Join Merlyn Mind co-founder and CEO Dr. Satya Nitta as he discusses the history of artificial intelligence, major milestones and luminaries in the field, the latest advances, what AI can and can't do, the phenomenon of ChatGPT, and Merlyn – the AI-powered assistant for education.
This talk was delivered in January 2023 at FETC in New Orleans, LA. The excerpt below was edited for length and clarity.
Machines can't really think. They can't really reason. They can't really understand where a student is, and this is something teachers do effortlessly. When a student's struggling, we reach for metaphors and allegories. We understand whether the student is paying attention or not. So one of the very last things in my opinion that will ever be automated is the ability for a machine to teach.
If you go back to the original goals of [American cognitive and computer scientist] Marvin Minsky and company, they wanted machines to teach. Basically that's not something that will happen until we actually build truly intelligent machines, and I'm not sure that will happen in my lifetime. I'm not going to say never, but we don't know how to build really intelligent machines because we don't really understand how intelligence works in the human brain, which is the only example we have of actual biological intelligence.
All the excitement today is about generative AI. People are giving ChatGPT prompts or DALL-E prompts and it's generating images and it's very exciting, but a subset of people are worried – is art or people's ability to create art obsolete?
But the really interesting thing that's happening is – understand and converse in natural language and perform actions in the real world. And that's actually pretty interesting. So if you start saying, look I can now talk to a robot and say, ‘Open the can of tomato soup,’ it'll understand what I said to it. It couldn't before, but it will now. Or if you say to a computer, ‘Open my Google Drive and search for this presentation,’ it couldn't before, but it understands it now.
So this whole understanding of language and performing actions in the real world – translating language to actions – is really where I am particularly focused, me and my company are very focused, and we think this will have huge implications for how we interact with computing. And that's kind of what the Merlyn Mind company is all about.
So if you look at where we were and where we are now, we had to learn complex computer commands in the past. Computers now understand our language. We were overwhelmed by technology. Now our favorite tools are seamlessly connected with AI. And we used to waste time on managing all this technology. We’re now focused on what's important.
So that brings me to what we do at Merlyn Mind. One of the problems we focused on when we first started this company is that, we looked at teachers today – and this is pre-Covid and post-Covid this is even more true – teachers are spending a large portion of their time juggling technology, applications, connecting things to HDMI switches, plugging in their display, trying to problem solve, and they're going away from the most important human task of teaching students and giving them motivation. So we felt this was a great example of a problem that was scoped, that was contained, but that could still really make something very useful and help people to do their job far better.
So what we did was we built Merlyn, the world's first digital assistant purpose-built for education
So what will happen over the next next few years is, first of all, AI won't displace people, but people using AI will displace people who are not using AI. So people who embrace AI and repurpose it to do some very useful things for them in education, both in the classroom with the kinds of things we're doing with Merlyn or outside the classroom, will be the people who'll succeed in this industry. Because this disruption’s upon us. This is as big of a disruption as anything we've seen in computing in the last 20, 30, 40 years, and definitely the biggest advance in AI that we have seen at least in my time in the field.
So what is likely to happen with with the latest advances? Things like task automation, like navigating all the technology in class, opening your presentations, answering questions, switching to HDMI 2, taking attendance, all these things will be things that AI will basically do – it's already doing, that's what we're doing at Merlyn.
But the next few things are on the horizon. This is exactly what's going to happen.
• AI tools will generate assessments. We are amongst at least a dozen other teams who are working on trying to create question generation as a tool to help teachers.
• AI tools will grade. So today you know if it's multiple choice, largely it's computer graded, but if it's some written free response by a student, it isn’t. But we are getting to a point where an AI tool could grade and could give a teacher some summary of what it thinks a student’s essay is and why it thinks the essay is actually on point or not, and the teacher can review what it did.
• Lesson plan creation. We keep hearing about this over and over again. Teachers come and complain that this is a huge amount of time. They take great pride in it, but it's filled with friction today. So I envision a time in the not-too-distant future where you'll basically work with an AI, create lessons that are grade-specific, content-specific, with the kinds of images that you want or the types of videos that you want in there.
• Certainly I think if you repurpose ChatGPT or DALL-E, all these things, writing can improve. It doesn't have to be a substitute for you to write. You can actually write with it because the way I like to think of it is, if I'm trying to compose something, I can compose it using what I know, or I can try to compose something using what I know and being informed by all the knowledge in the world that's at my fingertips now. That's the way I look at AI, right, which is, we're now standing on the shoulders of all the knowledge humanity produced that's digested, given to you in the moment and for you to build on top of, and that's pretty powerful. I think I'll become a better writer because of that, and humanity will progress and we will all start writing better, more precisely. We're not all going to be writing like each other because there's still huge amounts of room for human creativity. It's not something I would worry about here. I just look at this as yet another useful tool and another automation in the long march of automation.
• And things like summarization of articles. I read a piece of news today, I want to go show it to my students. Can this thing summarize and give me something that's grade-specific? So as you saw from the examples earlier in the talk, ChatGPT explained who Shakespeare is for a 10th-grader and for a kindergartner. So these kinds of things will also happen.
So finally when I look at what education is, there isn't a better quote from Yeats where he says it's not the filling of a pale but a lighting of a fire. So at the end of the day all these tools are here to take the mundane away from the people who do the job they do so brilliantly, all of you teachers out there – which is you fill your students with motivation, imagination, purpose … you're igniting their passions, you're making a difference in their lives. And that's what these tools will allow you to do better.