations with you. Tom, learn from you and share some of your insights with the Partnering Leadership community. Thank you so much, Tom Tali.

[00:36:04] Tom Taulli: Thank you very much.