Sam Liang, Otter.ai CEO: AI captures everything

0:00This BBC podcast is supported by ads outside the UK.

0:29CFO. LinkedIn has a word for that. Bull spend. Now you can invest in what looks good to your CFO.

0:38Sam Liang:LinkedIn ads generates the highest ROAS of all major ad networks. You'll reach the right buyers because you can target by company, industry, job title, and more. So cut the bull spend. Advertise on LinkedIn, the network that works for you. Spend$250 on your first campaign on LinkedIn ads and get a 250 credit for the next one. Just go to linkedin.com slash broadcast. That's linkedin.com slash broadcast. Terms and conditions apply. Do you ever feel like you're drinking from a firehouse? Paycor's intelligent HR solution empowers leaders to turn down the pressure. Their unified platform includes payroll, talent management, compliance software, and a lot more, connecting you to the people, data and expertise you need to drive long-term business results.

1:30Visit paycor.com slash leaders and go from workflow to workflow. That's paycor.com slash leaders. Hi, I'm Zoe Kleinman, the BBC's technology editor, and this is the interview from the BBC World Service, the best conversations coming out of the BBC, people shaping our world from all over the world. If you're not a little bit afraid, then you're not paying attention. We have never seen a people so united. Do not make that boat crossing. Do not make that journey. Being born in America, feeling American, having people treat me like I'm not. We're more popular than populism. For this interview, I met Sam Liang, chief executive and co-founder of artificial intelligence transcription startup Otter AI in our London studios.

2:22Sam Liang was born in China and moved to the US in 1991. He received a PhD from Stanford University before joining Google, where he led the search engine's location services. He co-founded the California-based Otter AI in 2016. The startup has evolved from a voice-to-text transcription service to offer AI-powered recordings of live events, meeting summaries and content searches. He tells us why he thinks we won't be typing as much anymore and how avatars could soon take our place in meetings. He also explains why he wishes Otter had been around when he arrived in America in the early 90s.

3:02Sam Liang:You know, when I first went to America, my English was so bad. People didn't understand me. I couldn't understand other people. I wish I had order at that time to help me. At that time, I literally carried a Sony Walkman with me in all the classes I went to. Did you? I had to record it because I couldn't understand the professor in real time. I had to record it and I had to listen to it several times to fully understand the lecture. If I had order at that time, I could have studied much better.

3:44welcome to the interview from the bbc world service with sam liang

3:50Sam Liang:we built order to capture all the voice data first then use ai to transcribe it use ai to analyze it and also use ai to aggregate all this intelligence so that we can extract intelligence, extract insights out of it. So it's not just a meeting note taker for a single conversation. More importantly, we're building a platform to create this conversational knowledge engine to organize thousands of meetings, connect the knowledge in thousands of meetings. So how is AI shaping who gets heard? Is it changing the voices that we hear and the voices, if what you're saying is correct, the voices that we will remember and the voices that we understand?

4:45Is it curating that? And will people lose out as a result?

4:50Sam Liang:The power of AI is that it's able to capture everything. It's able to try to interpret everyone objectively. Human beings are imperfect in terms of their capability to listen and understand. Everyone unconsciously, when they listen, they don't hear everything. based on their past experience, they may misinterpret certain things. They may unconsciously ignore certain things. So AI actually can fix that. What about accents and underrepresented languages? You know, there are 22 official languages in India, and most of the world's largest chatbots can only manage about half of them at the moment. Yeah, accent.

5:43Sam Liang:I'm not a native English speaker myself. I was born in China. I grew up in Beijing. I went to America in 1991. Even after 35 years living in America, I couldn't get rid of my accent. So when we started Otter, we built our own speech recognition technology ourselves. We actually stress test our system by speaking to our own speech recognition engine to make sure our engine can recognize our own accent. Of course, we also collect tons of speech data from all over the world to train the engine, to stress test it, so that it's able to recognize all kinds of accent, American accent, British accent.

6:40Sam Liang:In America, there are so many immigrants from all over the world. Everyone speaks with different accent. Because you have a responsibility, don't you? If something is transcribed wrongly because the tool hasn't heard it properly, because it doesn't understand the accent, then you are at risk of misquoting somebody, aren't you? And essentially sharing misinformation. Exactly. It's a very challenging task to understand people's speeches, especially with an accent. In addition, if you're just speaking about common topics, most AI can handle it relatively well. But oftentimes there are challenging words such as acronyms, jargons, unusual names, especially immigrants' names.

7:32Sam Liang:They have unusual spelling. How do we recognize them correctly? They're all challenging. So we build this system that constantly learn new words, constantly learn new acronyms and jargons. And there are cultural nuances as well, aren't there? Some things mean different things to different cultures. Yeah, even the spelling is different between UK and America. So we have to handle that. We get it right, obviously. Yeah. Of course, those are the word-by-word transcription. It's a lot of challenges we are handling pretty well. But more challenging is about understanding the meaning. So on top of the word-by-word transcript, we use AI to try to understand what people mean.

8:20You must have a treasure trove of data of thousands, millions of billions of conversations. What biases have you had to correct within that data that you've used to train your tools?

8:33Sam Liang:That's a good question. So far, we focus on one type of data. It's the business meetings. We focus on the enterprise intelligence. Of course, every business is different. I wouldn't say they're biased. But let's say, for example, we know that the majority of senior leadership teams tend to be men, don't they? Because that is an existing bias in society. And so your tool is trained on business meetings, senior leadership teams, it's going to be men. Does that lead to assumptions by the tool then that women's voices don't carry as much gravitas or they're not as important or they're not as visible?

9:13Sam Liang:That is a problem. That is a problem. Because as you said, most business leaders are men. So there are not enough meeting data where women spoke in. It is a problem. AI actually can help correct some of the problems or detect some of the problems. Because we have a special technology that can detect when a man speaks or when a woman speaks. We can actually measure the speaking time, whether the men interrupt the woman often. I bet they do. Or whether the men dominate the session by speaking too much. Maybe the men speak 90 % of the time, only leave 10 % of the time to the woman who speaks. So that's the first step to detect problems.

10:06Sam Liang:and also can remind people when that situation happens to help people be more mindful. I think that's a first step. The training data problem, it takes some time to fix, but we can start addressing that problem by detecting the problem and remind people of that. You grew up in China, in Beijing, and you came to the US in the early 90s. What do you think about the current tensions between the US and China? Do you think it's helping or hindering the progress of the technology? Would it be quicker if everyone was working together? Of course, different countries have different political system. I moved to America because I want to get freedom.

10:52Sam Liang:I like that freedom. But of course, America has its own problem. Whether you're a Democrat or Republican, every party had their problem. Of course, between countries, a lot of conflict happens due to misunderstanding, due to different beliefs. So this is why communication is so important. Between China and America, they speak different languages. They have different cultures. They don't always understand each other. So I hope AI can help people understand each other better. When they understand each other better, they can resolve more problems. It will take time. And a lot of things is lost over translation.

11:38So is that partly why you are doing this? Is there a personal reason?

11:42Sam Liang:It is. It is a problem that hasn't been solved yet. I hope AI can help solve that. It will still take time. When you first came to the US, how easy did you find it to communicate? It's hard. You know, when I first went to America, my English was so bad. People didn't understand me. I couldn't understand other people. I wish I had others at that time to help me. At that time, I literally carried a Sony Walkman with me in all the classes I went to. Did you? I had to record it because I couldn't understand the professor in real time. I had to record it and I had to listen to it several times to fully understand the lecture.

12:30Sam Liang:If I had allowed it at that time, I could have studied much better. I feel like this explains a lot about where you've got to now and why you're so passionate about this. That's part of the reason, because I needed to like this to understand other people. And I think probably millions of other people need this too. Did you feel isolated? Yes. You know, when I couldn't understand other people and I couldn't convince other people, I do feel isolated. I still have trouble communicating. You know, that's a problem I want to solve. And I hope other AI can help me solve that too. You're listening to The Interview from the BBC World Service.

13:26Do you ever feel like you're drinking from a firehouse? PayCore's intelligent HR solution empowers leaders to turn down the pressure. Their unified platform includes payroll, talent management, compliance software, and a lot more, connecting you to the people, data, and expertise you need to drive long-term business results. Visit paycore.com slash leaders and go from workflow to workflow. That's paycord.com slash leaders.

14:21Average New Policyholder Data for Accident and Illness Plans Pets Age 0-10. You know how sometimes you find yourself replaying a conversation in your head and thinking about how you could have said things differently? Sam Liang does that for real. The BBC was recording our chat, but so was he. He said he'd be listening back to it on his phone later and asking Otter to give him some feedback on how well he answered my questions. I wonder what it had to say about me. He also arrived in his running shoes. He was getting ready for the London Marathon shortly afterwards. OK, let's return to my conversation with Sam Liang.

14:59Where are we drawing the line here between a useful AI tool that's recording and tracking stuff to be helpful and surveillance?

15:08Sam Liang:We're definitely against surveillance. Big Brother by the government. I think we should definitely block that. But for business communication, we see the bigger problem in most of the enterprises is that information are fragmented. Most people don't have access to all the information they need to do their job. There is a class action lawsuit going on, isn't there, in California, which accuses your AI transcription bot of recording confidential conversations without consent. So there are people that are worried about it. We totally understand people worry about it. For that particular lawsuit, we don't think that has merit for many reasons.

15:55Sam Liang:First of all, all the audit users consent to the terms of service we set up. Other users need to get consent from other speakers in the room or on a video conferencing platform to agree before they use Otter. So it's under the control of that particular user. It's similar to the way people use, for example, Sony Walkman in the past. It's a cassette recorder, right? Sony is not responsible for how you use it. The owner of that cassette recorder has the responsibility. I suppose what's different now is that that cassette was not likely to be broadcast anywhere, was it? Whereas things that are recorded now and transcribed can be shared hugely widely on the Internet.

16:50Sam Liang:The sharing is controlled by that user as well. By default, it's only accessible by that one user. That user can say because there's 10 people in this meeting, it's in their best interest to allow everyone to have access to that meeting notes. Because why do we meet in the first place for the 10 people to communicate with each other, to share knowledge with each other? But if people forget what they discussed, they actually waste that time, which is actually very expensive because human time is very expensive. So for you, the onus is on the user to be responsible with Otter. Absolutely. We provide the tool.

17:39Sam Liang:We provide a platform, but the user needs to get consent and to also manage who have access to that. That's kind of the social media approach, isn't it? It's the same as document. When you write a document, the writer controls who can have access to it. I'm interested in the goal, the long-term goal of Otter. Is real-time transcription, translation, both of those things, is that what you're aiming for? It is a real-time transcription, but the information can be accessed in both real-time and post-meeting. For large enterprise that with 10 ,000 people, 20 ,000 people, there may be millions of meetings that are happening every year.

18:26Sam Liang:The enterprise actually invest a lot of money to pay people to go to meetings. They need to make sure the time spent in meetings generate good return in terms of a business value. In five to 10 years time, maybe you and I won't be meeting at all. It will be our digital twins or our AI clones that are meeting on our behalf. Where does that leave Otter? There are several aspects. One is actually we are also building technologies to create a digital twin for every person. Eventually, you can send your digital twin or your avatar, which has your knowledge, to attend some meetings on your behalf. Have you got one?

19:09Sam Liang:For example, you are sick, you're on vacation, you're double booked, you could send your avatar. Okay, have you got one? In our team, we build a SAMS avatar already. It's not perfect yet, but it can answer a lot of questions on my behalf. Do you like it? I like it. It's trained based on thousands of meetings I spoke in in the past. It also used a lot of documents I wrote in the past. So it knows how I would respond to a lot of questions already. We know that Mark Zuckerberg has one too, or is working on one for talking to staff at Meta. Is that what you use the SAM avatar for as well? Yeah, that's what we're building as well.

19:58Sam Liang:We also try to make it constantly learning everything new. The avatar I built two months ago uses all the historical data, right? But two months later, there's something new. So that avatar needs to continuously learn new things. When I started out, I used to transcribe stuff by handwriting it down. I would listen back to interviews and write them down. And then I started typing them into Word documents. And now we are increasingly using AI tools to do the transcription for us and then to go back and check, you know, that the transcription is accurate. I'm slightly worried that I'm not really using my hands anymore.

20:38I don't handwrite. I don't type. What are we going to do with our hands?

20:44Sam Liang:Of course, you can use hands for many things. You can play tennis. You can play violin. You can cook. Badly, but yes. So our prediction is that we'll increasingly rely on voice to interact with AI. Because voice is the most natural way for us to communicate. Typing is actually a skill we learned in later part of our life. When a baby started to communicate, the baby first used voice to communicate for many years before he or she started learning writing or typing, right? So writing and typing are harder skills to learn and to perfect. Unless you're a professional writer, most people are pretty bad at writing.

21:40Sam Liang:They're also afraid of writing, but talking is much easier to do. So do you think that this is the beginning of the end of writing and typing? I mean, might we forget how to do it? It's possible. It's the same. If you think about it, I drive a Tesla for many years now. I use the full self-driving mode. It can take me from my home garage to the garage at my office by itself. so my driving skills has been degrading have they you've noticed that yeah it's already happening does it bother you i hate driving manually you don't like driving anymore i don't like driving anymore because i i'm afraid that i may cause an accident so do you think the tesla is better at driving than you are exactly it's already driving better than me because it's trained by millions of miles of driving data.

22:40Sam Liang:The car has eight cameras compared to two eyes I have. But you don't need eight cameras. Your two eyes are doing the job. You can have 10 driving lessons and get your driving license. You don't need all of that. Eventually, you don't need that. So it's by the same way, AI can help you write way better. AI is a way better writer than 95 % of the human beings generate. But what happens to our brains if we're not training our brains in the way that we do when we learn to write and type, when we learn to drive? Is there a worry that we just won't be as sharp as we have had to be? Not necessarily. When you speak, you are still using your brain, right?

23:25Sam Liang:You're expressing yourself. You're thinking aloud, especially when you are discussing with your colleague. There could be five or ten people. There's a lot of interaction happening. Ideas can be created by talking. Of course, if you like to write, you can still do it, but you don't have to do the tedious part, such as taking meeting notes. Do you think Otter will still be going in 10, 20 years, or do you think it will be part of something else? I think so. You know, we're announcing a new platform called Conversational Knowledge Engine, which we think will capture as many conversations as possible, organize that in a platform so people can access that knowledge so that they can do their job better.

24:15Sam Liang:and also enables agents to help you with some of your workflows. So that's a huge market and it will take another 10, 20 years to grow. People may stop writing, but people will never stop talking. Will they ever stop having meetings? No, they won't stop having meetings. People will still talk all the time. They need to communicate. They need to build human relationship. so if you think about that how many conversations are happening in the world it's just so huge

24:54Thank you for listening to the interview you'll find more in-depth conversations on the interview wherever you get your BBC podcasts including episodes with Karim Bagheer boss of Africa's biggest AI firm the former Prime Minister of Australia Julia Gillard and the musical icon Ringo Starr Until the next time, bye for now.

25:42Amana, and Criterion. Save big money on Maytag appliances at Menards. And check out our weekly flyer on Menards.com for all the great deals happening now.

25:52Sam Liang:Save big money at Menards. May is Maytag month, and Menards is passing the savings on to you. You need reliable and energy-efficient appliances to conquer the day. Menards is your one-stop shop with appliances in stock and ready to take home today. Check out our top appliance brands, including KitchenAid, Maytag, Whirlpool, Amana, and Criterion. Save big money on Maytag appliances at Menards. And check out our weekly flyer on Menards.com for all the great deals happening now. Save big money at Menards.