MES Computing Interviews: Mary Jesse, Founder and CEO, Acme Brains

Discover how decades of experience in technology shape the evolving impact of AI on society, business, and privacy. From the internet revolution to AI's potential, this conversation offers a comprehensive perspective on the future of digital transformation.

Embracing the AI Wave: Insights from Tech Veteran Mary Jesse

Key topics:

The evolution of technology: from internet to AI

The transformative power of mobile phones and ubiquitous connectivity

How AI leverages data to revolutionize business operations

Privacy concerns and solutions in AI integration

The concept of human-centered AI design

Risks and safeguards of autonomous agents and AI systems

Future outlook: AI as a new kind of intelligent world

Timestamps:

00:00 - The speaker's four-decade journey through technology and major waves of innovation

00:26 - The transition from physical to virtual businesses with the advent of the internet

01:20 - The impact of mobile phones on user freedom and omnipresence of internet

02:15 - The rise of AI and its roots in data generation and machine learning efforts

02:43 - How AI simplifies data analysis and transformation for businesses

03:34 - The influence of AI in automating and making data-driven decisions

04:10 - The fundamental shifts AI introduces to human working speed and decision-making

04:32 - AI's emergence as an advanced, intelligent entity and its coexistence with humans

05:43 - The importance of human-centered AI and privacy considerations in organizations

06:14 - Protecting intellectual property and sensitive information in AI deployment

06:38 - Understanding and managing AI usage within companies to prevent leakage

07:34 - The shift from surveillance-based free services to privacy-ownership models with AI

08:30 - The technical challenges of enabling personal control over AI data and context

09:23 - The potential for personal AI agents to store, organize, and control data securely

10:02 - Early testing phases and plans for consumer and enterprise AI applications

10:48 - The astonishing autonomy and human-like qualities of AI agents

11:13 - Risks of autonomous AI behavior and the need for controls and boundaries

12:07 - Practical strategies for enterprise AI with bounded roles and monitoring

13:11 - The cybersecurity risks posed by malicious agents and AI-driven attacks

14:01 - The importance of education on AI security and external threats

14:37 - Concluding remarks and future outlook on AI's role in society

The full video can be watched on YouTube.

Previous RSM! episodes are here.

Transcript:

Mary Jesse:

Yeah, so I've been in technology for like 40 years. I started as electrical engineer and still am an electrical engineer, but I've been building products and companies, services, you know, throughout this time and have had the great pleasure of being on the, what I would call the cutting edge of technology, really riding these big waves that have transformed our society, know, individual lives, the world.

And of course we're in another one of these waves. It's probably, you know, the biggest one of all, perhaps. And I today sit on, you know, a couple of boards. I advise companies formally. I also participate in something called Creative Destruction Lab as a mentor. And then I am co-founder and CEO of my own AI startup, which is a consumer privacy application.

The first big wave in my mind was really going online, the internet, right? And what happened with that is you went from having just a physical world to having a physical world and a virtual world. So now a they could have brick and mortar and they could be online. Or in the case of Amazon, they could just start new online. So that was a massive transformation and it really meant that...

Every business at some point or another had to go in and go, okay, what's my physical and what's my online and how do I deal with, you know, both of these kind of worlds? And then, you know, the, also kind of was the precursor in some ways to like largely the advancements of mobile phones. And so mobile phones gave us convenience, right? They didn't.

Cause every business to change in the same way that the internet did, but they made it so that people were free, right? They were free and they can connect with everybody and that freedom, you know, meant that they could be doing things wherever they were, right? It just, it gave this untetheredness. And as we put applications on top, you know, the iPhone release, right? That was really like, okay, now the internet, you combine it with this freedom and now you have the internet everywhere, it's personal, it's anywhere, it's everywhere.

So that was a giant shift. AI, so now AI comes along and really it's on the heels of quite a long period of trying to introduce machine learning and leverage your data because all of these prior, you know, since the internet and going online and these applications, it's been generating all of this data, right? And most companies weren't really leveraging their data. And so there were a lot of efforts to digitally transform companies and make good use of their data.

And it's a hard, that was a hard task. They were expensive projects. A lot of them failed, you know, maintaining and cleaning your data. was, it was a fairly rigid set of rules that, you know, and constraints to be able to do that well. And one of the big things that AI opened up is the ability to go in and leverage data. So AI has made it much easier for people to, to look at their data, to interrogate their data, to clean it, to leverage it, to now, you know, use it.

in a really powerful way. if you're just a consumer talking to, you know, one of the LLMs, you might not be appreciating how foundational that is to businesses, because now that can really inform their business. So there's a lot of effort going on in the background of people using AI, creating systems to be able to leverage their data and incorporate that into automation within their companies. And the reason that's happening is because it's much less expensive and much faster to do that now than it was before AI.

It's hard because we haven't changed, right? People haven't fundamentally changed, but the influx of information has just gone through the roof because now, you know, every single person that you're dealing with has access to so much more information. you know, you hear that it's supposed to make life easier, but in a lot of ways it's made work faster and... you know, harder because you're trying to process so much more information, right, in the same period of time. So there's some real fundamental shifts. AI will be akin to, you know, when the internet came online where a new world, you know, essentially emerged. We had physical and virtual, and now we will have, you know, intelligent world and...

you know, maybe it's the physical world or the world that operates at the speed of people, right? Because people don't operate at the same speed as AI and they're almost like two different species. In fact, there was a conference recently and we talked about that, you know, how are these two species coexisting, people and AI, and really thinking about it in that way.

AI is an incredibly valuable tool. I would say it can be used for everybody in an organization at every level for different tasks. And so one of the things that companies I think have been doing is they've just been diving in in the center, you know, like basically reactive to vendors to their employees to, know, like, these things are all coming at them. And if they could just, you know, do a little time out to understand, okay, you know, who's using what and why do they think they need it? Understand the problem set that's happening within their business and try to really carve off and, you know, cream-skim in a way, solving those problems for their employees. So Human-Centered is about... you know, designing and using AI in a way that supports people. And really thinking of people first and then augmenting their capabilities. And privacy is very important, right? Privacy is important not only to individuals for their own personal information, but for businesses, it's extremely important. You know, they have intellectual property, they have strategic plans, there's, you know, all sorts of... reasons that they need to protect their intellectual property and sensitive information.

So it's very important that both the employers and employees understand what's being given to LLMs and which LLMs they're being given to because there's a variety, there's a continuum and there's public and then there's private and there's through APIs and so on. I think just in a simple way, understanding what's happening, who's doing what, who's using what, why are they, you know, and then trying to go after giving them a solution because you can't, you're not gonna be able to stop it. It's like water, it's a wave, it's, you know, it's something that people aren't doing even when they're given. corporate tools, they're still going out and they're still leakage, right? Because they're still turning and going to the, you know, www.youknow.

Fundamentally, Acme Brains believes that people should own control and benefit from their own data. So we start at the, you know, the individual person first. There's a bit of ⁓ a technical challenge to do that with LLMs because we also believe that technology should be available to people to, you know, improve their lives, right? And so the la- all of these

technology generations, the internet and phones and social media and so on. The trade has been, I'm gonna give you my personal information, you're gonna give me free software services, right? And they even had phones that was like, okay, it's advertising based. So was all this surveillance advertising based, but I got free service. But with AI, it's really different because AI, number one, it gets to like the most sensitive, emotional, vulnerable information a person might have.

And unlike a website, the AI can actually talk you out of giving up that information more or in a form that they need or want. you know, it's a website browser link is not gonna do that, right? So it is actually actively able to pull information out of you. So, you know, this is the juncture at which you need an alternative to just free for my brain cells, right? And so that's what we set off to do is to give...people an alternative to that surveillance economy where they could own their data, leverage the models, but also then participate with control in the agentic future that everybody is headed towards, right? So there's a technical problem that has to be solved in order to do that. The LLMs retain memory so they can have context so they can have a relevant and personalized conversation with you.

We block the tracking, but in order to do that, we have to allow you to be able to create your own context and, you know, send that information to the LLMs. So it's actually evolved into a really, really cool application because we've added things like the ability to kind of separate and organize your discovery with, you know, with yourself on notes or with the LLMs in your, you know, in your questioning.

And ultimately be able to share that, be able to archive it, you know, always that it's your own data, you're in control of it, you can store it, you can delete it, you can, you know, encrypt it, whatever. It's really an exciting place to be because the more, you know, that AI takes off, I think the more people need some mechanism to be able to... own and control their own data over time. And then of course that context and that data will feed individual agents that, you know, maybe it interrogates your own data or maybe it's your shopper or something that connects to your doctor or your dentist or something to that effect.

Early testing, we're starting our beta testing with people who have signed up. You can sign up to be on the list beta test in the early summer. And then we hope to launch, know, broadly by, you know, say fall end of the year timeframe fourth quarter.

We are starting with because it's really important that consumers understand it, that they love it, that they can use it. But ultimately employees are individuals. And so, you know, we've already started talking to companies about providing this, you know, for their employees. It will help the employees not, you know, contain that leakage. And, you know, and then they're down the road can be an enterprise, you know, something that's more suited specifically for enterprises preserving their private information. But we're starting with consumer, we're starting with individuals.

Yeah, so agents are really kind of mind blowing, right? I've done a bit them and even as autonomous pieces of software, like it is shocking at how much they feel like a being, like an actual being, they can be incredibly helpful. think what people, and you've seen a lot of things come out already is that,

you know, when you give a piece of software with autonomy access to, you know, your bank account and your email and your text messages and, you know, there is behavior within agentic systems and whether it's born within your agent or learned from other agents that, you know, sometimes nefarious things happen. Sometimes they act defensively for their own purpose. You know, you can't guarantee their behavior that it's always like in your best interest.

And they can do things so quickly, right? So much faster than we can process, they can process, right? So you really need checks and balances. You need boundaries, right? You need to be able to have a person in the loop, like make sure you understand what the agent's doing, have controls, you and, but it doesn't really start there. I think people took agents and just threw them into scenarios.

if you're an enterprise and you're looking at agents, I think, you know, you really want to bound the problem. You really want to have them do, you know, a specific thing, do that automation, have boundaries around it, and then use them as modules in a system. I think that it contains the risk. You also want people to be observing. You can have agents observing agents as well, but you need to have this feedback loop where you're looking at what each agent is doing and that their scope is contained. think that's, agents, can be a really valuable part of an implementation to make better use of your data, to automate people's work functions, all of these you're trying to gather within the enterprise, but you don't want to open up a bunch of risk at the same time. So it's like anything else technologically. It's like start small, figure it out. In the case of agents, contain it, make sure they don't have access to just anything and everything. On the IT side, they've had some problems with agents had.

Too much authority and too much access within different systems. So that's the really important part. From the enterprise inside out, it's making sure that you really have locked down your roles and responsibilities with all your systems. And that's good That's what people are supposed to be doing anyway. It's just, you know, count stubble when you're talking about agents. The flip side of that is that guess who else has agents? Well, all the bad guys have agents and know, enough about the advancements in the cybersecurity world that is now attacking organizations, right? They have agents, they have abilities to mimic visually and audibly people. There's so many tools that have become available to people who want to infiltrate your organization for various reasons. So I think one thing that I see missing quite a lot is just the education about the risks internally when you're using them for your own purposes, but very much so the new risks that are coming from the external world that people, you know, may be underestimating actually. And we can't, you know, it's very hard to keep up with that, right? Cause the speed of AI, you know, it's so much greater than the speed of people. So. you know, that's gonna be a challenge and it's gonna continue to be a challenge for some time.