Dharmesh Shah is the co-founder and Chief Technology Officer or CTO of HubSpot, a leading marketing, sales, and customer service software platform. Born in India, it was during his MBA course at the MIT Sloan School of Management that he met Brian Halligan, with whom he founded HubSpot in 2006. In addition to his role at HubSpot, Dharmesh has supported a number of technology startups and writes about entrepreneurship and technology. We spoke with him on a break from his recent annual convention, held in Boston, about the upcoming trends in all-pervasive Artificial Intelligence (AI). “We can now accelerate marketing and sales with AI, for example, by creating fully personalized content,” he explains.
Q. How can AI help companies be more productive and improve sales?
A. We will start with sales. Broadly speaking, the first thing I think AI can solve is the automation of tedious tasks. We have been with this automation process for some time and now we have reached a higher level. For example, today’s software allows you to write a blog post, something that was not possible before. It can provide a first draft and can also capture your tone of voice, your communication style and create videos to improve sales. AI already understands the context of the company and of a conversation, the needs of a specific customer and can help the agent who is speaking with them to offer them what they really need, other related products or services…
Q. A recent HubSpot report concluded that when customers have a problem, they want to be assisted by agents, not chatbots.
A. Chatbots have evolved a lot and for multiple functions. For example, they are capable of providing information about all the company’s products and in a very natural or human way. But we don’t want to fool customers by saying they’re talking to a human. We don’t do that, we don’t advocate that. But it is true that customers want to obtain information at any time and about any product or service and have competent assistance at all times. And this is being driven and enabled by AI. The same thing happens with emails. If I receive valuable information for me, that offers me what I am looking for, I don’t care if it was written by a person or an AI application. I’m not looking to make a new friend. I just want you to answer my question. I want to solve my problem and I think this is where AI is going to be able to help a lot.
Q. One of the next trends in marketing is going to be moving from specific applications to AI agents that solve almost everything by themselves. What is that transition going to be like?
A. We are very much at the beginning of that evolution. Now, the way we use most applications, especially chat applications, is step by step, first do a Google search, then another step… What the agents are going to allow us to do is carry out all those multi-process processes. steps in a single procedure. For example, if we have to launch a marketing campaign, that agent will understand if it is for a new launch, if it is a specific event… and will provide everything that we must do and also provide the materials that we will need at each step: from a note from the press to specific articles, materials and content to feed social networks… And, in the meantime, humans will be able to dedicate themselves to solving the most complex issues of the process.
Q. Are you surprised by the development that AI is having?
A. Yes, it surprises me, because I’ve been working in technology for over 30 years and I actually tried to build like a very complete chatbot 5 or 6 years ago, but it was frustrating for clients, it didn’t work. Very soon after, with the arrival of OpenAI’s ChatGPT, we could see what AI was going to be capable of. In the course of humanity, it is one of the most transformative inventions that humans have ever made. In reality, it is the first time that we have machines or tools that are really capable of developing a form of thinking, a form of creativity.
Q. Is it good that machines replicate our way of thinking?
A. We call it artificial intelligence, but I think it is wrong, that it is an inaccurate term, that it is not the correct word. A better term would be to call it alien intelligence because by calling it artificial intelligence we seem to assume that we want it to think like we do. Maybe we don’t need you to think exactly like us. If we, with our way of thinking, have not been able to solve health or geopolitical problems, we still need another different model that does not have the limitations that humans have shown to have since the beginning of time.
Q. Is that why some AI decisions or answers to our questions surprise us?
A. Sure. AI thinks very differently from us and that is why it sometimes confuses us. We got confused and wondered why he would solve the problem we posed to him that way. That’s because he’s not like us. He is capable of much better reasoning. And that will allow us to cross branches of different sciences. That was unthinkable before. Even scientists are taking note of it… They say this could advance humanity. In terms of solving problems that historically we have not been able to solve.
Q. AI, specifically ChatGPT, reached one million and also one billion users in record time, long before any social network… Does that popularization help?
A. Yes, it’s incredible. And I think the reason for that is that ChatGPT caused an experience similar to the one we had the first time we talked on the phone or sat in front of a computer. We get immediate results and we don’t have to learn anything. You express yourself in your own language, whatever it is, and it gives you an answer. It may not be the right or perfect answer, but it is a surprising answer. And that’s what has made adoption at that level so high and so fast. Until now, when we wanted to do something, we had to understand the software, perform several steps to edit a photo for example. You touch here, swipe here, click here… The AI allows us to simply tell it what we want to do and it does it for us.
Q. What will be the next AI trend?
A. I think the next thing will be that the models are going to be even more intelligent. Hundreds of millions of dollars are needed to develop these new models and that limits innovation to some extent. There are actually a handful of companies that are big technology companies, that have that capacity. Until now, there have been a handful of models that are super powerful, costing a billion dollars. Maybe we don’t need an AI that knows Shakespeare, Newtonian physics, and all of biology. I just need an AI that fits on my phone, that can help me do the things I do in my daily life. And that might actually be more valuable than the big models. I think we’re going to see models of different sizes and we’re also going to see this type of agent model that allows us to do multiple steps in one step. For example, I will tell him that I want to go on vacation to Italy and he will find the best options according to my calendar, the cities I already know, my tastes…
Q. You already know what Stephen Hawking said about AI…
A. That it would probably be the worst and the best thing that happened to humanity.
Q. Yes. What do you think?
A. I can argue both sides of that debate. I understand where the anxiety and fear comes from and I think it is well placed. I think I know what we have to do. The big tech companies that are producing the AI models need to have watchdogs. One thing that I think helps a lot is that in addition to business models for private companies, we have open source models. The open source models are now at approximately the same level of capability. The open source model we know exactly how it works and what it is doing and what it is or is not capable of doing. We can evolve it as a community. We have to have at least enough confidence in our own system for people to use our models over the alternatives that are out there. So I think it raises the bar and makes things safer.
Q. What role should regulation assume?
A. I personally do not believe in initiatives by individual countries trying to put restrictions on AI. It’s like we need to slow it down. Let’s put it on pause for a year so we can all catch up on what’s going on here. And I understand the desire to do that. But I don’t think it’s practical, because, we have global organizations for it. So you as an individual country may decide that you are going to pause AI, but the reality is that this is a global movement. And the rest of the world will continue to move forward. You may not like it, but it puts you a little further behind from an innovation perspective. So I don’t think it’s a good move.
Q. What would be the solution?
A. I think the best move is to say: let’s understand what the risks are. We are going to regulate the most important aspects. We’re going to offer financial incentives to providers of these critical models to ensure we’re building something secure.
Q. Do you see global regulation for AI as feasible, something that is also constantly changing?
A. It is not easy to regulate something in progress, totally in progress. Yes, I don’t know if there is any limit that we have to set. I’m looking for precedents regarding issues like this in the past. In our history it has been very difficult to try to find the right balance. How much regulation, how to enforce it? I don’t think we’re going to have easy answers for this. My hope is that this is a bit circular, that one day AI will actually be able to help us solve some of these problems as well. Well, there should be some kind of World Constitution around AI or technology. Although maybe, I don’t know, I’m being too optimistic.
Q. What worries you the most as a parent about this development of AI?
A. Well, I’m a little worried. I have a 13 year old son. We all grew up with different forms of technology, right? We had the computer, the calculator or video games before this. But we were able to take advantage of that technology and learned mathematical calculation even though calculators existed, right? The challenge I face now as a parent is my son asking me why he has to learn to write, when there is a GPT Chat that can write for me and will do a much better job.
Q. And what is the answer?
A. Then I have to explain to you that this is not really about writing or not, but about thinking. Writing forces you to think and express yourself. It’s like learning to play a musical instrument. We can listen to music and learn to play an instrument. It is very useful and fun to enjoy both things, to enjoy what has already been created and the process of composition, interpretation… You can enjoy these things. Even if a computer can do them, why wouldn’t we enjoy it? I want to have the experience of writing a poem or drawing… whatever your art form is. I struggle with that. It is like taking away Humanity’s desire for creativity and expression. Especially for the upcoming generation.
Q. Is there no real risk of reducing that human creativity?
A. Yes, that is one of the risks of AI. But, there are also benefits. AI gives you a new brush, gives you a new way of expression that you didn’t have before. For example, my son really likes manga and wants to put together a story, but he doesn’t have the drawing skills to be able to do it. He has the ideas. Get to know the characters and the story you want to tell. He doesn’t have the physical skills yet, he hasn’t developed them. And now thanks to AI, you can produce something that would not otherwise be possible to produce. It’s like I’m writing a book or something and I want to make an illustration. Now I can ask the AI to do it and it does a decent job much better than I could do, so it’s beneficial. We have to do it in moderation. Like when our parents warned us not to play video games too much…
#DharmeshShah #HubSpot #dont
