How Sensay is building AI replicas to manage email and business

0:01I'm Ashton Addison from the Crypto Coin Show and today on blockchain interviews we have back with us Dan Thompson CEO of Sensei here to talk about AI digital replicas and everything else in the artificial intelligence World Dan welcome back to the show it is a pleasure to have you here oh it's always great to be on this show thanks for having me again very welcome AI is

0:22moving faster than ever I'm hoping eventually it will do all my work for me I'm excited to get your industry Insider insights into to the world of AI and what you and your team are working on at Sensei I'd love to start off our conversation with a recap for those who haven't seen our previous interviews and I know there's a lot of updates as well on what exactly Sensei is focused

0:44on and then we can dive into all things AI yeah well I mean that kind of touches on all the points that I guess we're going to talk about but the you know hopefully the next interview will be my replica coming through maybe we're not that far away from that so Sensei literally creates that exactly that we're building digital immortality and creating digital replicas that can be

1:07used in contemporary use cases for businesses individuals as as a tool for Knowledge Management within organizations as well as automations and as essentially as a perfect personal assistant for yourself so imagine having a version of you that can pre-draft your emails or M basically you know have your entire knowledge based from your digital

1:34footprint from your previous messages from the documents that you work from and having something that combined with a sufficiently advanced AI replies in your or is able to reply or at least draft replies depending on your preference to those different to different questions that we get asked every day you know we're online for a certain hours a day and there's

1:56therefore a lot of time where we're asleep or maybe enjoying some personal time and our replicas will essentially enable us to or our teams and people around us to not be stuck and not not be limited to the availability of our own time so essentially they act as an extension of ourselves that can share our knowledge experiences skills with

2:19people globally 247 in any language and it's it's a fascinating to us it's literally building a version of Science Fiction where we can build these perfect digital replicas of people that can act on their behalf for them wow no it sounds great to be able to cut some of those hours of work and have your digital replica Do It but how so how if somebody was interested in

2:43starting on this working with the AI you know the AI has to learn information about the person and the job that they're doing to start doing it effectively so maybe you could walk through how that would work if I was like Hey I want someone to take over Crypto Coin Show how do I get teach this AI to learn about being me and the work that I'm doing yeah so it's always kind

3:08of backwards the way people imagine things to be more difficult or easier the easiest things are things like your voice or your image right so it's it's actually very quick and easy to train the image and voice there's a few great you know projects out there and companies out there developing the technology for that so you can go and quickly record very very little of your

3:29voice really like 30 seconds is enough and maybe two to three minutes of high quality video content to train the actual sort of imagery and voice to be similar to you obviously the higher quality the longer the better but it's it's relatively simple the trickier part is really you know searching for the information that you want it to have

3:53and filtering through your own life to understand which parts of you that you want to have replicated and who we are is probably the most fascinating part of what we're building and how we build it because the subject of identity is and are in existence is you know been contemplated for Millennia but when it comes to actually replicating that

4:16you know you've got a version of yourself that you think you are you've got a version of yourself that you put out there on social media another version of yourself maybe for work and then you have this subjective approach to who you are that is reflected in the people around us so you may be you know one way to your closest Family you may be a completely different person to a

4:36colleague or someone you're interviewing you might be a completely different person to you know the a random person you meet on the street just because you've had I don't know a bad day each of these interactions with these people are you know one version of yourself that they perceive you as and those might be completely different to your work Persona or your your social

4:57media Persona so it's it's really complicated to try and take all that into account and understand the nuances and the languages we use when presenting ourselves in certain formats on certain platforms and that's that's what we try and do we try and help to you capture the relevant information for the relevant use case for the replicas so that they become useful and

5:21you know actually represent the person they're meant to be a copy of mhm now that's really interesting because I feel like it's you're really taking it to the next level because when I first think of okay I need an AI to help me with my inbox I think okay well you know I'm just typing a reply to an email they're not listening to me they can't see me

5:42you know it's I feel like it's easier to get away with it so to say that hey this AI can do it and they can't really tell that the AI is responding but when I actually add in my own voice and video it's really taking it to the next next level of like hey you could you know I could send voice messages or hop on a call with you and it's not even me

6:01it sounds like you know otherworldly yeah I mean there is still the young Kenny Valley issued itself which is that you know people get a bit creeped out by when they're the fact they're interacting with you know non-human Robotics and AI which is definitely true and one of the things we've come to realize when creating this technology is that you know people

6:25will as as much of a wow factor is the video and the voice as people do kind of recognize that it is you know a AI doing it for on behalf of the person sometimes it's okay CU it's kind of you know you get the information you need but a lot of the time people just prefer the text messages and we're also realizing that people there's an uncomfortable aspect of if someone has

6:47their replica acting for them on kind of autopilot like like I do a lot of the time not everyone's quite as Brave to let an AI just do their do their replies for them automatically so we're working in kind of human oversight feature to it that will allow essentially the a bit like having your emails drafted rather than automatically sent for you

7:09so that you can just check them it should it will save everyone a huge amount of time by having the you know drafted for them every morning rather than having to type them out and process them all yourself manually but it's definitely a lot a lot more comfortable for both the person sending and the receiver knowing that the human essentially you know approved the

7:29message code going out a bit like having a personal assistant who can send mail you can do things for you and can use your email address to send things for you but you know with your kind of approval first just makes it feel a bit better which is weird because you know people do trust Pas all the time they you know it's an entire industry in certain countries so there's you know

7:50there is a you know a precedent for this that people have used a lot you know secretaries and PAs to do all their mail and emails and messages and calls on their behalf historically but then suddenly as soon as it's done by a robot which arguably is more accurate suddenly there's something weird about it and so it's kind of understanding where that practicality

8:12versus Nuance sort of preferences come in that's a great point and I feel like there will be a transition period where it feels Eerie at first and then it's normal because they realize the advantages of you know it's it's like having a self checkout machine it's never going to you know give you the wrong chance and after a few years you're like of

8:32course we prefer the automated version over the PA who who could make mistakes and be subjective and on that email front I feel like a lot of salespeople at least you know they're sort of typing the same thing to each client like trying to close the sale is very repetitive you know it could definitely be more efficient and I see huge advantages for AI there on the sales

8:56side now I'm sure that there's many many applications in all of the different Industries sales being a great one what are some of the business areas that Sensei is looking at as obvious loow hanging fruit first use case applications to automate with AI these different business uses yeah and want to touch on the automatic checkout service I mean

9:23that's a great example of you know this kind of technology that does get there and automates things and I remember when I first saw that it was yeah you don't really trust it so much but now yeah it's so normal you just you can't you know you almost don't want to go to a person because it's just inefficient MH and someone else made a similar comparison the other day like saying how

9:41weird it was to first see people with mobile phones and how easy and common that is now and because of the convenience of it and that's what we see this as as well the convenience of having a literally Mini U the digital youu that can go out and do stuff for you is going to be huge so yeah in terms of Industries and initial applications for it we see two two major

10:02ones the first one is for any company that deals in Knowledge Management and actually deals knowledge in general right so you're looking at Consulting you're looking at legal firms education Healthcare pretty much a lot of different industries that have a heavy requirement on an individual sort of knowledge understanding of a certain topic when that happens obviously

10:26there's a huge brain drain when people move around from organization to organization when they retire when they're just offline they go on holiday the maternity leave they sleep so having replicas of high performing high quality people enabling them to have to share that knowledge even when they're not available having the ability to share that knowledge far further and wider

10:51than their physical and mental capabilities can so taking I don't know let's say one of the leading professor of the world in a particular subject in MIT and enabling them to teach their subjects to anyone globally on a very intimate basis you know into places that knowledge is completely scarce and you know unavailable and the dream of even going to a college like

11:15that would be impossible suddenly becomes very very accessible for everyone you could look at you know High performing experts the consulting firms and lots of other big you know companies insurance companies or whoever pay to come and visit their company to give speeches they pay tens of thousands of dollars just for the just for the speaker having a replica of that

11:36person means that not only do they have the speech for the half an hour hour hour and a half but they have this kind of interactive replica they could that people can directly ask questions to and actually get more out of that knowledge and ask the questions they want to so replicating high performing key people with really high quality

11:56replicas is the sort of main one of the main use case for replication technology and one of the most applicable Services the knowledge within you most organizations is pretty scarce fleeting can disappear and usually relatively inaccessible so replicas kind of solve a lot of that you then can look at the second major use case which is a very scalable

12:20essentially message automator think autofill but on like crazy steroids you've got you know you take a bit of your personality doesn't require too much background knowledge usually your internal Communications from the organization some access to the documents that it should be able to reference and then you can actually have this kind of automator that basically

12:43works as a personalized autofill that would draft out messages for you based on the previous conversations help you writing summaries booking things into your calendar you know basically all the sort of usual day-to-day admin that we all have to go through to get anything done your you know rep will be able to do pretty much most most of it obviously

13:02with some human human oversight if required otherwise you know people like me would be very happy to have it fully automate their life and sit back and while it goes to do the interviews you know no that it does sound like a dream and those industries that you mentioned that's huge obviously Consulting the healthcare industry especially in America you know they're

13:26spending double the defense budget if we can automate some that with AI make it more efficient and I feel like the big four accounting firms you know accounting and Consulting they do both that's probably millions of millions and millions of dollars they could actually put more in their pockets if they could lower the human capital side and make more Automation and

13:51and speaking of that I'm curious on your thoughts on you know the good and the bad of AI and part of it is obviously it's automating so many tasks making it more efficient and on the other side maybe there's an argument that there's less humans that need to do work and then there's people out of a job or maybe jobs just shift into a new kind of Paradigm where they manage

14:16the AI do you think that this kind of technology is going to replace people Al together or just kind of shift their jobs I think there'll definitely be a shift more than a replace you we see this as as an empowerment tool as an extension of ourselves certain industries are more at risk than others from being replaced and certain jobs will inevitably be lost and I guess

14:41shifted into different jobs I think the main use case for replicas as a whole and AI in general are that you know they will enhance our lives in a lot of ways freeing us up to do a lot of things we want to do it will be difficult to start there'll be a lot of much FAS moving shifts in the industry and technology which will disrupt a lot of things and create a lot

15:05of inequality to begin with but if you do believe in the good of AI and good of people then inevitably AI does fix a lot of other problems that we have around the world as well that essentially mean that work almost becomes could be almost become irrelevant there may become a global basic income and living standard that we could achieve using AI yeah there obviously a lot of

15:28conspiracy and a negative approach to that I try and St on the positive positive side I think that there's always risks in any kind of new technology but I think the AI will do a lot of good for us there may obviously a lot of disruption it's a big change you know but same with any any kind of historical Industrial Revolution it's you know there's there's a lot

15:52of people who are afraid of the change a lot of people fight back against it and inevitably the change in the convenience and the e and the productivity of people who embrace the new technology inevitably win out so I think it's it's a lot like that I think you know some of the biggest concerns run AI are privacy concerns which is fair a lot of us sacrifice our

16:12privacy on a day-to-day basis just to use the latest smartphones SmartWatches laptops so I think there's always you know some kind of playoff against privacy when it comes to convenience but I don't think we should be afraid of AI in general I think that the you know the way it's currently set up you know with L llms as a as a sort of a go-to approach to AI

16:35you know it's still got quite a lot of limitations it's not quite Skynet just yet and it's in fact quite quite a long way off and so even when it comes to this idea of creating AGI and sentients it's possible and it is a you know in theory project faster and faster but personally you know do I do we think do I think it's going to have a big negative effect on us I don't think

16:58so I think the advancement of humanity as a whole will continue plenty of opportunity and I think we're probably more of a risk to ourselves than AI ever will be that's an interesting perspective and I feel like smaller companies startups those in crypto that are willing to take risks or try new technology be more Innovative than you know some of

17:22these big four companies that I mentioned you know they have huge levels of bureaucracy and it takes time to integrate something like AI That's going to automate everything I feel like they're more willing to take risks is Sensei developing a technology that's also focused on small startups whether it's just you know the email part or something that hey we have a really

17:45small company how can we automate as much of this as possible so that we don't end up like one of these big four that have 20,000 employees and our overhead is huge how can we automate and sort of keep the amount of people small and keep the a big well that's it I think that the age of those sort of big companies is kind of over AI really does put a shift in that and yeah so

18:08when it comes to the way we structure we pretty much have it as a SAS model so we have Enterprise plans for larger corporations and organizations but it can be purchased and used by the individual as well so the packages do really range and the same accessibility the same features are all there the only difference is on Enterprise models they you know essentially have a internally

18:28host version of Sensei which sits you know sort of blocked off away from General platform and internet and in the same way that they you know they use other llms very sort of cautiously to not have access to their entire priv private data because so far in the use of AI and most people's awareness AI has been trained mostly on public data and all that

18:55public data whe it's been open source whether it's been you know farmed off S sites you just scraped from any any kind of public websites that have have scraping capabilities you know you end up with this huge amount of like public data that's there right on open and out there on the web but there's also a huge amount of corporate data and an organizational data that is still

19:17relatively untapped so there's a huge amount of private data you know kept on these systems internally isn't available to the public overall and not definitely not available to some of these other organizations now they may have made some private deal behind doors but fundamentally you know a lot of this data is still still theirs and still closed off and they're being very

19:36cautious about how they approach AI with good reason so a lot of them have got emerging Tech departments AI departments or someone specifically looking into you know which Technologies can be applied to them that would be relevant so I think there's still a lot of opportunity for a lot of smaller startups a lot of smaller companies to be very agile here and very fast mov

19:59moving in a way that the bigger bigger guys can't I mean you know there's always stories about how people are completely blown away by being shown chat GPT or how they see you know a random consultant from one of the big four for example going to a particular pitch and just pulling out something that was very obviously you know produced by chat GPT so there's there's

20:19still time there's still lots of time available the big guys do move a bit slower and definitely integrating is going to be slower but we are making sure that we can cater for organizations at all level and I'm very careful to say organizations here because it's not just companies that can use this kind of technology to make themselves more efficient it's any kind

20:38of organization including governments companies you can look at nonprofits Charities anyone really anyone can can really benefit from having you know essentially a replica Workforce it's a great point about the Privacy aspect of it and also with these big corporations you know if they have information that sort of privileged information people are normally paying for it

21:02they're not going to go and input all of that into an open AI model that's that's public so what you're what from what I understand what you're saying is with the Enterprise models for example if you are building a startup and you know this is like information that you want to keep away from everybody so they can't just copy your idea you're able to sort of have a

21:25more Enterprise privatized language model that you can build and train and sort of segregate so that hey our clients can use this but we're not going to like let everybody train from it yeah exactly so you know we're very aware that is a big issue for a lot of lot of companies so we have this capability to essentially run run a Sensei node within their existing

21:48systems that has its own model completely separated from the rest of our platform and yeah is available to larger Enterprise customers that want something like that and so that you we know that's really important for them for us it's you know just a part it's just another element of security in our system that we've had to put in from the early days so yeah

22:10it's yeah we know we understand it's important for a lot of people so we just we make sure that's always been something that we can offer and you mentioned you know the progression of AI is it gets exponentially better year-over-year I think the open a model is getting 10 times better each year and we still might be away away a ways away from you

22:33know sentience but it goes fast how is Sensei planning to keep up you know in 2025 AI is going to get 10 times better you know what's on the road map or what's the goal in making this functionality 10 times better and getting it into the hands of the right corporations and having the right customers yeah it's interesting I think someone recently published that we are

22:56no longer following Mo's law it's now now Mo's law squared which is actually crazy so it means that we've accelerated Beyond Mo which is amazing and yeah in terms of keeping up so you know obviously we want to provide as much utility for users now we're we're kind of predicting where things are going to go but we really think that you know by working on and improving constantly to

23:19build the best quality replicas of people what it does enables us to you know stay ahead by focusing on that sort of Personality engine how we really capture understanding what it is that makes us us you know what which parts of us make us sound more human and less like chat Bots and how that can be replicated how the different nuances of humanity how the different

23:42characteristics how the different you know emotions of us can be presented in digital format and by doing that I think that it's it's not the easiest thing to think about it's definitely not the easiest thing to build so by focusing on how that we you know how we build those out we become the experts in that specific field these replicas will just start to get more and

24:06more utility these different agents that are being built out in the space and the different utilities that can be applied to the senstive replicas will just be added on bit by bit and they just become more and more useful in the same way you know mobile phones went from being just kind of like oh you're an extension of yourself as as a quick way to phone

24:25people through all the way through to smartphones and continue to develop and add features on no it's a it's a really exciting time and I'm definitely going to get an AI to take over my email as soon as I can because it's just one less thing that I have to focus on and I can focus on the highle business you know with these startups there's a lot to do

24:47with less people and automation is good for small companies in crypto and outside of crypto as well so what is the best way for entrepreneurs or small startups or the big four to start looking into this technology and also to follow along with the updates as Sensei gets better with the AI training and everything else your team is

25:11releasing yeah absolutely so by want check out our website sense. that's senns y check us out on social any social media any channels so ask Sensei but get in touch I mean reaching out to us whether it's from a larger company or a small or medium startup or just company in general by getting in touch we are taking on a few design Partners right

25:36now for larger especially for larger Enterprises that we can actually build out solutions that they are looking for specifically and we're always happy at this early stage to be you know agile and have the ability to solve some of the simpler problems we've got quite an ambitious road map over the next two years both in terms of sales and on boarding quite a substantial number

25:56of customers but also terms of the features the inputs and the sort of external channels that we're working with so even now you can create your replicas you can have them on your on your own website as a widget on telegram on Discord very soon like WhatsApp and slack and a few others and then obviously the email drafter is coming at the end of this month so there's you

26:17know a lot of different opportunities for applicate applying the technology already but we're always happy to speak to anyone who has a specific you know use case or Niche that they would like to address and we can see what we can do so it's it's still at a very very fun time where we're really enjoying you know solving a lot of really interesting

26:36problems yeah all the best for your team on solving those I'm really hoping that it gets done because it's only going to help everybody be more efficient in business and I can also leave a link in the show notes to those socials in the platform I'm looking forward to that email drafter you mentioned there I'm definitely going to check that out and I would love

26:58to follow follow along and I appreciate your insights on AI as a whole in the industry and the growth it's so hard to keep up I need you to come back on the show more often so I would I would love to have you guys back on you know in the next quarter if not too much has happened since then if not sooner so much for taking the time

27:17Dan no worries well next time it will be your replica interview mine love it looking forward to it thank you thanks