Future Focus Part 2: A Look Beyond with Peter Mulford & Peter Coffee

Join us for an exclusive webinar as renowned futurists Peter Mulford and Peter Coffee engage in a candid dialogue to explore what lies ahead. In this session, Peter Mulford, Global Partner + Chief Innovation Officer at BTS, will interview Peter Coffee, VP for Strategic Research at Salesforce, for a one-of-a-kind opportunity to gain direct insights from two of technology’s leading visionaries.

Together, they will dissect the trends and disruptions shaping our world, looking beyond the immediate noise to identify the critical shifts that will define 2025 and beyond. Join this session for a glimpse into what strategic actions you should be taking now to prepare for the future.

ABOUT PETER COFFEE //

Peter Coffee is the VP for Strategic Research at Salesforce, and is now in his 18th year with the company. Additionally, he is the co-founder and President of the Foundation for Intelligent Life on Earth, sponsoring work on climate change, conservation, STEAM education, and the exploration of earth and space.

ABOUT PETER MULFORD //

Peter Mulford is a Global Partner and the Chief Innovation Officer at BTS, where he leads the Innovation & Digital Transformation practice and Global Insights and Design Laboratory. Working with Fortune 500 firms, he focuses on developing innovation leadership, design thinking, and disciplined experimentation capability to foster a culture of innovation. Peter’s writing and research have been featured in CLO and Training magazines, and he is a sought-after keynote speaker, often delivering addresses at HR conferences, Technology and Innovation events, and company off-sites.

Transcript:

Mike Anastasiades: Welcome to the Future Focus: a look beyond with Peter and Peter. 

Peter Coffey and Peter Mulford. This is part 2. By the way, hopefully, you got to catch part one. I am Mike Anastasiades, Solutions Director with HIKE2. We provide advisory services and solutions for companies across industries, industries such as public sector institutions. And you know, professional services companies, high tech companies, law firms we are a consulting firm with with a focus in in several different areas, cloud solutions certainly Salesforce, Snowflake. You know, other other similar cloud solutions, data and analytics, innovation, human centered design advisory services. We, we help firms, we help organizations that are looking to digitally transform really make the most of what they have today from a sort of from a current technical investment perspective and from a future technical investment perspective. Of course, we’re doing this to help help our clients solve some of their biggest business challenges, and I am thrilled to be joined by our guests.

Peter and Peter. I’ll start with Peter. Mulford quickly got Peter Mulford. global partner and and chief innovation officer at Bts. He, he leads the the innovation and digital transformation practice at Bts, a management consultancy focused on strategy execution. He helps leaders and and teams from Fortune, companies from around the world succeed with digital transformation. Last time we chatted. I don’t know if he remembers this. I’m sure he does. Peter had been facilitating an offsite meeting at the Vatican with Pope Francis. Peter, have you had any interactions recently with the Pope? Let’s that’s my st question to you.

Peter Mulford: Yes, no, no, no, I haven’t, not since then. But but thanks for asking.

Mike Anastasiades: We can leave that, maybe towards the end. Maybe I don’t know. Maybe something will come in and then maybe he’ll email you in the next hour Peter’s based in New York City, Peter Coffey, Vp. Of strategic research at Salesforce. Peter holds an engineering degree from from Mit and Mba. From Pepperdine. He’s been a faculty member at Pepperdine, at Ucla. He’s lectured at Stanford, at Caltech, at Harvard Business School, Sloan School of management at Mit he’s been at Salesforce for years, and.

Peter Coffee: Isn’t it.

Mike Anastasiades: I was at Salesforce for quite a while. He’s without question not only one of the top thought leaders at Salesforce with a top, thought leader in the tech industry. Peter ordinarily is based in Bellevue, Washington. Today he is in the fantastic city of Chicago. I’m gonna pass this in a minute over to Peter Mulford. He’s gonna be interviewing Peter Coffee during today’s session. But before I do that some logistical items for for the team for those of you listening and watching. If you do have questions, please use the Q&A functionality in Zoom, we’ll we’ll keep a close eye on that, and be surfacing those questions as they as they come up. 

And this webinar is meant to be an introduction to the HIKE2 Innovation Summit. A journey into what innovation could look like for your companies and for your firms. Innovation Summit is going to be at in Pittsburgh, Pennsylvania, August, Wilson, African American Culture Center in Pittsburgh, March th and So if you’re in the Pittsburgh area, we’d love to see you. If you have colleagues or friends in Pittsburgh spread the word. We’d love to see you. We’d love to see them there with that. Enough with the Commercial. Let’s get over to Peter Mulford to get this started, Peter, take it away.

Peter Mulford: Thanks a lot Mike. And and, by the way, in your introduction you you did do a faithful rendering of what BTS does. But you forgot to mention we are also a happy HIKE2  customer. You do. You do great work for us, so I throw that in for you as well thank you for for all you do.

Mike Anastasiades: Thank you.

Peter Mulford: And welcome. So the other, Peter Peter Coffee! Hello again.

Peter Coffee: Good to be with you, Peter.

Peter Mulford: Yeah, yeah, it’s wow. As Larry Mcmurtry put it, life sure does meander like a stream. I think the last time you and I spoke we covered a lot of ground. We got into cybersecurity. We talked about quantum computing, and of course we talked about AI, and I think AI is probably where I’d like to begin, because that’s the one where it’s not the only place where there’s change, but, good heavens! The amount of changes is.

Peter Coffee: Yes.

Peter Mulford: Astonishing we could. We could start by building a bridge between our last conversation. And now, when last we spoke, you were, you were very optimistic about the potential for AI to reimagine work and to to make a better future for all of us. So perhaps we could start by having you catch us up on what’s going on in your head? What are your thoughts about the current state of AI? And is there anything that’s happened since the last time we spoke that has particularly surprised you as a futurist.

Peter Coffee: Well, not so much surprised as that. There’s an expression that many people can tell you what the world will look like, and they’re called Futurists. But those who can tell you when it will look like that are called billionaires. It is very much about the timing in April of So going on years ago, I wrote something internally about why people should not focus too heavily on conversational generative AI as being the new path that AI would henceforth follow. And suggested that in the same way that visicalc on Pcs completely changed people’s relationship with quantitative modeling compared to what it had been when it was cobol on a mainframe.

You know a cobol on a mainframe. You could when computing was capital, intensive skills, intensive. You could only have one. There was actually a photo essay in the Life Science Library book of the engineer that said, today, the computer is becoming almost as important to the engineer as his slide rule, and every major school of engineering has one.

You had a computer, and it did general purpose things. And then there was this, this metastasization, where all of a sudden everyone could have interactive, bespoke iteratively refined, quantitative modeling on a machine that pretty much anyone who had a question worth answering could afford to buy. And I suggested in that we should expect to see that trajectory followed by generative AI as well, and I think many people would agree that the advent of Deepseek only a few weeks ago now might have been that punctuation moment when suddenly wait a minute. It’s not going to be a massive, highly centralized, extremely general purpose capability that transforms the way we relate to machine intelligence. It’s going to be everyone being able to have the one they need curated on the data they trust, that is relevant to their problem and have it be something that learns with them and adapts to them and serves them personally and interestingly, this, I think, is very much the architecture that apple is proposing with what they call apple intelligence, where to the greatest possible degree the intelligence and the data live at the edge. You don’t overwhelm and saturate a shared network or require literally nuclear power plants to be built to power, the training of large general purpose models. Rather, you push the power as close as possible to the point of use. You maintain possession, control, and trust in the data as close as possible to the party with skin in the game and we have already seen that that trajectory is, I think, well underway, and I think that’s an observation that I could have said was inevitably going to take place months ago, but that now I think people would say, Okay, the reality check event has happened. I now believe that that curve is hockey sticking, and that would be the biggest change, not so much a surprise, but now I can put a time on it, and the time is now behind me. Instead of an estimate of in front of me.

Peter Mulford: So are you saying you’re a billionaire, Peter.

Peter Coffee: No, I am not because I did, because I did not, in fact, predict that timing. Had I sold Nvidia Short on one particular day a few weeks ago, and then been prepared to harvest over a matter of a few days. I could have profited handsomely, but I did not.

Peter Mulford: Got it.

Peter Coffee: Now you asked you asked if there was anything else, and the other thing that I think would be a really important transition that’s taking place. Eric Brynjelson, at Mit, talked about the Turing trap, TURI, NG. As in Turing Test, where the early notion that it will be intelligent if you can have a conversation with it, and not be sure that it’s not a human being that that will be the proof that it is now what we call intelligent, pointed out that the result of that, in effect, definition of intelligence was to channel decades of AI research toward that goal of being able to carry on a conversation. And Eric’s ironic observation is, Yeah, but we have things that can carry on a conversation. They’re called human beings. What human beings do quite poorly is entertain multiple, differing hypotheses simultaneously, even in a room full of people. The st good idea that’s offered anchors the conversation, and it’s difficult to break the conversation away from refining the st good idea into well, what about some genuinely different approaches the idea that your AI could be better than a human at entertaining several ideas simultaneously Marvin Minsky used the phrase, Society of mind to suggest that what you really want is an AI that isn’t like a human, but is like a room full of humans who are not intimidated by each other, and can maintain original lines of thinking that synergize each other and building systems that can do that is something that I hope we will try hard to do. It is not the path of least resistance. When chatgpt through a conversational interface on an Llm. It effectively choked down the possibility of multiple hypotheses into something that now has a single thread of conversation with a human user. And the metaphor that I’m using is that if you’ve ever been at the beach and seen a plane fly by towing an advertising banner. Well, you could have a towing a much, much larger banner, or you could have a drone show where there are many, many small entities that are able to cooperate collaborate today. A drone show doesn’t just merely move from static image to static image. We had a drone show at Seattle on New Year’s Eve, in which the drones were delivering animated displays like an airplane flying around a rotating globe, and I did a little bit of investigating on. How do you do that? And the authoring software for these things now is very abstract. This is the animation I want to achieve, and the back end works out all the flight paths of all the drones, so they don’t collide with each other. They’ve achieved that level of abstraction where I can now visualize the outcome and have the innards coordinated for me in this. In this physical world we’ll try to imagine, ais, that work more like a drone show, and less like a towing the biggest advertising banner you’ve ever seen, and you’ll see where I hope we can go. It’s certainly what we’re salesforce trying to do with our agent force offering is to build the infrastructure in which orchestration of multiple agents will be a natural path instead of just one bigger and bigger, smarter and smarter Oz, the great and powerful. Behind the curtain.

Peter Mulford: Well, okay. So as as always, Peter, there’s a lot in there that was a that was a cognitive fruit smoothie that was very dense with nutrients. Let me see if I can take a step back and.

Peter Coffee: Please, do.

Peter Mulford: We can wander over that terrain step by step. But I heard you talk about a shift from AI moving data centers to edge devices.

Peter Coffee: Yes.

Peter Mulford: You talked about what I think it was Mustafa Suleiman at Microsoft, referred to as the shift from you know AI as the chat bot to polymathic thinking. You know.

Peter Coffee: I would not disagree with that characterization. No, yeah.

Peter Mulford: Yeah. And then, you. You referred a little bit to Agentic AI, which I know.

Peter Coffee: Yes, I do.

Peter Mulford: When I know there’s something that a Salesforce where you work is is really leading the charge on that.

Peter Coffee: Thank you for assessing it that way. I appreciate that.

Peter Mulford: Yeah. Yeah. And then, of course, related to all, this is the, you know, comparing where we were in when you saw some of these shifts to where we are now. Nothing surprising but nonetheless enlightening, and it sounds like buried beneath all of your your comments. You see, there’s some really interesting potential for reimagining how work might get done, particularly if we can move away from the you were talking about the the Turing test systems to more of this, this polymathic thinking. So let’s.

Peter Coffee: And another key development, because there should always be would be that the prospect of attempting to regulate this is no longer prospective. I think that the I think that it is since the last time we spoke that what people refer to as the European Union’s AI Act has now been put on paper, and in theory some of its dates, its critical dates have have now arrived, and I have Cavills with that one calling it an AI act, I think, overstates its power to control, because if you read the document, it is really and some pages of guidelines, as Captain Barbosa says in Pirates of the Caribbean, it’s not so much a code as a set of guidelines, and it is a directive, if you will to the Member States to have their legislatures write actual laws, and if you do, a simple text search of the AI Act for the phrase, Member States should do something as opposed to, you know, must. You’ll discover that there’s tremendous opportunity for variations in implementation of that act, and the idea that now there’s going to be. One set of rules is is vastly overstated, and also the implementation dates of some of the most important provisions of that framework are years ahead. So the idea that it’s being going to be regulated is no longer a hypothetical, but the idea that it’s already being regulated is, I think, overstated.

Peter Mulford: Well, I think we’ll get into that. And of course you know that right now, as we’re recording this group of Internet politicians and business leaders are meeting in Paris this week to talk about this exact topic at the summit there. So let’s

Peter Coffee: Yes.

Peter Mulford: Let’s rewind the tape and let’s start at the let’s start at the very beginning. So,

you know, I think I you would probably agree, or you. You’ve noticed that many of your peers in and around the industry have described AI as variously as the steam engine of the th Industrial Revolution. I think that was the expression that was being thrown around Davos during the World Economic Forum. Yes, and that kind of language definitely makes head headlines. But if you look past all of the hype and the noise, and you know, really plant your feet in the ground.

How transformative is AI, I mean, really, compared to some of the past technology shifts that you were referring to at the the start of the call and and related to that for the non technical leader trying to get her head around this what is it that actually makes it unique relative to some of these other transitions?

Peter Coffee: Sure. Well, the phrase th Industrial Revolution was certainly heavily brute about following a world economic forum event. And yes, the the steam engine was often invoked as a metaphor, but at the time I took a pretty strong position. That, mistaking the artifact for the Revolution was was a mistake. To avoid that.

Peter Mulford: Okay.

Peter Coffee: The steam engine made power scalable. It previously was the case that if you needed a horsepower you needed a horse, and if you wanted horsepower. You had needed horses. Well, Robert Fulton’s steamboat was a horsepower motor.

But what made it the breakthrough that it was was that that motor was not so massive that it used all the payload capacity of the boat, and that the technology to build a steam engine, that small and import it to the Us. From Bolton and Watt in the Uk. Was treated probably about the same way today that getting an Nvidia chip in Beijing would be. It was treated as a strategic technology, and the massive centralized steam engine could not be distributed to a small artisanal workshop. You had to build a massive factory to use all that power. So what happens? Manchester, England doubles its population twice in a decade, and there’s a massive migration of farmers to cities to become factory workers.

The migration of the people was the revolution. The steam engine may have catalyzed the Revolution, but it was the transformation of an agrarian workforce to an industrialized factory workforce and a capitalistic model. That was the true revolution. And so from there the transition from steam to electrical power becomes interesting, because now you can build small electrical motors and not just it’s hard to build a pocket sized steam engine. But you can build miniature electric motors. You can have hand tools, you can have workshops and power tools. And so we added, from scale of energy capacity, we added distribution of that capacity and notice.

The revolutions are cumulative, not successive. We had power. Then we had distribution of the power. Then we had a rd revolution, where we added information to the Watts, and to the volts and dampers and the crude watts of energy. Now we added communication and control to that network, and that was arguably a rd revolution. And then, of course, we had digital computers to to perform calculations on that. I was at the Exxon Baton Rouge chemical plant at the time when we put in the st digital control room and we literally went from an environment where at the beginning of the day, you would go to a wall-sized cabinet. Open it up, and there would be the pen and ink chart from the chart recorder over the night, and you would look at it to see if anything bad had happened, and if nothing terrible had happened you would file it away.

And we put in the st digital control room. And now it was possible to do quantitative analysis and get pre-failure warnings of things, and it turned into information and not just a record of power consumed. And that was the rd So what is the th revolution? It is the addition of machine power to optimization and planning, so that now we don’t merely have systems of record. But now we have systems of recognition and forecasting and prediction that can actually do, you know, useful

Peter Coffee: foresight. And again, optimization, rather than merely recording of what we have done. It can give us guidance onto what we should do, and that is arguably what the th revolution really is, from scalability of power to distribution of that power to communication and control that add value to that power. And now, any time dimension making prediction and optimization of what we do with that information become available pretty much to anyone.

Peter Mulford: So if I if I follow your analogy. So you were making the point that, hey? If we go back to the st Industrial revolution. The steam engine was the catalyst. But what you notice is the transformation really had more to do with the migration of labor, and what that labor did, and its transformative effects on society around it.

Peter Coffee: Yes.

Peter Mulford: In this shift. You’re making the point that we have a shift towards, you know prediction and accelerated computation by any other name.

Peter Mulford: What’s the? So if that’s the if that’s analogous to the steam engine, what is the actual transformation you see happening, or that you can imagine might soon happen.

Peter Coffee: Well, it is already happening. The good news is, it is already happening. As William Gibson is accurately quoted as having said, the future is already here, just very unevenly distributed, and you can find pieces of this future already taking place. I commented on how you had that agrarian workforce all move to the cities to become factory workers. Well, at the same time, their children who had previously, as soon as they were old enough to hold a tool and stay out of the way of the sharp instruments were working side by side with their parents and becoming collaborative problem solvers. But then, once Mom and Dad are going to the factories. The kids have to go to classrooms, and you have a factory model of learning which is still with us.

The children still show up on time, sit where they’re told, follow instructions and go home when the bell rings. So the paradox is that even as employers are saying, I don’t need that kind of worker. I need a collaborative problem solver, but we have a massive inertia in the way we educate young people to produce effectively the future factory worker in a world that isn’t really going to be wanting that. But what you can already see is that there are elementary school teachers. Who say, How can I ever teach a child to write when a generative AI can write the essay for them? I have seen essays by former college professors who have recently left the field because they have despaired of the possibility of ever getting a student to do the work required to deliver an A or a minus grade of written work when they can get a B plus in  min from a Gen. AI.

But the elementary school teachers who are seeing the lemonade instead of the lemon, are saying, Wait a minute now. Instead of laboriously learning to craft a paragraph, I can have an elementary school student be be shown. This is a prompt that was given to an AI. This is the paragraph that it wrote.

“Criticize.”

And you can begin the process of becoming a critical evaluator and not merely a laborious minimal composer of content so much earlier now, and we can raise a generation of critical thinkers who can all have in effect a personal tutor at their elbow and in their ear, helping them become better at what they do, and helping them learn to be improvers of work, instead of merely having to struggle, to learn to do that work.

Peter Mulford: Okay, so let me let me let me double click on that for a second. So what I heard you say there and again, as always. If I put words in your mouth that don’t belong there, just pull it out and.

Peter Coffee: I’ll let you know. I’ll let you know.

Peter Mulford: But you you shined a light on perhaps one of the st areas where AI could be transformative. And that’s in education. And one of the things you notice is that for some educators, not everywhere, but for some, they’re seeing that there’s an opportunity to look at AI, not as something that’s a shortcut to garbage, but actually can be a tool to stimulate.

Peter Coffee: Yes.

Peter Mulford: Critical thinking. And you probably saw the news in in your your home state? Oh, actually, no! You live in. Do you live in Washington State, Peter.

Peter Coffee: I’m in the State of Washington. Yes.

Peter Mulford: That’s funny. For some reason, just because you’re big brain. I just assumed Silicon Valley. But you’ve probably heard in California the Csu California State University system just gave Chat Gpt. students. And I think it’s faculty. So now everybody gets it. And this is, of course the largest organizational deployment of Chat Gpt in the world. Yeah. And of course, what? What must be sitting downstream of that is both students and faculty that are of the same mind as you who are realizing. You know this isn’t something to be.

Peter Coffee: What one hopes.

Peter Mulford: Snaker.

Peter Coffee: Webinars. Yeah.

Peter Mulford: One hopes so. What I heard you say there is things you talked about, one way in which it’s transforming how something gets done, and in this case Education heard you say, what goes along for the ride is a mindset shift.

The have to go together in order, for in this situation to be transformed, you have to think about it differently. So if we hold those those twin gargoyles in the air for a second.

Peter Coffee: Okay.

Peter Mulford: And step out of education, and perhaps look at the business world where many of our listeners yes, there’s again a lot of buzz about the potential for agentic AI to you know, various. I’ve heard hundreds of millions, or even billions of dollars in annual productivity. Of course, salesforce is is one of the companies that that’s shouting in that direction.

Peter Coffee: We we hope to be helpful in that area. Yes, yeah.

Peter Mulford: Yeah. But but the other thing about you, Peter, is is you’ve always had your feet firmly planted in the ground, even though you’ve been an optimist. So from where you are standing, obviously, AI isn’t going to transform everything overnight. But where do you see the most significant economic opportunities to turn those those dials in, say the next to years I mean to it’ll probably be everywhere. But if I’m a business leader, and I’m thinking where I should be paying attention. Where do you see it? In the next to 

Peter Coffee: I try to avoid making predictions by cheating and looking at things that are already baked in. And just a matter of this is definitely going to happen. I sometimes say.

Peter Mulford: That’s right, and you become a millionaire. I want you to.

Peter Coffee: Well, I point out that the light that left Proxima Centauri and a rd years ago is going to hit Earth today, and I don’t call that a prediction. I think that’s really going to happen. I’m pretty damn confident that’s going to happen.

Peter Mulford: You can.

Peter Coffee: And the same way, demographic change is not a prediction. If I know how many year olds there are today, I’ve got a pretty good handle. On how many year olds there will be years from now it’s not really a prediction.

Peter Coffee: And by the year you’re going to start to see a population in which the number of people over and the number of people under are approximately equal. For the first time in human history all of our institutions of education, hiring training and development, how we promote people, how we retain people in organizations, and crucially how we care for their health are going to be transformed by this certainty, that we will be dealing with a population of over s. Over s. Over seventies like we’ve never seen before. These will be the wealthiest, best educated, most demanding, and least technophobic older people you’ve ever seen. Today’s today’s year old, let us remember, was in their twenties when this new thing called a Macintosh came along, and they were explaining to people this radical new idea of, you know. Window icon menu pointer. Well, they didn’t get less tech, savvy? As they got older, and a friend of mine said, Well, you know, Peter. It’s not that the tech that the older getting tech savvy? It’s that the tech savvy are getting old. And so all of the things we think we know about things like online banking, online virtual healthcare, these are going to become readily taken up by the largest and most demanding customer base you’ve ever seen. So if you’re in healthcare, if you’re in education, if you’re in human resources, if you’re in financial services, that demographic certainty of a tech, friendly affluent, educated, ready to take advantage of your service as customer is something on which you can place very big and very low risk bets health and life sciences in particular, because the current healthcare system simply cannot scale to deal with the size of the population that is not going to be dying in their sixties, because they won’t have been smoking, they will be in better physical shape. They’re going to be lasting longer. So instead of dying in a matter of months following their st heart attack, they will be with us for years of needing lifestyle, monitoring management.

What that means is that the role of AI in healthcare is crucial. I was with someone from Pfizer just yesterday who pointed out that % of medical professionals time today is spent on paperwork. And if we can have AI meaningfully accelerate the process of preparing and completing reports, transcribing notes, taking the vast quantities of unstructured data that are currently, you know, the atherosclerotic plaques of the healthcare system and streamlining that well, doing that math that would be roughly a rd increase in the amount of medical care that a given medical professional can provide simply by giving them the advantage of those power tools to do the things that are not high value, but are necessary parts of the healthcare process.

Peter Mulford: Okay, so let me let me pause and linger on that point for a second. So what I think I’m hearing you say there is is kind of interesting. You’re pointing out that you know, rather than focus on the the path. Technology is taking one place to look to see where technology would be interesting in the future, is demographic.

Peter Coffee: Well, all of the important changes will not be technologies looking for something to do. There will be demographic or climate, related environmental necessities creating a need for a kind of capability that is not scalable with current technology that does not have the Talent pool that it needs to do it with humans alone. There was a comment in the chat stream about the role of the payers, the health insurance providers, and so on. In all of this. And I’ve had some very productive conversations with them where they’ve I pointed out that they’re currently perceived as almost barriers to healthcare delivery. And there’s been some considerable controversy about that perception. Of late they have an opportunity to change completely their role in the system, because medical care is about pills and procedures and office visits. That’s what it does. But a payer can do things like establish an affiliate relationship with the frequent purchaser program of a major grocery chain and start to give you health care, cost incentives and discounts. If you’re purchases of food, indicate choices that are, you know, healthier choices, or I can take the data from wearable devices, fitbits and apple watches and and other others, and make that inform the personalized assessment of risk. And, in fact, John Hancock Insurance. At point a few years ago said, from now on, all of the things that we offer in the area of life insurance will have some opportunity to be informed by this quantitative data, we can get on our people adopting and practicing low risk and risk reducing and health improving behaviors, because that affects the assessment of their risk.

So this is the demographic need or the environmental necessity combined with the availability of people to do it or not, the scalability of the current process or not will tell you where the AI will be pulled in and made to work as opposed to what will come out of a laboratory, looking for a place where it can be offered to work.

Peter Mulford: So what what I like about what you’re saying there is. You’re as always. You’re encouraging us to take, you know the long view right and get out and look at the world oper of opportunities from the future back.

What I wonder about, though, is at the same time is is you’re encouraging us to take the long view. In the present moment. You’ve probably noticed it’s simply the case that consultants and vendors and academics alike seem to be really fond of of ringing an alarm bell and saying, you know AI is impacting every industry right now, and that is dangerous right now. Now I I’m sure you would agree that the reality is a little bit more nuanced. So in in your opinion, if you, if you dial back to the present moment.

Peter Coffee: Yeah.

Peter Mulford: And you you. From where you sit you look across industries and tasks. Where do you see the most rapid change happening right now? And what would you say? Are the specific use cases that are driving that change.

Peter Coffee: Well, the long view of what’s going to be needed when you bring it back to the present moment can sometimes create a substantial urgency of doing something right now. Because if you know, you’re going to need something, and if you’re going to have a critical need years from now, and it’s going to take years of work to have that implemented. Then you have one year to get that started, or if it’s going to take years. You should have started it years ago and the connectivity of data, the management and governance of that data, who owns what privacy protections will be applied. What techniques can be used. For example, what Apple has talked about. I don’t think they coined the phrase, but the phrase differential privacy. And there’s another one called homeomorphic Encryption, where I can protect certain aspects of data that should not be widely distributed. Personalized aspect while still preserving epidemiological value, or still preserving climatological value of that data? Do I want people to know where I personally have been driving every minute of the day? No. Would I be okay with the data on the movement of my vehicle being aggregated with the movement of other vehicles to optimize traffic flow in a city and reduce the amount of pollution pumped out by cars sitting in bumper to bumper traffic. Yes. And so there are techniques that can be used.

Given massive connectivity, given massive computational power and given regulatory frameworks that prevent a race to the bottom of doing it in a sloppy way, because it’s cheaper if we can raise the floor on, on, how strict the regulation and control needs to be, so that the worst case becomes acceptable, then everybody can enjoy the benefits of having that available at a lower cost. You don’t want to race to the bottom. That’s the path that we need to work actively to prevent.

Peter Mulford: That’s interesting. So what I think I heard.

Mike Anastasiades: Hey? I’m sorry, Peter. Peter. I’m sorry. I just wanted to quick, quickly jump in. And I know. Yeah, there, there’s a question out on the Q. And A. And I did. I want to continue to.

Peter Coffee: Please. Yes, absolutely. The more interaction we get, the better. Yeah.

Mike Anastasiades: Yeah. So this is from Robert Barker for Peter Coffee. You hinted at companies. And and you know, AI evolving and growing together. Is there a risk of human AI group? Think, Peter Coffey.

Peter Coffee: If what that means is reduced. Diversity of thought, the the phenomenon of generative AI increasingly finding that the easiest data to get is the output of other generative AI, which is either the cost reducing technique, so-called distillation for producing models at lower cost or just the inevitable result that if you have it reading everything on the Internet, an increasing fraction of that is machine generated.

That’s already been shown quickly to reduce the quality of the output that you get. I hesitate to use the word creativity. It sounds too anthropomorphic, but the the degree to which you can be impressed by the possibility the Aol will produce a useful surprise, quickly shrinks. If the AI is consuming each other, if that’s what he means by human AI groupthink, I completely agree. And among the reasons why it’s so important to have our educational system educating critical thinkers, collaborators, and problem solvers, instead of merely mastering subject matter. Knowledge is precisely that we need to keep that human for lack of a better word, that that human element in the system, so that it is continually stirred so that there is continuing fresh oxygen being stirred into the pool before it becomes stagnant. I think Robert had another follow up question.

Peter Mulford: Yeah. And I, you know, I just want to chime in on that as well. What I would say to that is just echoing Peter Coffee’s point by keeping a human in the loop. That also means that if you, depending on how you interact with AI, it is actually possible, by dint of your interactions to keep groupthink from happening. And you know, the easiest way to think about it is, are you using AI as, say, Google on steroids, where you’re just asking for an answer to something, and then accepting what you get, or you know, Peter, you you said you were you, you inferred you were loath to anthropomorphize it, I would push back on that. Say, you do want to anthropomorphize it, because if you start to think about it as a very well read very naive coworker. Yes, it changes your your dynamic with it, and.

Peter Coffee: I’ve used that that metaphor of imagine that you have an intern who does nothing but read stuff, but actually knows nothing about the business. They may make suggestions that times out of are okay. Cool idea. But you didn’t know this. Okay, fine. Now it knows this. So there is an opportunity for interaction there. And Robert talked about talking about companies developing their own AI instances and using their own data. It is tremendously important that the curation of data not become a moat outside of which you don’t pay any attention to what’s happening.

People talk about the Moore’s law rate of improvement on transistor density and use that as a proxy for the improvement of computing price performance. But at point I drew the curves based on the cost of something as simple as the modem that used to be a separate box next to your PC. And then became something that every PC. Had built in the rate of cost per unit performance of connection.

Improvement has been vastly greater. The ability of your computer to connect to and ingest data from all kinds of sources has grown more quickly than its ability to do computation on that data. And this is not generally noticed because people think, wow, the machines are so much faster. Yeah. Well it’s kind of a paradox that people say, if my machine is so much faster, why do I? Why am I not really feeling like, I’m getting a faster computing experience. And the fact is the quantity of stuff that you’re asking your machine to digest and and find value in. It has grown faster than the computational power of the machine it used to be. You would ask it to read a document. Now you ask it to watch a video and tell you what happened. That’s a much more intensive task. And it’s been particularly interesting to see just over the last few years how the architecture of the processing power of a laptop computer has has changed to meet that change of workload. If you look at what an apple m. Chip is designed to do compared to what say an Intel? Chip was designed to do? You can see that the processing power is not merely growing, it is reshaping itself around the kind of workload that we expect AI to do. I resist the idea of calling the result an AI chip but it is a better front end to an AI system in that it is optimized for ingesting and extracting information from the kind of data flows that are now so increasingly abundant.

Peter Mulford: Yeah, I think just to put a fine fine point on that, Peter. If you wanna get a a an experience that’ll really illustrate what Peter’s talking about. Just sign up for ChatGPT Pro and ask it to do some deep research for you on something called the o reasoning model, and what you’ll notice immediately, you’ll actually have an opportunity to see the AI going from step to step to step to step. It takes a while, and by a while, I mean, we’ve grown so so accustomed to just typing in a prompt and getting a response, that oh, wow! You know it’ll take you maybe a half hour, but a half hour to do a week’s worth of work. And you know the thought, I don’t know Peter, about you. But the thought that we might actually get to a point where a reasoning model could move as quickly tomorrow as, say, a a Google search would work today is, that’s kind of a mind blowing mind blowing thought.

Peter Coffee: That’s not a controversial projection of the continued growth. People who say, Oh, Moore’s law is leveling off have consistently been proved wrong every time an S. Curve begins to plateau, the next S curve takes over, and this is something that Bill Joy observed years ago, is that the aggregate of one S curve following another S curve following another S curve can be a hockey stick that goes on for quite some time, but there are at least Points. Now that you’ve unlocked that I want to be sure we explore.

One of them is the work being done by Tom Malone of the Center for collective intelligence at Mit, which asks simple questions like, What’s the IQ of a group? It is not the average of their Iqs. It is not the IQ. Of the smartest person in the room. High IQ. People, with radical differences in viewpoint, might actually have a lower intelligence in any useful way than either of them would alone. And when I had a conversation with them once I asked. Have we screwed ourselves by going from an email model where I see a question, and I get a chance to think about what I think about it before I see anyone else’s opinions to more and more instantaneous chat or video conferencing. Where now we’re back to the old behavior of the st person to blurt out something halfway interesting anchors the conversation. And and Tom looked at me and said, We don’t know yet. But yeah, we need to explore that. So you may need to think about optimizing your manner of collaboration to preserve actual independence before aggregation. And these are words I’m picking from a National Academy of Sciences paper that talked about this so-called wisdom of crowds effect. It turns out that if aggregation comes too soon you don’t get a wise crowd, you get a dumb herd.

And so that’s of the aspects is that we need to craft environments in which human creativity is not in effect suppressed by the technology getting so good that we no longer get the chance to do independent thinking. We do get group thinks that’s of the issues. Another thing I want to make sure, we spend some time on, because it scored quite high in a survey of attendees. Interest is what’s the impact of a quantum computing model on all of this? The hardware is still, if it is on the transitional zone from science project to prototype of of a real machine. But already people thinking about how would they do? Computation on a quantum machine where, for those not familiar with more than just the phrase, imagine that that bits can be either or but they can only represent one of those States. If their quantum bits or so-called qubits, then those bits can, in effect simultaneously represent all States, and you can conduct computations that are now not feasible in terms of time or hardware in ways that now can literally do in minutes what would otherwise take hours, or even centuries, to do with more classical devices. This is already being applied to tasks from financial risk, assessment to grocery shelf, restocking merely by thinking how would I do this? On quantum hardware? New algorithms are emerging that had not come to people’s minds when they were thinking about a single threaded von Neumann machine architecture like we’ve had. And this is such a fascinating result of merely thinking about, how would I do this if I had quantum hardware?

People are writing software that already achieves significant performance breakthroughs even on the hardware that we have today. And I love to see that kind of unexpected synergy and catalysts taking place. It’s it’s so much fun to see that happen.

Peter Mulford: Okay. So you just you just managed to open a door to a completely different universe. I think we should step through on on quantum quantum computing. But perhaps before we do that one last thing you said a moment ago, that click on is you? You talked a little bit about AI regulation. At the very start of this you referring what was going on in Europe.

Peter Coffee: It’s an important question.

Peter Mulford: So maybe we could. Before we step through the door of into quantum computing, we could linger for a moment talking about ethical AI development and deployment. And you know there seems to be a really interesting debate that if I were to parse it into there’s this argument that use cases should be regulated, not the technology and vice versa. So I wonder? I’m sure you’re familiar with the debate. Do you agree? And if not, what is your point of view on the the right way to do that? And and what would you regulate, or would you recommend? We regulate.

Peter Coffee: The history of attempting to regulate technology in general and information technologies in particular is rife with. Examples of this seemed like a good idea given the technology we had at the time, but the regulation became a perverse block to doing things better. For example, early attempts by the EU to regulate information. Governance resulted in administrators of machines being prohibited by law from doing perfectly ordinary administrative tasks. Because if you’ve got a well-intentioned law that says you may not manipulate a person’s owned and created information without their permission. What happens when an employee leaves a company, and you want to remove their account from the corporate system.

Literally, you are destroying information created by another person without seeking or obtaining that person’s permission. Literally you are breaking the law, and the degree to which legislators fail to think in sufficiently general terms before they write proscriptive rules is a long observed pattern, and that can easily happen. This, which is why, if someone says you should regulate the use, not regulate the tech, I think that’s where that comes from. And I think there’s a lot of validity to that.

The other observation is something that was said by someone at our Federal Trade Commission at a meeting. I was at once when he said just because you’re doing it on a computer doesn’t mean you need a new body of law to control it. Fraud is fraud. You don’t need a new body of law against cyber fraud. What you need to do is work backwards and identify the areas in which your current laws make unconscious assumptions based on the technology that you’re using now and peel those away so that the law can do its real job of controlling that behavior. And I’ll give you one of my favorite examples. A case in which a British court was was being asked to adjudicate a claim of forgery and the dialogue between the judge and the attorney is wonderful, because the judge says, my learned esquire, the statute of forgery requires that an instrument be used to obtain access to an asset is the crown, prepared to introduce into evidence the instrument, because this would normally be a document with a signature that had been, you know provably, you know, not the real signature, and so on. And the lawyer says, Well, my lord, the the instrument was a pattern of bits in a computer, it no longer exists, and the judge nods and says, Okay, and and the asset to which access was granted, given, well, my lord, the act of forgery was used to obtain access to information on trading in progress, and that asset no longer exists either. And the judge said, Gentlemen, what the defendant did was undoubtedly a bad thing, but Parliament has yet to write a law that makes it a crime. And that’s such an interesting question is, if you look at your laws, do they make unconscious assumptions, unexpressed assumptions about the manner in which a bad thing will be done, and then regulate that behavior, instead of sufficiently defining the harm that would be done and making the harm. The crime, instead of the action, which is now an anachronistic, you know obsolete idea, the crime! It would be as if it would be as if a law against murder assumed it could only be done with a knife and didn’t know anything about firearms. It would be almost like that. So that’s of the questions about AI regulation is, are you clearly defining the harm you seek to prevent and criminalizing that harm? Or are you taking a shortcut and criminalizing the use of a particular technology in a particular way, in a way that becomes obsolete and irrelevant.

Peter Mulford: Interesting. So stay tuned.

Mike Anastasiades: Quick time check guys. We got about  min left.

Peter Coffee: Yes, indeed, yes, indeed, thank you. Thank you.

Peter Mulford: Thanks for that. Mike. So let’s let’s let’s jump to quantum computing for the the minutes that are left. Now you at the just. A few moments ago you were you were talking about qubits, qubit stability, and inferring error, correction, and you know the rest. But for anyone that you lost. I could imagine.

Peter Coffee: Might have lost a few people there. Yeah.

Peter Mulford: The conversation. If we bring the we bring the conversation down to earth just.

Peter Coffee: Happy to do that.

Peter Mulford: Simply put, what do you see at this moment in time as the most promising near term applications of quantum computing that a non technical person ought to be thinking about, even if she or he is kind of lost in in, or is not really up to speed with what the.

Peter Coffee: The most urgent understanding of quantum computing is this. People often make statements like, well, you know, computers can only think in ones and zeros, they say. Well, computers aren’t like people. They can’t see the shades of gray and anything. They can only see black and white. And if you go back literally to When Norbert Weiner wrote cybernetics, he actually shows why, with the hardware available at the time. Ones and zeros were the only cost effective way to represent data. There was no technical reason why we couldn’t say everything is a to one in increments of th We could have done it that way. But the economics were such that zeros and ones were just the best way to do it then. Acting like that’s some fundamental property of computers is a mistake, because with quantum bits we can, in fact, represent shades of gray, or even some or even multiple shades at once.

Why does that matter right now?

To a huge extent, your ability to live in a digital world depends on the availability of encryption that you trust when you’ve got https at the beginning of a website instead of just Http, when you make a purchase with a credit card online, when you have your home control, your thermostat, your car online, the assumption that data streams can be encrypted can be tamper evident, if not tamper proof with technologies like blockchain, that confidence in the integrity and reliability and privacy of the system is all based on one assumption in most systems today, which is that if I take very, very big prime numbers and multiply them. It’s very difficult to figure out what those numbers were after the fact. It’s a -way function. I can produce the product cheaply, but factoring the resulting number back to its original factors that were multiplied to get. It is very, very difficult with current hardware. That’s simple idea. It’s not a complicated idea. You can illustrate it easily with with any schoolchild who knows what prime numbers are.

Quantum computing makes nonsense of that reliance on that simple property of numbers. There are algorithms which, executed on quantum hardware will already shred most current modern forms of encryption. There are algorithms that are quantum proof. National Institute of Standards and technology is well into the process of declaring quantum resistant encryption techniques they’re more expensive to use today. You probably won’t use them today.

But if you’re recording data, or if you’re transmitting data, it is already known that bad actors are capturing encrypted data streams, knowing that in the next years they will be able to attack those records with, as they say, cryptographically relevant quantum computers or Crqcs, as the industry calls them. So if you are not today handling your data in a manner that anticipates the imminent possibility of it being attacked by a quantum computing, equipped threat actor Federal Government has already told its agencies under the previous administration, that they should be assuming that by their data streams will be subject to effective quantum attack. And that changes behavior. Right now, today, today, you need to be asking, what do I need to do to protect data if I care about it being still private. years from now. Speed speed trading on Wall Street. Not so much. But important national security or corporate information. Maybe that needs protection. That’s good, for, you know, to years, in which case the advent of quantum computing is a is a future that’s now.

Peter Mulford: Got it. So what I think I heard you there say there, with just  min to go. There was a lot in there again. But it sounded like you were saying. If you’re you know, if you have intellectual curiosity about how the systems work, you know, maybe listen to Brian green over the world Science Festival, and he’ll take you on a journey through quantum mechanics. Yes, for the rest of us. You’re talking specifically about Pqc, right?

Peter Coffee: The practical reality is this affects data security. This affects the stability and reliability of our electrical grid and other infrastructure. All of these things assume that they will be protected by robust encryption, which becomes considerably less robust in a quantum relevant world, unless there’s been anticipatory action taken which needs to be an urgent priority.

Peter Mulford: Well, Peter, as usual, my head hurts after spending an hour with you, but it hurts in the best possible way. So I think maybe I’ll turn it back to our host Mike to close it down, Mike. Thanks for bringing us together again. That was that was a lot of fun.

Peter Coffee: Thank you, Peter. I appreciate the questions, the questions. I never know what I think until a question is asked. It’s a Schrodinger’s brain kind of thing. So thank you for the dialogue.

Peter Mulford: Terrific.

Mike Anastasiades: Questions, great answers, great session. Thanks. Everyone quickly wrap up. I mentioned our innovation. Summit March th and So again, if you’re in Pittsburgh in the area. Please stop. And we’d love love to see you. That is our Summit just quickly on that, you know, top leaders talking about you know where they think. You know, business is headed in the future very similar to this session that you you you saw today and listen to today. Breakout sessions, panel discussions, workshops led by many industry experts. And again, sort of a fun and engaging way we think for you to network. Talk to peers, industry, leaders, other like minded professionals, I’d say so again, March 26th and 27th in Pittsburgh. And and please you know we’re hype, too. So if you could, you know, follow us on Linkedin, you’ll you’ll, you know we host thought leadership sessions on a pretty regular basis. Usually every weeks or so would love for you to join those enormous thanks to Peter and Peter as we get to top of the hour here for delivering a a very thoughtful, very impactful session today, a million thanks to all of you for joining and thanks for participating in the Q. And A, and have a great rest of the day, everyone, and we’ll talk to you all soon.