E20: AI Agents & The Intelligence Age: Hype vs. Reality

In this episode of the CRM show, hosts Max, Chris, and Rod discuss the latest developments in AI, including insights from Sam Altman's article on the intelligence age, the current state of AI growth in corporate settings, the emergence of AI agents, and Salesforce's position in the AI landscape. The conversation highlights the balance between AI advancements and the role of humans, the challenges of AI adoption, and the potential for AI to augment jobs rather than replace them.

Takeaways

• AI is evolving rapidly, but the role of humans remains crucial.
• Sam Altman predicts we are close to achieving general intelligence.
• Corporate adoption of AI is still in its early stages.
• AI agents represent the next generation of digital assistants.
• Salesforce is positioning itself as a key player in AI.
• Usage-based pricing models could disrupt the SaaS industry.
• Model collapse is a significant challenge for AI agents.
• AI will assist in various tasks but won't replace humans entirely.
• The future of work will involve a collaboration between AI and humans.
• AI adoption requires overcoming barriers related to trust and accuracy.

Episode Transcript

Introduction

Max: Welcome back to the next episode of the CRM Show. This is a show where we talk weekly about all the latest AI news. Joining me today are my co-hosts, Chris and Rod.

Chris: Hello, everyone.

Rod: Hi everyone.

Max: Great. So welcome back. How are you guys feeling today?

Chris: Definitely less tired than you must feel, Max.

Rod: Definitely, yes. I was at a finance IT event yesterday focused on AI called FinJS. As a result, my night was a bit too short for Halloween.

Max: That's interesting. I mean, with AI, everything kind of runs 24/7 anyway. So I guess your nights and days will only get shorter with AI. Today, I think we have a jam-packed agenda. There are a few things we want to talk about:

1. Sam Altman's piece on the intelligence age
2. The state of AI growth report by Iconic Growth
3. AI assistants and the concept of agents
4. Salesforce's recent Dreamforce convention and their AI revelations

So, without further ado, let's dive into our first topic: The Intelligence Age by Sam Altman.

The Intelligence Age by Sam Altman

Max: Recently, Sam Altman published an article discussing how he sees AI developing. A few key points stood out:

1. After many years of development, deep learning work is finally bearing fruit.
2. He believes we're about 18 months away from general intelligence.

Chris, I'd like to get your view on the article in general and on this idea that general intelligence is not far away. Yesterday, I attended a panel where people were saying it depends on how you define intelligence, and it's not as easy to build a general list for everybody. I'd love to hear your thoughts.

Chris: Thank you so much, Max, for sharing this article. I found it very inspirational, upbeat, aspirational, and optimistic. I think Sam has a point about how 10 or 100 years ago, we couldn't envision where we are today. He's really depicting a future of AI in our pockets, which isn't a new idea. We've all thought about having AI at our fingertips, making presentations with one click of a button and completing different chores with just a swift little prompt. I think we're getting towards that.

However, when it comes to really getting these complex workflows done, we're a little bit farther away.

Max: What about you, Rod? What do you think?

Rod: I'll take the contrarian view. First, of course, Altman is not just some unbiased observer outside; he's interested in this vision coming true. He's heading OpenAI and has many investments in the space. Of course, if he says the revolution will fail, that doesn't look good for his legacy.

But this is also the opposite perspective of others, such as Peter Thiel, who for more than a decade has maintained the thesis that we've remained stagnant in progress. Nothing has changed besides a little bit of bits, and so on. The big things in humanity remain more or less the same since the 70s, maybe since the 60s.

One quote from Altman's article that gave me a lot to think about was that we will have things that our grandparents would have imagined as magic. I was thinking, yeah, but that happens pretty much all the time, right? Let's think about our grandparents' grandparents. These are individuals who were born likely in the mid-19th century. At that time, there wasn't really electricity. The most advanced thing was the telegraph. Of course, for them, in the early 20th century, when their grandchildren were born, things such as the telephone and electricity everywhere, at homes and so on, would have seemed like complete magic.

I have the impression that this happens across generations. This thing of progress seems sometimes very slow. When we look from a macro perspective, it's actually very fast what things are going on. So here, on one side, yes, Altman, of course, wants this to become true. But at the same time, I'm thinking maybe, in fact, for our grandparents' grandparents, the change was much more drastic than for our grandparents now.

Chris: Max, I think you just opened a battlefield.

Max: That's the whole point of the show.

Chris: Yeah, Rod, that is very true. And yet at the same time, I think what you're saying is, I mean, lots of things would have impressed our grandparents, true. And I think AI, nonetheless, is something that we all can look forward to. In terms of adoption, I think Sam also makes that point in his article about how it's going to be an incremental slow adoption.

I think back to probably 2007 when the first iPhone appeared. I think there were articles saying, "My God, nobody really wants to have a buttonless or keyboard-less phone. Typing is so difficult," et cetera. Everyone was bashing it. And yet where do we stand today? We're at iPhone 16 now. It's not special anymore to buy an iPhone because it's so normal and standardized. There are so many other competitors and products out there to have a smartphone now. And most of them come with a keyboard-less screen. That adoption also took a bit of time to come. And I actually feel very optimistic about AI for that reason.

Rod: Yes, I mean, I'll show what happened is happening and so on. But something that his article was missing, and also in general, I miss a lot from these very pro-AI views, is: where is the role of humans in all this? He says something along the lines of, our children will be learning from AI tutors that will be able to teach them everything in all possible languages at any time and perfectly and so on. I think, okay, so where does this leave humans?

Historically, people have said, okay, we will move towards more and more creative jobs. And all of a sudden, pretty much everyone will become, say, a content creator, a YouTuber, an influencer, a coach. And this is how it will be. Here's something along the lines also of, so we will be doing things that we don't consider jobs currently. And I'm thinking, okay, but what does that mean? Because we're saying we'll move towards creative tasks, but we're seeing that AI is becoming really good at creative tasks.

Also, this idea of everyone becoming an influencer - I'm not so sure if you look at this week's Meta conference where they showcased detailed clones for influencers. The idea is that if you receive a lot of direct messages as an influencer, you can just put an AI that will answer your fans, and they will not really know that it's an AI they're talking to. And this is also with a 3D or a model view of your face.

So also the idea of swapping the creative human task, like being a personality that people like and pay for, maybe long-term that will not happen anyway. So I wonder, in all this vision, where are humans left?

Max: Well, I guess I'm the human that's left out, to be honest. I think yesterday when I was at a panel, I was listening to a very similar argument around humans. So ultimately, humans steal from humans today. But in the future, as you mentioned Rod, a lot of interactions that we have today with humans might get replaced with AI. And then the next question I have in my head is, what's the difference between you talking to a person and you talking to an AI? To a certain extent, if intelligence is really there, if I were to call up and make an order, I might not know if the other person is either an AI or a human, because it could be an AI that was talking to me in a more general tone, in a more general sense.

So I guess a lot of routine tasks can and will be replaced. However, when you think about truly novel stuff, as well as truly critical decision-making, I don't think we're there yet. This is where I think when Sam mentioned that in 18 months we'll get to that, I am quite curious what that will be like. Because I don't think my decision-making skills are extremely good, but I'm actually looking forward to what that AI would be like. Will it be able to add on to what I lack, perhaps? So I'm quite pro and excited about it, I must say.

Chris, you were trying to say something?

Chris: Yes, I think Rod, your argument or the argument that's always in the room, this pink elephant of what happens to humans when AI takes over, I think it's quite overdramatic and definitely lacks a bit of that time component to it. What do I mean by that? I think at the end of the day, human beings will evolve along with the technology that is currently present and is getting adopted. We've seen this with a lot of other technologies, from the automotive car replacing the horse to the computer replacing secretaries. But I don't feel like we have less work, and I don't feel like we have more unemployment.

So I believe it's a bit of a two-dimensional view to say, okay, humans will be redundant. I'm sure humans will make sure that they will stay relevant. I think that's sort of the first view on it. I think a second one, and we need to cut it because I forgot it, is that even if what Rod said was right, that it's doomed time for humans and we all become redundant, even if that was true, I think the question is, can you stop this progress? And I think history actually shows us that human curiosity is something that is very, very difficult to stop.

I think it's almost like, maybe it's almost too late if you want to take that pessimistic perspective. It's already unleashed. Lots of people feel excited about this new technology, the possibility of deep learning. And there will be people going through with it. And I think, yes, you can resist it and you can try to contain it, but there are many, many more people that are willing to go and explore. And I think that's really what, in my opinion, also makes up humanity and human beings.

To your question, Max, what's going to be the next frontier when it comes to our work? Perhaps it's the idea generation. Perhaps it is the time where we actually come up with original human ideas. And that might be the next frontier.

Max: Cool. I think, just to recap, from an intelligence age perspective, I think we're all excited about it, but there are also risks and opportunities that we need to think about. And humans are always going to be in the loop, at least for now. That's my biggest takeaway.

The State of AI Growth

Max: If I want to shift gears a little bit, let's talk about the second thing, which is the state of AI growth. Iconic recently did a survey on all the C-suite executives around 200-plus companies, trying to understand where and how they're adopting AI. Just a headline: what's interesting for me was that in there, there was a chart where, not surprisingly, 91% of the companies they surveyed are in the US, 8% is in the UK, and then 1% is somewhere else. I don't know if that really corresponds to what's actually happening in the world from an AI development perspective, or if it's just because that's their network.

The second highlight that I find interesting is that at the moment, because it's so nascent, a lot of larger corporates are investing in both infrastructure and applications at the same time. It's almost like they need to build their own picks and shovels in order to develop the right thing for the organization. So that to me is quite interesting. It feels like we're still at the nascent stage.

At some point, I think Rod and Chris, we talked about this previously, which is around how infrastructure kind of just gets collapsed and then applications just flourish. I'm keen to see when that will happen. I think just a highlight on that is on the application side. They were talking about deployment more for internal use cases today, especially from a large corporate perspective. And a lot of it comes down to HR, customer service, trying to just make things easier, engineering and operations internally.

I guess to that, maybe I'll turn to you, Chris, just to get a bit of your thoughts. Is that what you're seeing in that adoption perspective? We talked about adoption earlier. And then the barriers to entry are still very much on, I guess, more on the performance of the LLM as per the report.

Chris: Max, first of all, thanks so much for pointing out a caveat to this report, which is very US-centric, I would say. And I think this is something for the audience also to keep in mind. And I think the other thing is that I think the report also had six predictions for the future across three themes. And to me, it really feels like a good summary of what we've been discussing actually in our show, right?

They talk about the rise of specialized use cases. So we spoke about startups really focusing on one specific case, whether it's customer service chatbots, for example, as a very specialized use case versus building up yet another foundational model, which I think is also the second theme that the report is quoting as being really hard. I mean, just in last week's episode, we spoke about how expensive it is to actually build these models up and that you need 100 million in order to even start building another foundation model that probably isn't as well trained as the existing models. So it's really become a battle of the big companies and the big guns that actually do have that kind of funding.

And then yet at the same time, we're also talking about, the article talks about exponential growth and sort of like where this is coming from and lots of it comes actually from enterprises utilizing maybe available foundational models, but in a proprietary way. What does that mean? It means to use it and enrich these models with proprietary enterprise data to really build an edge. And I think that's also something that we've seen so far when we talk to bigger companies and how they actually leverage AI.

Max: How about Europe?

Rod: I want to connect this report to what we were just discussing now on the pace of progress. So we can talk about how fast or how slow this is happening. But what we're seeing here in the report, but also what I can observe in the industry, is that things move much slower. So here in the report, one of the problems that is criticized or reported is that there is a lack of expertise in companies. So people cannot implement this.

For example, now that I was yesterday at this event, this is my second year at this finance event for IT experts. And last year, it was all about saying, "Yeah, we started exploring ChatGPT and doing some prompts. We still need to get approval and so on." So they were like one year later than everyone else in terms of general adoption. This year, it was all about showcasing copilots, pretty much ChatGPT-like interfaces that are embedded. For example, if you're an investment banker, then you can chat directly on Microsoft Teams and ask, "Hey, can you please give me the latest stock price for Walmart?" And then this copilot will pop up and say, "Right now this is showing like a chart or something."

And in general, this means that the finance industry, in this case, is at least one year behind. So what is the current progress of state-of-the-art in AI? Then also when I talk to companies and so on, the type of use cases that they're having, the type of use cases that they're showing, it's also very early, it's still very explorative.

Another thing from the report that I also found interesting, again, shows how there's always dissonance between what we believe in and maybe what is happening on the ground, is who are the vendors, what is being adopted. Where often we talk about alternative newcomers, new cool models and so on, but for the enterprise and for this specific subset of companies being reported, which is primarily US-based, pretty much everyone is using OpenAI, followed by Google, the Google platform, the Google models, and so on, which might come as a surprise, right? Because we keep saying, and also often here in the show, we say, "Hey, Google is so behind. What's happening with them, and so on." But on an enterprise level, because Google is already so trusted, so many people, so many companies have already implemented the Google Cloud in the infrastructure, in their environment, this is what they're adopting.

Max: I find that fascinating. I think the Google point definitely is something that I guess I've been thinking about. About a decade ago, we kind of saw Google as just a consumer brand. I remember there was even an article saying that Google was not able to sell into enterprise. But it feels like they completely turned it around at the moment, especially on this development on AI. I think it kind of just shows you the amount of investments they have gone into from the infrastructure perspective. Super interesting.

And then I guess one thing that I get out of it is the application ROI haven't quite reached the level that we would like yet. But that doesn't mean that it's not rapidly developing. And Rod, you mentioned financial services. I can be pretty sure, I recently spoke to a couple of CIOs and COOs of the largest banks you can think of. And even for them, it takes time to even approve those use cases, and also takes time for them to implement. And because of that, there's a massive lag between what