Kevin Pang - AI Beyond Knowing - Explore Emotion AI with Neura - TOKEN2049 Singapore 2025

Today's session is about AI, but it's not your traditional AI talking about generation art and stuff like that. We're just going to talk about something very different. It's about AI beyond knowing which is explore emotions with neuro AI. So I think before we even start, I want to share about some of the current traction we have some success right. So this Grammy awardw winning R&B icon Neo and uh and obviously more celebrities as well.

He's currently working with us right to create emotional avatars of himself so that they can engage more I think emotionally at least right the film engage with the um his community and his fans. So that's what we are doing and of course uh with that in mind it help us to actually accelerate our web development as well as the mobile app. So I think that's achievement we want to kind of share before we actually start the uh presentation proper. So some key statistics right and I think this is something that everybody probably know as well. Um over 90% people prefer talking to a live person.

Again while this statistics come from the US I think is actually much true to uh other markets as well. I think everyone has experienced nowadays when you actually get onto website for instant exchange website right or going to your utilities website to seek for uh support most of the time a will chat up and say how can I help you and all that right but if you ask people at least more than 90% will prefer talking to a live agent including myself right so most of the time when I actually see the bot pops up I'll just type live chat which means that I want to be transferred to a live agent to chat with me right same thing for Google same thing for any other providers That's human because bots are bots, right? You feel that you're talking to a robot and you feel that the robot couldn't solve your problem despite AI being advanced and all that, right? And of course a lot of times half of the adults find that the customer service chatbots are unfairable because they tend to answer certain questions which you can find already in FAQ right which you can find. So that's what they can do right now.

They can't really solve your problem. They cannot like for instance perform a task for you, activity for you, delay a bay payment for you or make a payment installment kind of plan for you as well. You need a life agent, right? So most and you're looking at you know using them for health uh purposes and all that like for instance you're talking about uh online remote doctor services most of the time in triage where you actually speak to a nurse you would prefer to speak to a live person right versus a board. So which is why only 30% would actually kind of use it.

Probably the G gen Gen Z, right? Gen Z are not very good in talking to adults potentially and so I mean talking to B is something that they're used to but that's probably only 30%. The rest are pretty skeptical to say that hey how how is that tried? She's able to understand my health issues and again right we are always talking about JGPD with uh you know generative AI that's almost 100 million a week. So that's actually up and coming but the problem is that people are not able to use that to engage for human interaction now and of course if you look at purely the trust on AI because purely they know that it's a bot only 40% of the world actually kind of trust it right so why people prefer humans because when you talk to a human right there is something that's lacking in AI right now or bots is actually empathy when you talk about empathy the person emphas uh understands that you are urgent you have urgent need you are angry you are sad you are actually kind of anxious with your needs and all that and try to attend to you.

That's the reason for that. It's not of customer service. Of course, you have bad customer service as well. But majority of the time, you still get your service done. Someone on the other line advertises with you and there is established emotional affinity, right?

Which most of the time when you talk to a bot, it doesn't establish that, right? Of course, when you talk about complex emotional issues, you are almost shouting. you are like you know you really need to get it done and you are you know there's something urgent going on a bot doesn't recognize that it just say sorry can I help you can I help you right but if it's a live agent a human it actually recognize that and actually help you a problem and of course when you talk about it right they feel that there's a trust factor because a human there's someone they can hold accountable to or that you're actually talking to a real human right and of course most importantly assurance right when you talk to a live human agent it will just assure you that hey don't worry Mr. so and so it will be solved I rees escalated to my manager I understand your pain wait for me be patient with me things like that are very assuring to a customer which a current AI bot is unable to kind of establish right so what are what is the real frontier we are looking so we have actually gone through things like basically very much IQ right which is essentially reasoning generation and efficiency we have done things much faster better and much more complex so you actually have like chat GBT 5.0 zero able to do much complex task that previously like maybe even last year is unable to be done.

However, that's purely on their so-cal intelligence quotient level or the IQ level, right? The next step isn't IQ itself is EQ, right? So, long time ago when people was just talking about, you know, IQ being uh, you know, able to gauge the success of a person. However, later on it was debunked that actually it's not IQ, right? It's something called emotional quotient, EQ, right?

That comes upon where some entrepreneurs despite knowing that they don't feel that they are highly educated right they do not look too smart they are not like geniuses and all yet they are very established and very successful in the business versus people who are actually of high IQ right so the thing is that he has established that EQ is a much more important aspect in terms of success and human interaction so emotion AI isn't just a particular feature in AI right it's not just something that can turn on and say it's a module it is foundational. It needs to build at a foundation level to establish it and build it based on emotion AI purely rather than just you know uh uh kind of like technology kind of AI. So where does the AI fall shorts today right it forgets past emotional context it will just remember you as a person you have logged in there's a history of what you have done even with chat GBT right now right it knows what the task you've assigned it to be uh you have assigned it to do yeah you knows the history of the things that created for you but that's about it right there is no emotional context like for instance when you were actually writing I mean just to give quick example right when you actually asking chat GBT to actually help you draft a resignation letter are you happy to resign or are you set to resign it doesn't register that right so if there's a emotional aspect to it then the AI is able to understand the reason why you're actually doing the resonation you know you can actually help you support you better just a very brief example right it struggles with subtext and of course sarcasm and this is actually being different right so for whenever you type something in a text even in current times when you actually post something in social media you wanted to be sarcastic but guess what Most of the audiences think that you're trying to tell it literally, right? So if a human doesn't understand that much less an AI to understand that you're actually being sarcastic, but that can be changed if there's an emotional aspect to it. Of course, it misses the cultural nuances, right?

So I think different culture despite speaking the same language English, they express it much differently due to the culture. So what it means in this culture that seems particular positive, it means the words, but it's actually rather negative elsewhere and invokes negative emotions. So if there is a lack of emotional understanding of the particular uh nuances you know for cultural differences then obviously misunderstanding happens right so that's another thing that's missing uh of course there isn't emotional coherence right so the thing about emotions is not something that you know it just happen at the moment it just goes off which is why people stays angry like for instance you're angry with your boss you're going to be angry with your boss uh you know not just for the next minute and so he's going to be angry with to your boss I mean for the next couple of weeks right or sometime for a longer time period. Same thing as well with your friends and all that. The thing about emotions whether you are angry and happy and all that is sustained.

There's a period of time. It's not just going to happen and disappear decipate very much immediately. Right? So there isn't such emotional coherence across time in existing AI. AI is just AI at this current moment in time.

So this is where we are trying to find this shortfall and weaknesses and try to kind of do something different. Right? So our architecture is built at the foundational level. So it is a problem we identify at the system level which is a foundation level rather than just a feature where you have it on and just a style right say oh I have an AI I just put a wrapper around it is emotional and that's it right no we actually be at the foundational level right so the first thing is longitudinal emotional memory again when we actually look at memories right what does memory invokes for us it actually helps us to retain sorry when emotions right what does it actually help us retain it retains memory good memory or batmar. Like for instance, you would remember a very good meal you have, right?

The taste of the food and you remember the the you know, basically the people that eat with you. You're having a very good time having that meal and you have that emotion of being very happy. You will remember that almost years later, you can still recollect. Same thing, right? If you go to a restaurant and the food is horrible, the service is horrible, you will have that very negative emotion and you will remember that all the time and next time when people mention it, even 3 years later, they say, "Yeah, don't go to that restaurant."

Right? But if there is no emotional context, let's say, for instance, you're having a quick bite. Let's say you get uh, you know, a sandwich, you know, from the fridge over there, have a quick bite and all that. Guess what? Tomorrow, you just forgotten what did I eat for lunch?

Because there's no emotion. There's no something they're saying that, you know, it's tasty and all that. Unless you're just grabbing something and suddenly you bite on it and you didn't expect it to be so tasty. You remember that yesterday you have a tasty sandwich from the bridge. So again it's tight to emotions right so we actually develop based on that.

Next thing is multimodal fusion engine right. So in so basically we have different input systems not just text not just voice we are able to actually base it on you know inputs from cameras inputs from uh you know visuals in order to understand what is the body language what is the emotion you express on the face at that time. So that's something that's very important. It's not just about text, the voice, the nuances in voice, you know, the inputs, the movement of the body, you know, the eye gestures and all that that is all registered in this engine itself. Of course, emotion condition decoding, right?

For instance, as tied into the first one is that certain people, you know, certain culture move certain different ways, right? Like for instance, certain culture they will bow, certain cultures, they move their heads in a certain way and certain cultures doesn't move at all. So that is something that is designed in there. And of course different uh humans of different ages have different nuances as well. So that's actually looking into that.

And of course the cultural adaptation level right how people speak how people talk about it how they actually pause and all that. That's a culture aspect that is actually also being registered in there. And last but not least personality anchors for instance right understanding the human that's interacting has a certain personality is actually being registered and kept as a personality record rather than just say oh this is so and so is asking certain questions. So that's our architecture is based on that. So what are the use cases that we have that have you have impact the first and foremost which I mentioned just now is healthcare right for instance remote doctor remote doctor consultation first you need to definitely go through a triage which is the nurse right that actually understand what's your problem before handing over to the doctor so that's something that can be handled rather than having humans it's difficult to find qualified trained health care professionals uh already in hospitals much less having them in remote uh places as well right so the thing is that that can be handled by AI emotion AI Because in health care in hospital people are in pain they're sick.

If they're not sick if they I mean they're happy that they have nothing right they wouldn't be actually seeking a doctor's consultation for treatment. So they are already having that very negative emotion. So the thing is that there must be emotional affinity to to have at least empathy to understand the problem and able to kind of like you know uh uh engage with the patient. So that's an an actually use case as well right education. So there are certain AI engines right now that actually performs the role of actually like you know performing causes right whether with a AI avatar that you know teaches or lectures particular lessons or you know kind of like uh snacks that go by itself.

However, there's never an emotional aspect where there is a recording of whether you are actually engaged understanding your face. Does it look like that you are totally confused? There's no recognition of that. So, if emotionally AI, it can actually understand that as well. Of course, retail.

So it's actually all based on your mood's recommendation rather than just click click click and then you know it doesn't know it just you know current uh huristics only understand that you are very focused on buying certain uh products but it doesn't understand why you're buying certain products right why you what mood do you have at the moment in time now if you have the emotional aspect to understand the person why they're buying products then it actually can have better recommendation based on mood and that creates a better shopping experience for the the the customer as well last but not least I think everyone has played uh online multiplayer games or single player games before right even at the time you know when you know technology is actually very you know kind of like you know generations behind people still get attached to non-player characters why because there is a emotional attachment even though it's not emotional now imagine if we apply emotions to the non-player character as well right the the NPC can actually interact with you emotionally then you'll feel more engaged with them right so that's another use case so what we do is that how do we learn from society? You know data sets are not enough because they are we actually have real conversation emotional reactions and of course real criticisms all recorded to understand to build this particular emotional engine right and of course the web tree is is itself right so web tree by define by definition we look at the decentralization so this AI is not centralized so you want it to be decentralized right into your mobile app and all that's why we mobile app you can actually use your mobile app to create the emotion AI based on the foundation engine we have and that's why the decentraliz comes in right the utility of the web tree financial services and all that will be decentralized that's the whole idea behind that and again right so in future emotional agents are the fabric of digital life not because you know they they can do things better because that's already done but they actually connect better so that's the future of it the future is definitely decentralized with the power of machine AI so thank you very