The sudden rise of ChatGPT is helping to bring artificial intelligence to the mainstream, especially when it comes to search engines. Vitali Mossounov, Global Technology Analyst at TD Asset Management, discusses some of the big players involved and their battle for search engine dominance.

Transcript

Kim Parlee: Big tech. The big tech firms are dominating the news flow right now. Two big themes have emerged. First, ongoing massive layoffs, and second, AI – artificial intelligence – going mainstream, particularly in the form of ChatGPT. With more on what both of these mean from an investment perspective, we’re joined by Vitali Mossounov. He’s a Global Technology Analyst at TD Asset Management. Great to have you here.

Vitali Mossounov: Good to be back.

Kim Parlee: Let’s start with AI. I think that probably the mainstream, or the– most of the world is hearing about AI, really, for, not the first time, but more often now. Is this the beginning of a renaissance, or has this just been happening for a while?

Vitali Mossounov: It’s been happening behind the scenes, and we haven’t really had any reason to pay attention. But big businesses have been all over this. And so, we’ll talk about it later probably, but Alphabet (GOOG, GOOGL). Take a company of that stature. They’re using AI to optimize the ads that they serve you. They’re using AI to autoplay the next video on YouTube when you’re done watching, so you don’t have to think about what to watch. And they’re using AI even to complete your search results as you start typing.

So far, AI has been a tool of productivity and efficiency for businesses. And I’ll call it improving consumer choice, or maybe steering us. But again, we haven’t had any reason to notice. But now, with DALL-E, and ChatGPT, things are changing. For the first time, we’re noticing it, a killer app in some sense. And that’s the– I wouldn’t call it a renaissance or revolution, I’ll say sea change.

Kim Parlee: OK. Sea change. Well, let’s talk about the sea change. ChatGPT, I mean, people have been quoting this number everywhere, I think it’s used, and this is an old number, by 100 million people after I think two months, which is the fastest-growing, I’ll say, application in a broad sense since– well, since ever. I think that’s the fastest we’ve seen that kind of adoption in anything. Tell us what ChatGPT is, and maybe why it’s so compelling for those who haven’t used it.

Vitali Mossounov: Right. It’s very compelling for those who have used it, certainly, given the number of 100 million. But the term I’m going to put on it will take the air a bit out of the sails, and that’s it’s a chat bot. It’s literally a chat bot. And the purpose of ChatGPT is to mimic natural human speech in response to human queries. It does it quite well, obviously, and it’s very fascinating.

But taking a step back, there’s an organization, a non-for-profit called OpenAI, founded eight years ago or so by Elon Musk, and a number of other people, but that name has kept coming up the last year, Elon Musk. So I thought I’d flag that. Behind the scenes, what you’re really talking about is a very large language model. And that language model, its job is to really predict the sequence of words, right? And that’s what ChatGPT is so good at – understanding what words should come next, stringing it all together, finding context for it, and then making it seem as though you’re talking to another human being on the other side.

And so, it’s been a little while this has been in the works. 2018 was when the first iteration of this type of model came out. This most recent one – 175 billion parameters. In my head, simply equate those to neurons or synapses in your brain, so tremendous size.

The trainers have gone. The researchers have gone. Trained it on the open internet. 300 billion words. We’re talking about 500, 600 gigabytes of data. But not only, I’ll say lastly, have they trained this model on the open internet to read Shakespeare and to explore Wikipedia, but then they’ve used two techniques: supervised learning, and reinforcement learning.

And what they’re doing there is giving the model, giving ChatGPT pre-set, pre-engineered responses, different alternatives, and as ChatGPT is choosing one or the other, they’ve then given feedback and said – no, no, no , that would have been a better response. So they’ve really cranked up the learning of this model and the results are, well, they speak for themselves.

Kim Parlee: It’s incredible to use to people who play around with it. I know my son has been playing with it. I’ve been playing with it. Teachers are terrified of it because they think it’s going to be writing essays now.

Maybe you could just tell us what the impact is going to be. I mean, it is a chat bot, to your point. A very large language model-supported chat bot to do what it’s doing, but maybe just talk a bit about what’s happening between Microsoft (MSFT) and Google too. Because Microsoft, of course, has come in, bought it. They’re going to be, of course, integrating it into what they’re doing. Google’s took a hit because they were trying to do something that competed and it didn’t look as good when they did it.

So, how is this all going to play out? I mean, is this– I mean, just maybe respond in terms of what we’ve seen happen so far.

Vitali Mossounov: That’s a question for ChatGPT.

Kim Parlee: I know. [LAUGHTER]

But say it like a surfer.

Vitali Mossounov: Say it like– [LAUGHTER]

Kim Parlee: Anyway.

Vitali Mossounov: Oh well, the application so far, I would put them in a novelty bucket, just to start with that. We’re certainly finding that developers are using it to code more efficiently. We’re finding that kids are using it to cheat, much to the dismay of teachers. But you can use it to get great suggestions for recipes for a meal in the evening, and then have it lay out the recipe for you. So it’s got use cases. I still put them in the novelty bucket. But they’re novel enough for all of us to be captivated by this, and as you said, for most of us to spend some time playing around with it.

But let’s leave that off to the side and talk about the new wars, the search wars, then. So, Microsoft, an early investor in OpenAI, and therefore ChatGPT, three, maybe four years ago, and no one was really paying attention to this. Of course, Microsoft, all about the cloud. We were talking about other things. Now, they go ahead, and with this most recent iteration of ChatGPT-3, the powerful one, that came out I believe on November 30. And Microsoft made a lot of noise about it. Recently, they had an event, they said we’re integrating this into Bing, our search engine, which nobody uses. And so they explicitly went– and their CEO did in fact, and said we’re coming after search, we’re coming after Google. They’re big margins. They’re over. The party’s over. And so Google, on the back foot, trying to respond, held an event immediately after.

Kim Parlee: It didn’t go so well.

Vitali Mossounov: It didn’t go so well. Kind of a bit of a publicized flop. But this is sort of the background and the essence of what’s happening between the two. Now, I guess, where does this go? To me, it’s a little bit nuanced, because right now, the way that the stocks trade, you’d think that Microsoft is going to eat Google’s lunch, dinner, and every other meal Google’s ever eaten.

I don’t think it’s going to be quite like that. And the first reason is, I would say, consumer behavior and distribution. If you consider the revenues that Google generates off of search, 90% plus come from mobile. And through mobile, you’re either getting it, well, from an iPhone or an Android device. If you think about an Android device, who owns that operating system? That’s Google. And they make sure that Chrome and the Google Search bar are preinstalled on it. Very hard to displace.

And from an Apple (AAPL) perspective– well Google, we don’t know exactly. It’s a closely guarded secret, but it’s rumored that Google is paying, or Alphabet is paying, something in the range of $15-20 billion just so your Safari browser has Google as its default search. And so, the distribution channels for what search you use are very much locked down by Alphabet, at least for now.

And so, the way for Microsoft to win is through PR, and they’re doing their best at that, but it’s to convert that PR and get you to install the Bing app on your phone. And so far, very few people are doing that.

Kim Parlee: It’s interesting because, really, it comes down to so much in business, it’s not always about the better mousetrap. It’s about sales and distribution, which really makes the difference at the end of the day. One thing, I’ll try and fit this in. But a lot of people have said, AI is coming after my job. AI. I’ve heard others say, actually, AI is going to be a huge generator of jobs as we move forward. Where do you stand when you take a look at this on the, I guess, medium to longer term?

Vitali Mossounov: Yeah, I’d say I’m going to have a less popular answer, but I think historically, you have looked back and said, every single technological evolution and revolution has been a job creator. I think this one is different. I think this one is different because it’s the first one that is able now to replicate human abilities and skills, and to an increasing degree. Not quite an awful lot yet as we can see, but I think in the long run as you go 5 and 10 years out, there are a lot of repetitive or routine tasks done in the knowledge economy.

First, that those individuals, I think, will be enhanced by tools such as OpenAI and what everyone else is building. But eventually, I see it as a risk, because these models get bigger and better. And that is going to create a bigger discussion that we’ll probably have a separate MoneyTalk for.

Original Post