How AI is Transforming Web3 With Doug Levin at AIE 2024

Hello, everybody, this is Doug Levin speaking, and, um, I'm gonna present today on how AI is transforming Web3. By way of background, uh, I have a background in, uh, development tools, including artificial intelligence, machine learning. Uh, I, uh, I'm deeply involved in the cybersecurity industry and, uh, Web3.

I've been involved in Web3 for a while, including specifically the, the overlap between cybersecurity and the blockchain, AI and the blockchain and development tools. And of course, I've also involved with open source, uh, having been the founder of, uh, and the first CEO of Black Tech software. Um, in addition to, uh, all that, um, I currently work at the Harvard Business School as an executive fellow, and in the past I've worked, uh, at Microsoft and on the Macintosh development team.

So, let's dive in. In the past, there have been three dominant players in the AI space. Uh, these giants or, uh, monsters, you, uh, if you will, uh, were the, uh, the federal government, which included the, the three letter acronym agencies such as the CIA, the, uh, uh, the NSA and the DOD especially, and of course, other agencies.

But in addition to that, uh, there were large companies. These are operating companies, non-tech companies, which, um, used AI for a variety of purposes. And then, uh, also there were, uh, big tech companies such as Microsoft, Oracle, um, especially Google and other tech companies, which used AI over the years.

Prior to the advent of the open ai, um, offering of, uh, chat, GPT plus, the other generative ai. The reason why these three dominant players prevailed is because computing costs were very expensive. Uh, there was extensive data requirements for not only training, but also operation of the models.

And there was a dearth of advanced skills and, uh, knowledge, uh, in the community. Of course, the, the DOD and um, CIA and NSA had, uh, the best AI guys, uh, in the industry. But now it's a little bit different, which I'll go into in just a second.

But what's important to know is also simultaneously, the AI systems were kind of arcane. They were inefficient. Uh, they had limited, uh, you know, at the time, of course, 20 years ago and more, uh, there was no open source software.

There wasn't a community to interact with. And, um, the, one of the most important reasons that we've learned about, uh, the, the full consequences of, uh, recently is that chat, GPT, uh, Gemini, Claude, and other generative ais did not provide a broad base of AI users to not only contribute to the AI community, but also the open source community and general AI usage. So, by and large, in the past, AI was a hobbyist tool.

Today's ai, uh, landscape has been completely transformed, uh, as all of you know, and this transformation, um, really to a large extent, um, was, uh, has been, uh, impacted by the chat GBT Gemini and the other generative ais, but also behind the scenes, especially in the, uh, dev community, um, by open source tools and software and frameworks, LLMs s SLMs, um, cloud cloud platforms have facilitated it, and advanced AI tools, um, are not only, have not only been in, in in place since the early 2000 and, uh, teens, but also, uh, there are all kinds of new tools which have come along. But every, everybody in the industry is aware today that, uh, innovation is relentless. It seems like every week there's a new LLM or SLM, which is announced, or a new tool or a new, uh, or the intention of bringing out, uh, new tools.

Uh, this relented re relentless, uh, pace of innovation is an extremely positive sign, which, um, shows that, uh, many of the use cases underlying these announcements are either going into place or in the midst of development. And what it also means is that the AI expertise and data is more readily available. This issue of data is also a critical thing that can't be, uh, over-emphasized.

Um, what basically in the past, data has been in silos, especially in corporations, and there's been a bureaucratic entran, which has limited the access to all this data. Today, corporations and governments and other entities have come to realize how important, how strategic the, uh, data is. And so as a result, there have been efforts underway.

I mean, things are not perfect, but there's certainly efforts underway to, um, to de democratize the data, get, give greater access, uh, to, to various operators in the company, um, workers in the company, information workers, and others who can, uh, who are then now more power, uh, empowered to, uh, to do ai. So the picture's different. Um, the monstrous are still there, but that, uh, you know, that AI hobbyist of the past has now, uh, become a much more formidable factor in the ai uh, landscape.

So the challenges to unlock Web3 adoption are also very clear. Everybody is concerned about, um, security and the blockchain, security and Web3, um, assets. And equally important is, are the scalability issues related to Web3?

Now, the scalability issues related to security is are very clear. When you have hundreds of thousands, millions of transactions during the course of time, security has to be highly scalable. But in addition to that, um, handling large numbers of just general transactions such as smart contracts or is also a factor.

And so, um, what also comes into play, and it, and to me, it's very interesting because I see this over and over again in startups as well as in corporate life. Um, there's great complexity in today's, uh, ui. Now, while the generative ais look a lot like Google, they're still complex.

Um, some of them are, you know, even generative ais, their first splash screen are complex. So UI and UX have to evolve. There also has to be seamless interoperability with, uh, the blockchains as well as that, uh, AI applications.

And, um, what, what you see throughout, um, you know, uh, tech society is that there is a, uh, lack of understanding and a lack of awareness around Web3 benefits. This has been an inhibitor to web, uh, Web3 adoption, just as the regulatory environment has become a, uh, thing, which has led to uncertainty a lot of questions and a lot of misunderstandings. So, um, you know, with every presentation, I'd like to, uh, I'd like to identify points of departure.

And we're there, we, uh, with this presentation, I'm announcing the end of, uh, web two. Now, it's not as all dramatic as that, and it's not as definitive, but Web two has been, uh, characterized by browsers and it's highly SaaS centric. And many of these applications are singular in, um, in, in, uh, in, in the way people interacted with them.

So you see developers using Atlassian or other tools, um, you, you know, people park themselves in their browser, uh, for most of the course of the day. And, um, interact with, uh, SaaS and web apps and, um, and, um, all kinds of other information on the web. And, you know, and separate and distinct is their email system, for example, like such as Outlook or Gmail.

Well, AI and Web3 will bring an end to the singular web app apps and the interaction on a singular basis, so we can talk. So we, so as a couple of examples, um, think about, uh, a supply chain application where it answers with a graph if you, if you query it or think about a wellness application, um, where it answers with a subscription, um, to be prescription, uh, information about how you, how to take the drug and, um, make an appointment and it automatically makes an appointment for you in three days. Um, or interactions with web apps, which are highly, uh, which are communication centric like WhatsApp and, uh, slack and, uh, email applications and voice like Siri or Alexa.

When you combine these and one calls to the other, and there's a, uh, there's a response, uh, facilitated by, uh, AI or facilitated by, um, you know, just systems which have been developed specifically to provide a holistic kind of experience for the user, like a, like a wellness system. Then what we have is the advent of the web, the type of web apps that are possible outside of the standard conception of a web app, being an NFT or being a smart contract. These kind of, this is happening today, and as a result, it leads to the idea that, um, web two has, um, declined and, or, or is in transition into Web3.

AI will be also the choice of in, uh, the choice of underlying technology, which drive interface, uh, personalities, customization, and multi-mobile, uh, modalities. The multim modalities are simply, you know, text, um, voice and, uh, other modalities. They're gonna be mixed up.

So, for example, you'll see a disruption with Microsoft PowerPoint, where you can talk to your, um, uh, either your mobile phone or your laptop and just, and ask it to read the, uh, your slides to you. And in so doing, um, the graphics or images in, in PowerPoint, as well as the bullet points and even subtleties can be a read back to you so that you can, um, read your own slides, which may have been helpful here because I could have, uh, caught a couple of, uh, typos. But you could also read other people's slides by simply, uh, asking, um, your system to read 'em back to you.

Dis distinguishing between dialogues and data is gonna be a frequent occurrence. A good example of this would be a maritime distress call, which is initiated by WhatsApp, uh, by a vessel owner, uh, perhaps operating in the Gulf that automatically connects to a monetary application and bypasses traditional channels and, uh, facilitates a swift, swift action, perhaps a, uh, if, if the vessel has been commandeered by, um, uh, pirates. Um, another example of this kind of disruption that we're seeing today are during the course of sales call, visual aids can be, uh, engaged and, uh, slides can be created in real time enhancing communication facility, uh, and efficiency.

So everything will be multi-dimensional conversations, and it'll, uh, have these multi modalities about them, and, uh, people's people will be able to interact more freely between these different applications. So this transformation is happening today in a limited way, but it's happening. So smart contracts are, for example, being used, um, uh, in a more intelligent and dynamic way where, uh, new businesses are considering or experimenting with, uh, smart contracts as a way of, uh, automatically adjusting to interest rate changes or automatically adjusting to prices in their markets, like, for example, in the, um, natural gas or, uh, oil markets.

Separately, another application would be a risk assessment application for Defi loans. You can see readily that, uh, Web3 can, uh, AI will drive this into, uh, Web3 and facilitate the, um, the processing of loans and ultimately the, uh, the loans being procured by, uh, potentially more people in a, in a greater geographic area. And, um, but simultaneously, ai, uh, is being used today as well as, uh, increasingly in the future to detect and prevent fraud in defi transactions.

This not only applies to, uh, smart contracts, but it also applies to defi loans and other, uh, vehicles of web. For Web3, uh, transactions, fraud detection will be applied to cyber currencies, and, uh, ultimately, um, what we'll see is much more personalized financial services coming out of, uh, uh, you know, available to people. Um, because of AI underlying defi.

Let me offer a couple of more, uh, uh, examples of what's happening in, uh, Web3. Um, here's an obvious one. Uh, generative AI models like, uh, doll, as well as others are creating unique forms of digital artwork and avatars.

And so in the NFT area, the ability for AI to generate novel images will expand the creative opportunities or, uh, possibilities, um, for NFTs. Increasingly, as time go by goes by, we'll see greater navigation of decentralized apps and, uh, services through chat box, um, assistance, various types of virtual assistance and, um, uh, Web3 applications will just simply be smarter, uh, be more natural to the user, in part because of personalization and much more interactive, much more conversational to help users, um, get the most out of their application, as well as the data that underlines the application recommendations and personalization will be increased multifold, and it, it's already happening today, but the ways of, uh, recommendation engines under Web3 are fairly limited and, um, and more singular. This, the recommendation engines, if you can imagine a travel application that just gives you, uh, greater recommendations based on your own personal data, as well as interactions with your email and interactions with your calendar, and interactions with various other things that you have, um, including bucket lists that will ultimately help in shaping your travel experience by simply organizing it better in the beginning, as well as enhancing it while you are traveling.

Um, this and other applications will show, um, will result in greater decentralized applications. Automated AI systems, which are already, uh, going into place but are, will be, uh, increased substantially in the months and years to come, uh, will be developed around trade, um, optimization of crypto assets and defi applications. Um, algorithms will be much more advanced, but will be more accessible to people in part because of the multi modalities and in part because of the personalization.

And this will help, uh, execute trades, rebalance portfolios, and maximize, uh, yields on your portfolios. Security tools will definitely get better with AI Today, we're seeing a, a major transformation in security tools. Uh, this week it turns out that, uh, the RSA conference is, uh, being held and AI is everywhere at, uh, the RSA conference.

And, um, but with respect to, uh, Web3 specifically, it'll protect networks, protect blockchains, protect, um, smart contracts and users from fraud hacking, and, uh, the various things to come. It'll detect, uh, anomalies such as, um, uh, embedded malware or, um, in the software supply chain, find anomalies and other malware in source code and binaries and, uh, suspicious behavior will be, um, uh, found and, um, in real time. And this will ultimately result in a limit to, uh, risks.

NLP as you've already seen, is gonna be ubiquitous. Um, you know, natural language processing is, uh, something that is built into Siri, for example, and Alexa and various other applications, uh, for customer service. Uh, these techniques are gonna expand and, uh, will become a prominent part of the, uh, blockchain where the human language rather than, uh, complex programming will enable people to, uh, act, uh, actively participate in the blockchain, but also transact, uh, in new ways in the coming years.

Tho those blockchain applications are computationally intensive. Um, um, witness, uh, you know, bit, uh, Bitcoin Mining and Transaction processing and AI will help facilitate this, uh, increasingly, uh, today, but also in the future. So what we see is AI becoming a very, very important part of decentralized applications and services.

Uh, things will just get smarter, more creative, more usable, and, uh, more secure as a result of the evolution of, uh, uh, Web3 and, um, our, our evolution from Web two, a singular type of environment to a, uh, multi-dimensionally interactive environment. In Web3, Web3 opens up new data sources as a result of AI, and, um, will ultimately result in applications which are much more accessible as well as much more powerful. This concludes my presentation.

Let me know if there are any questions.