On the Front Lines of AI - A Special EmTech MIT Preview, 2024
It's emerging. Do you hear it? Do you feel it? Technically. The trustworthy journalism required to make technology a greater force for good is here. It's always been, and it has barely begun. Hello, and welcome, and thank you for joining us today for this special preview session for our flagship event, M -TECH MIT. I'm Niall Firth, I'm MIT Technology Review's Executive Editor, based here in London. And with me today on the call is my colleague James O 'Donnell, our AI hardware reporter based in Boston. Hi James, how you doing? Good, good, thanks. Thank you for having me.
So what is EMTECH MIT, you might be asking. If you don't know, it's for about more than 20 years. We've been running it as our main event for place for discussions on emerging trends and technologies really before they hit the side guys. So we have our experts from all different places, businesses, talking about the issues, breakthroughs, and applications for technology in their business. The next one is going to be held at the MIT Media Lab on September 13th, October 1st, two consecutive days.
And the agenda is going to be focused on two of the biggest topics in tech right now. That's AI, obviously, surprise, surprise, and climate tech. James and I are going to be covering a lot of the AI sessions over those two days. So we thought it'd be fun to give you a little
teaser of some of the biggest topics, themes, tell you about a couple of the speakers we've got lined up, just to give you a feel for what it's going to be like on the days. So we're calling this session front lines of change because what it really means is we're going to be introducing you to some industry leaders who are already working through some of the gnarliest issues, problems, challenges that probably many of you who will be attending will be grappling with in the coming months and years. Do you give a chance to pick up some tips, tactics, maybe learn some new approaches for your own business or your life? So before we start today's chat, remember, if you've got any questions, throw them in the chat window.
I'll try to get to as many as I can during the call. Right. Let's get into it. James, hello. What are you thinking about ahead of the event? What's on your mind and what are you excited about coming up? Yeah, sure thing. I think top of mind for me as we head into M tech, and this year in general is definitely trust. As we think about AI, it's obviously an election year, we've already seen a couple cases where, you know, the trustability of AI has really trustworthiness of AI has really been brought into question and sort of, you know, entered the political sphere. So you had images of Taylor Swift fans embracing
and supporting Donald Trump, which Donald Trump embraced on X, or other platforms. And it turned out that many, if not all of those images were generated by AI. You also had Donald Trump making accusations about crowd size, images of crowd size being AI generated for for Kamala Harris. So even if you put the election examples aside, you also have the issue of AI overviews, which Google has launched this year, and it's sort of totally transformed how we get information through Google search. AI is helping to sort of filter out information, and sometimes it does a good job of that, and it delivers answers very efficiently. And then as you've probably seen other
times, it delivers bad information, misinformation, things like that. So the trustworthiness of AI, I think is a really top theme for us this year. And I'm excited that we're going to be diving into that at EmTech. One speaker I'm particularly excited about, her name is Yasmin Green. She heads up Jigsaw, which is a
unit within Google that's sort of, you know, in a way tasked to solve this problem. So Jigsaw does a lot of work trying to make the internet a more safe place, a more trustworthy place. So she's going to be discussing some interesting new research with us and some of her work in general. So she's kind of the perfect guest for us to sort of, you know, debrief how AI is affecting our trust in our political system, in the web, things like that.
So really excited to have her guide us through that. And that's cool. That's very cool. I know something you cover a lot is quantum computing. You're keeping an eye on what's happening in that space. And we've got companies out there at the event that you're going to be chatting to. Tell us a bit about that.
Yeah, quantum is in an interesting space right now. I'm sure a lot of you have probably heard that quantum computing is promising to do really great things for us. And what it's promising to do is basically compute things faster than ever before. So that could be really helpful with things like drug discovery, but also really helpful with things like training AI models more efficiently. Obviously, it's a big
question right now of how much computing time, but also energy and emissions are being spent for computing power related to AI models. So quantum computing is promising to change a lot of that. The problem is that it's also still in its infancy. So people have been working on it for decades. And when I talk to experts in the quantum computing realm, they say, that's just how the process is going to be. We'll get there eventually. They say it's inevitable that we'll start to shift towards that way of computing. But it's kind of a slog to get there, really expensive to research. And a lot of companies are
trying to do this in incremental stages. They're building really small quantum computers, which is good to demonstrate the technology. But it's actually not very useful right now. They're kind of too small to do anything useful, certainly not something as complicated as discovering a new drug or antibiotic or things like that, which requires millions of computations.
So one company I'm excited to chat with at EmTech, they're called CyQuantum. They're taking a bit of a different approach. So instead of trying to build incremental small computers, they actually kind of have this moonshot vision that, you know, we're not going to really build one until we can build it large enough that it can be useful. So they're testing out different components and
chips to sort of get there. And they are actually working in the state of Illinois to build sort of the largest quantum supercomputer campus right now. They have their goal set of building that sometime in the next 10 years. They're also working in Australia, I believe, with a lot of quantum computing experiments there. So
that'll be an interesting lesson in sort of what it's like to toil away at a problem that's many years away from coming to fruition, but kind of requires that incremental step to get there. So they will be there and really excited to chat with them. Very cool. I just want to say I've heard that there's some problems with my connection on the feed. So really, sorry about that. If I'm looking kind of weird and glitchy. You know, we are working on it to see what happens with that. And tell us to give us one more. Give us one more that you're looking at the event,
apart from competing. Sure. So it's not strictly AI related. But in a way it is, I suppose. So we're going to be talking to a company called Synchron, which if you haven't heard of them, they are sort of a competitor to Neuralink. So they make brain interfaces for people with all sorts of mobility issues or disabilities. So
that can be, you know, something like Parkinson's all the way up to paralysis, things like that. So they help people sort of interact with their environment or with screens, with computers, if they are not physically able to. And we're going to be chatting with them. It's a really interesting field. It's obviously also kind of in its infancy and very heavily
regulated. It takes a lot to get these sort of devices even to the trial stage. So from my understanding Synchron right now has about six patients with devices that compares to Neuralink, I think either has one or two, I have to have to check on that. But, you know, in its very early stages of getting FDA approval to actually get these patients, these devices implanted in patients. One thing that's sort of unique about Synchron is it requires supposedly a less invasive procedure than Neuralink. So Neuralink is a full on implant, whereas Synchron actually goes into a vein near the brain understanding.
And it can be implanted a little bit less invasively with some of the same characteristics. So I'm excited to talk to their CEO, Ricky Banerjee, about what it's like to sort of develop these another type of moonshot technology that's set to have some real impact for people. So that's one side of the conversation. But something else I'm excited to chat with her about is what it's like to share a space with Elon Musk, who is obviously, you know, very dominant in the media cycle right now. And so it'll be interesting to hear her thoughts on sort of the ups and downs of having someone like that bring so much attention to the field to this technology that people are still sort of getting to know, but also what it's like to have him as a major competitor. So that'll be a really
interesting conversation. Yeah, yeah, I mean that's something I wonder, I've had many people in the public can't name any other rocket companies apart from SpaceX. If you're competing in the space, someone with that massive social presence, that's huge PR presence, like how do you get anything? Yeah.
Done. Okay, let me tell you about what I'm excited about. I've still got a few under wraps for now, but I'll tell you about a few that I'm really looking forward to. One of the topics, marketing, generative AI and marketing, is this something that is going on right now, but not quite sure in what forms. I'm speaking to Rebecca Sykes, who's from the BrandTech group, and they have lots of sub -brands beneath them, and they've got various different ways they're using generative AI. People, creators, are using to create their own ad campaigns. It comes up with copy images, the whole thing, all in one go, and it propagates out, and she's probably going to come out there. One
of the dangers being if you don't really know who you are and what you stand for, and you're suddenly propagating a generative AI campaign based on that, it's going to come out really woolly. There's one thing I was like, and that is the idea of having generative AI models being a new audience for your brand. So the idea being that you are focusing your multi -AI models think about your brand. It sounds like weird. It sounds like something doesn't really make sense intuitively,
but actually it thinks so much now as proxies for other things. If you're using Gemini via search, it's based on the large language model, Llama from Meta via social. The idea is get your brand and see if it comes back to the things that you think of that are important for your brand. If you make an energy drink and you really want the brand to be associated with your expertise, feel it into various language models, and that's not what comes back. They come up with techniques to try and then tweak it by maybe doing a series of YouTube videos that can then feed into the model to try and get your brand ranking for that kind of stuff. So it's kind of like a new version of stuff.
I'm really looking forward to chatting to her about that. Then what else have I got? I've also got, I'm speaking to someone from FedEx. We'll be talking about how it's a supply chain, that kind of thing, how they're sort of doing the last mile problem with autonomous robots.
In your reporting, I'm looking at how sharing to AI is sort of super charging how we train robots to do things. That's right, isn't it? Yeah. So we're in a really interesting time for robotics. I think for a long time, people felt like, you know, we've been seeing videos of crazy robots for years, but they haven't really, you know, made an impact, especially in the home, you know, most of us probably, if we have a Roomba for a long time, that's probably been the most complicated robot that we interact with on a daily basis. That is, unless you live in one of the cities with robotaxis, which has obviously gotten a lot more common this year, San Francisco, cities in Texas, there's some other pilot cities, but for the most part, robots were kind of slow to develop. And then AI started to really change, especially
in the last year or so how roboticists were able to train robots. So instead of coding them to do one specific task in a laboratory, you know, move this tennis ball over here, or pick this box up and put it on this palette, instead of having to sort of manually code all of those tasks, people were getting really good at just feeding in data of example tasks and having the robot sort of learn how to do those tasks themselves. So if you were to, it's called, one method of doing that is called teleoperation, where you basically have a human manipulate the robot to, let's say, you know, wash something off of a dish using a sponge, you maybe coach the robot how to wash 80 dishes. And if you coach the robot how to wash 80 dishes, that might be all it takes for them to be able to wash any dish that looks vaguely similar to the ones that you taught them on. So just like large language models that can be fed,
you know, a corpus of text, and then be learned, be trained how to sort of generate sentences and predict the next word in a given sentence. You can pretty much do the same for robots, it turns out, where you can train them with existing data. And so that over the last year or so has made some really explosive impacts in robotics, not so much in home robotics still, I mean, that's that's a field that's sort of coming up. But we're still a ways away from having robots sort of do menial housework for us. But in terms of commercial robots, people may have seen lots of videos from figure AI and Elon Musk's Optimus robots and various different humanoid companies that are getting, you know, many billions of dollars of investment right now. And a lot of that investment in humanoids is happening because AI is making it easier for us to train these robots. So we're putting together some really great sessions for M tech to sort of
walk people through how those changes are probably going to happen, you know, what jobs are going to be affected first, what sort of industries are going to be affected first. But I can safely say that this year is, you know, things are changing by the month I can't, it's hard for me to even keep up with all the robot news that's coming out. Yeah, it's good, it keeps you busy. I've got a question from Shantanu about whether there be an issue with intellectual property rights, with AI and tech advancement. And actually,
that is something we write about a lot, something we cover a lot. One of the people I'm going to be speaking to, really excited about, is a company called Story Protocol. And they are like absolutely red hot right now. That if anyone who's on the call will have seen over the last week, they just raised absolute ton more money in a Series B round, valuing them something north of $2 billion now. So, but they're basically a blockchain company for, to protect the IP of creators. So, and I've got my hands up here as a massive disclaimer, having sort of knocking around tech reporting for about 20 years. I've been through a lot of blockchain stuff and I've like definitely edged towards the skeptical side of things, waiting for the blockchain to become the thing that's going to replace something and it never seems to happen.
But something's been happening in the last, last couple of years called smart contracts. And we've covered that a little bit at MIT Technology Review. These are basically contracts that are embedded in the blockchain that execute when certain conditions are met. I think that's when it has starting to get a lot more interesting. So basically, sort of protocol has got this, this system where they embed smart contracts in and you put your IP in there, whether it's like, you know, your images, if you're an artist or if you're a publisher or your work, and you set it up a smart contract that says, people can use it for these different purposes if they pay this amount of money. So people can license it to remix it,
republish it or for general generative AI models to train on it. They can then pay directly using the smart contract without any lawyers or lawsuits or middlemen involved. And it kind of thing is meant to happen all seamlessly. And it's part of this,
it's part of this bigger topic that we've sort of, we've written about a lot about sort of the sort of fight back between creators and the big tech companies who are kind of using these balls to suck up all of this, all this artistic work. And I think it's kind of an interesting solution to the problem. It comes out from a different way that we've done it in the past. We've covered it a lot from like the artists side, you know, a thing called glaze where you can put it on your work to stop people scraping it or Nightshade, our colleague Melissa has covered, which is where you sort of poison your work. So it completely messes up a model that tries to train on it. This is like a different version of that. It's like, all right,
I think they're going to take it anyway. So like, how can I, how can I monetize that? So I'm kind of interested in speaking to their co -founder, going to dig into what that means. Yeah, that can I add one more just a little bit to that? It's interesting you bring that up because I had an interview yesterday with a VP from Amazon and we were talking a lot about You know the issue of explainability how much can you get a model to actually tell you what it's doing under the hood? And that is actually really important. We
were talking about with intellectual property stuff because If you could imagine Let's say an ai music generator, which we've we've covered before um right now All of those generators are being sued because they're basically producing work That sounds a heck of a lot like the beach boys or beyonce or or whatever, you know, take your pick of artists And the artists are saying that's sort of copyright infringement if you're able to generate this thing using ai That sounds exactly like my style. And so what I was talking with this engineer about yesterday was is it possible? You know in the future to have models where the model can actually tell you Hey, this song is actually five percent Beyonce and ten percent the beach boys and that would probably be a terrible song but it could actually cite, you know What artists it's it's getting for its inspiration and then that could lay the groundwork for some sort of compensation system. Let's say Where if you know that your work was used in the generation of some sort of output whether it's a song or a poem or a book We're getting to a point where you can actually explain that and there might be you know Technologically possible to have some sort of Compensation system behind that so i'm happy you're you're going to be exploring something related to that at mtech Yeah, yeah, I think, yeah, I think it's a really sort of cool, sort of coming from the opposite end of trying to fight basically the inevitable, just seeing, you know, how can you make money out of it. I think I've probably got time to tell you one more talk I'm going to be doing. I'm
going to going to speak to Slack's new CEO, and that's going to be really fascinating. That's on the end of the second day, it's going to be something like to build up to and, you know, Slack quite recently bought by Salesforce. I think she's the third CEO in like a little over a year or so. So there's a lot of change on Slack that I want to find out about, but also they're really rolling out loads of generative AI stuff for all kinds of things.
So like at MIT technology view and probably lots of people watching and saying new Slack all day long for all of our interactions at work. And I know like I go on vacation, I come back, I have like this like obsessive compulsion to read every channel that I miss to see like, you know, what stories I miss, what, you know, what chats, what banter, what are the jokes that I didn't, wasn't, you know, and, you know, I know you're not meant to do that, but it's like, it's hard. There's so much to catch up on. And I was also trying to find like, I know I spoke to Jane O 'Donnell about something like this about a week ago and trying to like search through it. So they've been like rolling out these generative AI add -ons basically, you know, things that do like summarize discussions that you've had with people or like a more sort of turbocharged search for kind of certain topics, or can just like summarize a channel while you're away. This is the kind of the general vibe that you missed out on. But then, you know, there's, there's lots, lots to get into there, and especially about like how you sort of roll out AI across a massive company and make it work for people and also be cognizant of people's privacy and their direct messages to each other.
It's not being sucked up into the model. So I think there's lots and lots of sort of things to play with and talk about around that sort of workplace AI in particular. Yeah, I think we're in an interesting time for how much appetite we have for AI systems doing work for us in terms of how much privacy we want to sacrifice. And I guess it just makes me think of back when, you know, photo aggregators like the photos apps got started to get really good at recognizing faces and organizing your photos into, you know, this vacation versus that one. And at first, we may have been a little skeeved out by these companies having access to our photos. But if they deliver a good enough product, you start to kind of, you know,
give them a little bit more leeway. And I work with something like Slack, you know, depending on how they roll these out, we'll have a sort of similar pattern where at first people are afraid to give Slack our messages. And then over time, if they do a good job with it, you start to sort of lose that fear. I'm not sure. Right, right. OK, well, I'll tell you what, I mean, that's kind of summary of some of the stuff we're going to be talking about. I know, you know, we've got other people who are also going to be at the event. And we're going to be doing a lot of very
cool stories between now and then. I expect everyone to be to be reading out there. But just a few other highlights of stuff that's going to be there. There's an amazing session coming up that's going to have Amy Webb, the futurist in it, who's going to be talking about an idea of like a tech super cycle. And how you sort of navigate that as a business or a business leader, which is very cool. What I'm really looking forward to, definitely going to be like every seat packed in the house. Ray Kurzweil is going to be there and he's a very famous tech sort of writer and futurist as well. I suppose he talks a lot about the singularity,
the idea of when humans and machines are going to be melded together. He was one of the very first to be talking about AGI, artificial general intelligence, the very concept of which is like pretty controversial and contested. And it's on again, so we've written a lot about. So I can't wait to hear what he's going to say. That's going to be total fireworks, I imagine. And then also at the event, we're going to be unveiling for the second year, running our
climate tech list of 15 companies to watch. So basically, this is companies that we have picked carefully from around the world that we think have a chance of substantially cutting emissions or counteracting the effects of climate change in some way. Some really cool names on there. Some people will know, some that they definitely won't know, and that's going to be unveiled live on stage. And my colleague Amy Nordrum is
going to tell you more about that, tell you, James, but also you, the audience, more about that alongside our editor -in -chief. And in a special LinkedIn live that we're holding next month, it's called AI climate and the new rules of business. So she'll be talking more about AI and the climate side of things, giving you a bit more of a teaser of what else is coming up on our agenda. So that's coming up next month. So make sure you're signed up for that. But I'm afraid that is all we've got time for and our special producer. Thank you, James, for joining us. And thank you everyone watching
there for coming along for the ride. If you want to know more about the event, go visit us at mtechmit .com, that's E -M -T -E -C -H -M -I -T .com. And keep an eye out for your emails, because you're going to get a special email with a special discount offer for the event as a thank you for coming along to see me and James chat.
So keep an eye out for that. Thank you for joining us and hopefully see you all at Pentec MIT. Bye.
2024-09-01 11:01
