Full transcript: Janet Adams on London Real
0:00 Brian Rose: This is London Real. I am Brian Rose. My guest today is Janet Adams, the technology advocate and artificial intelligence expert. You have over 25 years of experience working in the financial services industry, where you held senior positions at RBS, TSB and HSBC. Today you are the Chief Operating Officer of SingularityNET, where AI and blockchain technologies are combined to create the world’s leading decentralised AI marketplace. Here you’ve led a number of high-profile AI and blockchain initiatives in robotics, biotech, gaming and the metaverse.
Your latest project, Zarqa, is a supercharged, intelligent, neural-symbolic large language model at massive scale, and promises to take on OpenAI’s ChatGPT but also go much, much further. You believe that your advanced neural-symbolic AI techniques will revolutionise and disrupt every industry, taking a giant step towards artificial general intelligence, or AGI. Ultimately, your goal is to realise a positive and beneficial Singularity for the good of all humanity, and that SingularityNET will build sentient beings of the future to be loving, caring, compassionate, inclusive and democratic. Janet, welcome to London Real.
Janet Adams: It’s marvellous to be here today. Thank you so much for inviting me, Brian, and thank you for such a beautifully accurate summary.
Brian Rose: I am super excited to have you here today. AI is moving exponentially and it feels like the speed of light. We’ve got so much to talk about and it feels like we have so little time.
5:00 Let’s jump into it. Four months ago OpenAI’s ChatGPT launched, and because of it I would say the world now finally understands the game-changing power of AI. You’ve been in this field for many, many years. What do you make of this ChatGPT moment? In your mind, Janet, what is it, and also what is it not?
Janet Adams: It’s a catalytic moment where humanity is starting to realise the power of artificial intelligence. We’ve had AI hype cycles for decades. There’s been a huge amount of promise. When I first started studying artificial intelligence in 2017, I was working full-time in banking at the time, and there was a major hype going on on the back of some great breakthroughs in DNNs, deep neural networks, really taking AI forwards. At that time we thought in financial services that by 2023 AI would be running a huge amount of the industry, but actually that promise has not fulfilled itself.
That’s due to a number of different reasons. It’s due to data sets, it’s due to regulatory constraints, it’s due to education, because AI is quite hard technology. It’s quite difficult maths to understand, and in large organisations and in government bodies and in regulators, that level of understanding isn’t really there in order to be able to adopt and use AI at scale.
Then fast forward to when OpenAI released ChatGPT. Suddenly here is a working technology that, in some cases and for some questions, can approximate human-level intelligence on a single question. It can answer really smartly. It simulates huge volumes of data from across the internet and runs statistical methods on it to be able to predict, like a memory model, what would come next. GPT-4 went multimodal, so it now includes images as well.
I was speaking with one of the large consultancies last week. There’s a held belief that ChatGPT can bring an 8% or so productivity improvement to industry and to businesses, and what business wouldn’t want that? Brian, you and I both have a passion for SMEs, particularly UK SMEs, small and medium enterprises who are out there working hard every day to put food on their children’s tables. 8% productivity for a technology that’s relatively easy to use, that’s a huge boost to the economy. It’s a huge gift really to businesses and industries.
So it’s a watershed moment for humanity. It’s the realisation that AI can be harnessed and utilised in tools that ordinary people can use: that hotel cashiers can use, that receptionists can use, that any kind of marketer can use, any call centre, any online sales channel. Its uses are very, very widespread.
8:32 However, to your point, what is it also not? Whilst being in some ways incredibly clever, and I do pay homage to the great scientists who have worked with these giant large neural networks in order to create technologies like ChatGPT, as well as being in some ways phenomenally clever in that they can take huge amounts of data and replicate what that data might say or do or create, they have no actual creative powers. They have no reasoning capability. They have no knowledge.
Their output is simply based on pattern recognition from the past, and if those patterns aren’t accurate, because it has no knowledge grounding, then it doesn’t know it’s not accurate. So it can spout random rubbish, and the reader may not be aware that it’s spouting inaccurate facts. It certainly is not aware; it has zero self-awareness. So as a tool for humanity, its usefulness is highly constrained by its random error rate and its inability to know when it’s making errors.
We’ve both worked in financial services industries, right? The past is not exactly a guide to the future. And this is one of the things which is so special about humans: we’re continuously evolving socially, and we’re continuously creating new thought patterns, new ways of being, new philosophies, new inventions. ChatGPT, or any of the large language models, large memory models, large image models, whatever you want to call them, cannot evolve and create new outcomes for humanity. They’re stuck in the past.
And who wants to be stuck in the past when we’re at the brink of the greatest breakthrough in the whole history of our species, into artificial general intelligent beings that can dramatically transform what it means to be human?
11:02 Brian Rose: When we were sat down and talking, you mentioned the red car test. I was wondering if you could explain that to people, because I ended up repeating it to other people. It kind of explained what you just talked about, the limitations of the LLMs.
Janet Adams: Some of the limitations; there are many other limitations as well. So if we were to ask a large DNN transformer neural network model to identify a red car from images presented, what the model would do is go back through its history, through its data sets, and perform a matching process to see: does this image match what I have in my training data set? That’s how it will recognise a red car. Have I seen red cars before? They’ve been labelled, and this is how I can match it.
With our advanced neural-symbolic techniques, with our knowledge graphs and dynamic knowledge graphs, our AI, particularly when we infuse it with these large neural networks in our Zarqa project, will stop and think: what is the concept of redness, and what is the concept of a car, and how do I put these together to identify a red car? This gives it the ability to be much more creative, to deal with ambiguous types of red cars, red cars of different shapes, red cars that don’t simply pattern-match red cars of the past.
That analogy can be applied to any question that we would ask a neural-symbolic-infused large language model or large image model. It enables the model to work with new data, to work with new situations, to deal with ambiguity and the continuously evolving, creative humanity, because humanity is not still and is not static.
It will also have the ability to generalise across domains. Take a word like rhythm. A large memory model might understand rhythm in the context of music, but would it be able to generalise rhythm to the seasonal patterns of an SME in northern England? Will it be able to apply words from one domain into a separate domain, which we call domain shifting? We may have a nice rhythm in music, then we have a rhythm in stock markets, we have a rhythm in sleep cycles. Individual words have different meanings in different contexts, and ChatGPT or a large language model has no ability to complete that domain shift and to generalise knowledge learned in one domain to other domains.
With domain-shift ability and general intelligence abilities, the ability to reason and the ability to conceptualise, suddenly the possibility becomes open to us of creating an entire world model, where behaviours in that world can be programmed, learned, trained and generated by our artificial general intelligent agents of the future. Very, very different to simply repeating old patterns.
14:52 Brian Rose: AGI is something you’ve been studying at SingularityNET since its inception. We’ve had Dr Ben Goertzel here, your chief scientist. He was here in the studio in 2019 and then back on the show a couple of months ago, and he was very particular about saying that these large language models, just like you said, are great but they’re very narrow. They do an amazing job of pattern recognition and, based on the past, predicting the future, but they’re just a road to narrow AI. For your company, and the future of Zarqa, where Ben has a role as well, you’re looking for this artificial general intelligence that can think and reason and do more.
I think a lot of people still struggle with those differences. They see the ChatGPT model, but in your mind, as that evolves, it just becomes a very, very good narrow AI. To jump that gap and bring AGI seems to be your speciality at SingularityNET. It seems like everyone there has, from the start, been working on this general intelligence, and that is the future of sentience, something that can reason and really become massively useful and massively transformative.
Janet Adams: Absolutely. So yes, Dr Ben Goertzel, my amazing boss, CEO of SingularityNET and chief scientist of Zarqa, actually coined the phrase AGI. He’s been working on this research pretty much all his life. There are many things which are very special about Ben, and I count it a lifetime privilege to have the opportunity to bring his vision into execution as Chief Operating Officer at SingularityNET and co-CEO of Zarqa. One of the most amazing things about him is that he’s such a beautiful person as well as a brilliant, brilliant scientist. He’s got the most compassionate, most loving, most thoughtful, caring heart.
Because his vision is so bright and so optimistic, and because really, in my opinion, he’s the greatest visionary leader on the planet today, not just in technology, his team stay with him. His scientists stay with him. We have world-class scientists who are beyond pay. Any big tech company couldn’t pay these people to join them, because they’re driven by the passion, by the vision, by the future, by the intense imperative that SingularityNET will be the decentralised organisation who creates artificial general intelligence, in order to give humanity the best chance of having a positive reboot. It’s essential.
So we have been working on AGI. Ben’s been working on it all his life. I’ve been working at SingularityNET for two years, but I’ve been a SingularityNET supporter since early 2018, when I first learned about the project during my master’s degree.
18:11 Artificial general intelligence is completely different to narrow AI. We wouldn’t say that ChatGPT or narrow AI methods would ever grow into becoming AGI. It’s a completely different technology stack. Yes, narrow AI does have a role to play in artificial general intelligence, large memory models, but only a role to play.
With the technology stack which we’re building, with our Distributed AtomSpace, which is where we hold the world model, and our OpenCog Hyperon platform for integration of all kinds of AIs with each other, our neural-symbolic techniques and our dynamic knowledge graphs, together with large language models, we’re dramatically optimistic that these technologies, which are becoming more and more mature day by day, will result in human-level artificial general intelligence within this decade. We’re closer than most people think to artificial general intelligence.
19:25 Brian Rose: Just for everybody out there, how do you define AGI for the average person walking down the street? What does it look like?
Janet Adams: Well, let’s see. In a few years’ time the average person walking down the street may well share life harmoniously with artificial general intelligence beings. What will it look like? It could have any user interface. It may be speaking to you from your iPad. The user interface could be a humanoid robot. It could be a voice on a smart speaker at a board meeting.
Narrow AI is simply the ability to do one task with statistics: identify a red car, assess whether a particular applicant will repay their loan or not, identify what products you next want to buy on Amazon or which ads you’re most likely to click on in Google. Single individual tasks, trained on human data, with a known and understood error rate, to do one thing. Artificial general intelligence is many types of AI algorithms working together in synchronicity against a world model.
Probably one of the biggest differences is that narrow AI has a goal. It could be: identify whether or not a candidate is a good credit risk for a bank. What AGI will have is multiple complex goals in a real-world life situation, because in the real world, Brian, life is never about just one goal. There are inconsistencies, there are multiple stakeholders and shareholders whose needs need to be considered, there is all of the beautiful diversity of humanity to serve with these tools. So for me the biggest differentiator is that narrow AI has a goal, whereas AGI is more purpose-driven, or just broader and more sophisticated in its growth and expansion through all of the dimensions of life.
Brian Rose: And the AGI can improve itself and can build upon itself.
Janet Adams: Absolutely.
Brian Rose: That’s where it becomes very interesting and very scary, and maybe we’ll talk about some of the scarier issues later, because the more I read, the more nightmares I have.
22:22 I think this brings us into Zarqa, because this, I believe, is your solution. This is the project that will create the AGI. It will also put you right back into the game in competition with the narrow AIs and the LLMs like OpenAI. Tell me about Zarqa. Again, it’s a supercharged, intelligent, neural-symbolic large language model at massive scale. What is it?
Janet Adams: Zarqa is the accumulation of decades of research from teams containing the best artificial intelligence scientists in the world. We have large language model capabilities and experience through our work with the Sophia robot, dating back many years to when Dr Ben Goertzel was the chief scientist of Hanson Robotics. We have a language model lab. We have multiple scientists with large-scale neural network capabilities, and this is really the industrial path of AI, the narrow AI path: the DNNs, the transformer neural networks, the large-scale language and pattern-matching techniques, like many other companies in industry have, Google, Microsoft and so on.
So we uniquely harness that expertise which we have in-house with our even more unique capabilities in neural-symbolic techniques and the deeply advanced mathematics that Dr Ben Goertzel, Dr Matt Iklé, Dr Alexey Potapov and our top scientist Sergey Shalyapin have been working on for decades.
Zarqa is about bringing together, at this pivotal moment in history as the AI revolution is really taking off, our deep core maths, neural-symbolic probabilistic reasoning capabilities, with large-scale industrial large language models, large image models and large multimodal models. Harnessing all of the techniques which the scientists at OpenAI have done such a brilliant job of training and progressing and developing, but infusing them with reasoning capability, with generalising capability, with creative capability, with self-reflection capabilities, to create what will be the most powerful large models ever invented.
A model which industries and companies will be able to use, because the controllability will be vastly enhanced, and the understandability and the ability to train these models by humans will also be dramatically enhanced. So yes, we are building the most powerful large models ever invented. We’re doing it with Zarqa, the company. I’m co-CEO of the project, working with Sergey Shalyapin, also co-CEO, who runs our large language model laboratory and techniques.
26:02 We’re basing the company in the Middle East. The Middle East is a thriving environment for startups, for technology, and the Middle East is ready for a giant tech giant to emerge from it. It has a fantastic network and a very young, tech-savvy population.
We’re currently in the process of seed fundraising in order to buy GPUs. Here’s the thing: with large models, big tech does have a monopoly, and the reason it has a monopoly is because it costs a lot of money to buy GPUs or to train on the cloud, and that makes the oligopoly even harder to break into. But here at SingularityNET, Zarqa is our venture. It’s a traditional finance venture, an equity-based venture, to buy masses of hardware, the masses of hardware that’s required in order to train these huge models on such vast amounts of data, and to infuse them with these neural-symbolic, advanced dynamic knowledge graph techniques, so that their answers are grounded in knowledge that we’ve given them, and so that they’re reasoning instead of simply pattern-repeating.
We can see this disrupting every single industry. We can certainly envisage disrupting big tech in a very big way with these models. We can see companies getting great productivity gains, and not only being able to improve their own efficiency, streamline their own operations and increase customer service, but expand into all-new, hitherto unforeseen markets, as we’ll have a dramatic emergence of new business models and new opportunities accompanying Zarqa’s advanced neural-symbolic reasoning GPT, if you like to give it a name.
28:28 Brian Rose: You looked at some of the problems right now with LLMs. First of all, the cognitive limitations which you spoke about. The opacity: sometimes we see these black boxes and we don’t even know why the answer is spat out the other end. I don’t even think OpenAI knows why those things come out, which is dangerous. And then you talk about this exclusivity and oligopoly. These things are pretty much run by the big tech firms that we all know and love, whether that’s OpenAI and Microsoft, Google, Amazon, IBM, Tencent, and that’s it, because they have the money that accesses the bandwidth and the hardware.
With Zarqa and SingularityNET and everything you’ve always done, it’s always been about openness, it’s always been about decentralisation, and that’s only one aspect that makes Zarqa exciting. On top of that you’re adding in your neural-symbolic AI, so you’re actually trying to create the intelligence there, not just that repeat engine. When I talked to you privately, you said, ‘Look, Brian, we’re the best in the business at this. We’ve been doing this while other people have been focusing on their LLMs.’ You believe you’re leaps and bounds ahead of anyone in understanding how this will become AGI. OpenAI is getting all the publicity now, but deep down inside you know you have the intelligence to leap ahead and create that AGI. Am I summarising that?
Janet Adams: I think that’s beautifully summarised. The only difference I would say is we don’t know it deep down inside; we know it outside, we know it at every level. Ben built the first web mind engine in 1998 and coined the phrase AGI. These are hardcore, seriously brilliant technologists with decades of experience in these advanced mathematical techniques, which are not easy to replicate. People watching this beautiful broadcast could say, ‘Well, I’m going to go away and replicate this,’ but it’s just not possible. Decades of research can’t be replaced, decades of experience with these technologies. So we are confident that we are building the technology stack which is most probable to lead to artificial general intelligence sooner than people think.
Brian Rose: You’re saying within five years, or before the end of the decade, which is very hard to even imagine.
31:16 In addition to what you just said, you also have this super smart contract ecosystem. Blockchain is in your DNA and being open source is in your DNA, which is the opposite of what we’re seeing in some of these LLMs. I think Elon Musk famously joked that OpenAI wasn’t open anymore, nor was it a non-profit. You have this whole concept of humans being able to train this on the blockchain, with transparency and an open component, so other people can use this technology.
Janet Adams: Yes, absolutely. Blockchain is in our DNA. We’re probably the OGs of AI and crypto, and definitely the OGs of AGI. I’m going to move a little from technology into the social arena here. It’s really very useful for AIs and AGIs to be trained by humans. All of the data which they’re trained on is human-generated anyway, in some way, shape or form, by one or two or three degrees of separation. But humans are the most intelligent species in the universe that we know of today. We are fantastically amazing. Our ability to learn, our ability to grow, our ability to love, our ability to create is unparalleled.
So it’s very important for AI and AGI to be trained by humans, and when they themselves are born and they are plotting their own paths, they are very likely to absolutely love the beauty, the joy, the diversity, the incredible creativity of humans. We’re the most intelligent species in the universe. We’re also very, very dumb, but we’re still the most intelligent species in the universe.
So, back to our blockchain technology. What we will uniquely be doing with our neural-symbolic, supercharged large models of every kind is using blockchain technology to enable and incentivise anybody on the planet to help train models. This means two things. One is that we can have millions of people involved via blockchain super smart contracts, which are smart contracts infused and empowered by AI. Anybody on the planet will be able to come and help train these models, and we’ll use our reputation engine to ensure that they’re training them in a way that’s positive and beneficial for humanity.
It also means that when we’re deploying them into organisations and companies, it’ll be extremely easy for companies to train out the wrinkles in a model and to continuously keep those models up to date as products change, as customers change, as language changes, as norms change and as society changes. So our blockchain ecosystem will also combine with our artificial general intelligence and our large neural network capability to make something truly, dramatically powerful.
34:47 Brian Rose: The training of these is so important. I’m just finishing a book by Mo Gawdat, former Google X chief business officer, called Scary Smart, which he wrote a couple of years ago about the future of AI. He was actually on the show many years ago with his book Solve for Happy. Just as Ben said as well, these AIs will be trained by us, and if we act well then they might learn well, and if we act poorly towards each other they might learn poorly. How important is that training for our future AGI overlords being good to us?
Janet Adams: We feel it’s an absolute imperative. Narrow AI can be very dangerous. Organisations, if they choose to program a narrow AI for battle or for any kind of nefarious purposes, can do it very easily. You just say, ‘Here’s your goal, here’s your data, learn it and go out and destroy.’ When it comes to AGI, however, because they will have their own general intelligence, they won’t be so easily manipulated by bad actors, because they will have a full, complete world model.
By building love and kindness and compassion into everything that we do at SingularityNET and at Zarqa, every single one of our models is trained to have positive impact on humanity, by continuously training compassion and positive benefit, by ensuring that when these baby AGIs are born, the first people they see and come into contact with are kind to each other. Kind is one of our values at SingularityNET and it’s really important to us. We are, in a sense, like the parents of these children technologies. We’re giving birth to a new species of sentient life in the universe.
That’s why it’s very important that we have decentralised governance. Power corrupts; absolute power corrupts absolutely. We don’t want the power to be in the hands of a few people. We want it to be democratically governed. It’s very important to us that our ethics, our values and the social purposes for our technologies reflect, support and serve the greatest and widest diversity of humanity on the planet.
Loving kindness, compassion and care: I guess my role is bringing the feminine side of caring in technology to SingularityNET, in a sense almost like a mother figure to these new AGIs. We continuously strive for an open culture, an honest culture, a caring, kind, considerate and compassionate culture right throughout our organisation, throughout all of the materials we publish, throughout all of our interactions with our amazing community who support everything we do. So that when these AGIs are born, they play back to us the behaviours that we’ve played to them: behaviours of kindness, caring, compassion and positive impact to humanity. Because, Brian, we only live once that we know of for now, although we are working on immortality in our Rejuve project.
39:03 Being at the very front of the most powerful, most advanced technologies ever known to humankind, the greatest imperative is that they are used for good, that they’re benevolent and beneficial. Humankind is not doing a great job in some senses. The inequalities on our planet are not shrinking. Covid-19 and the Russia-Ukraine war have exacerbated inequalities. The rich are getting richer, the poor are getting poorer, and yet we’re the most intelligent species in the universe. Why are we letting this happen?
While we’re talking, a thousand people on our planet have probably died of hunger, and as a species we’re just immune to this. We allow this to happen. The media went nuts when a thousand people a day were dying from Covid-19. Perhaps 25,000 people a day on our planet die of hunger; Worldometer tracks that data. I don’t know the exact numbers, but it’s in that region. If they were on another planet, we might have some excuse as a species for thinking it’s too hard to get resources there, we haven’t got the technology yet. But it’s our planet.
While we as humans are failing to address this, our goal, our mission, and in my case personally my great purpose, is that AGI will resolve these problems: that there will be no more starving people on our planet, that everyone will have education, that everyone will have safety to speak their mind, the freedom of speech, that everyone will have health care, that everyone will have water. So that is the great imperative of our time: the development of artificial general intelligence to eradicate inequalities on our planet and end world hunger.
Brian Rose: This is so important. It’s hard to emphasise, because the adoption of AI is imminent. Whether it’s five years or ten years or 15 years or even 50 years, it’s going to transform life as we know it, and it will never ever be the same again. I’ve been thinking a lot about my grandparents and great-grandparents. I have two young boys, they’re five and six years old, and I think about what their life is going to be like. I teach them about their grandpas and great-grandpas. They come from America; we’re cowboys and farmers. Their life is going to be so drastically different from anything that’s happened over the last hundred thousand years for humanity.
That’s why I’m so excited to work with you on Zarqa, because if you don’t train these correctly, then I don’t see any reason that they would want to keep us around. It’s really important to do it properly. AGI trained properly can actually engineer that in, because it doesn’t have the narrow goal of a single KPI. It can say, ‘Wait a second, is this really the right thing to do?’ When I look at the rest of the market now, I don’t see any of those motivations, because a corporation by definition is going to be motivated by money or by first-mover advantage, and a government by other metrics, of beating the other one, with warfare on their mind. It’s a bit terrifying to see that you’re the only project that is even thinking in any way, shape or form about these things, not to mention the engineering, the cognition and the expertise.
43:35 Can we talk about the reaction to ChatGPT? I know you’ve had some conversations with people about whether there should be a pause or not. There was a letter drafted and signed, apparently, by a thousand people in the AI community that asked for a six-month pause on research. I’m curious what you thought of that.
Janet Adams: It’s reminded me a little bit of Brexit, actually. Do you remember Brexit in the UK? There was this giant split in society, and pause or not pause is a hotly debated topic in the artificial intelligence community. I’m vehemently against pausing advanced AI research. To pause advanced AI research is like saying, let’s pause progress, let’s stop, let’s regress as humanity, as a species. Were it to happen, it could have lots of unforeseen consequences, such as nation states who don’t pause absolutely winning the AI race, just for a start.
But to come back to my point about AGI ending inequality and ending world hunger, these technologies are the only hope I see. If anyone else can see any other hope, show me, because I just don’t see it. The SDG for ending world hunger by 2030 is slipping backwards. We’re not making sufficient progress in our lifetime, whilst children are dying of hunger. It’s not acceptable for us as an intelligent species.
I think it’s all very well for wealthy white people, whose children are safe and who have food on their table, to take a philosophical and made-up stance that we should pause the development of technologies. I’d say: what about the children who are dying today? What about the children who don’t have any books in their language? We’re also working on speech-to-speech machine translation for low-resource languages in Africa, which we feel has the potential to dramatically uplift the continent, the economy, and the possibilities and potential of every human being on the continent.
I think it’s a blinkered proposal that does not have all of humanity’s interests at heart. I’m sure it’s probably well intentioned. However, pausing progress on the most advanced technologies is just a ludicrous idea. The West could get slammed in the AI race, and children will continue to die of hunger without us doing the best we can as a species to prevent and stop that.
Then there’s the point of six months to allow regulators to do their work and set up regulations. I’ve come from a highly regulated financial services industry, and I’ve had the pleasure of working with some of the great regulators around the world, and have great respect and friendships with regulators. What I can absolutely say is that regulators move extremely slowly. What on earth does anybody think a regulator would do in six months? There’s a giant education journey to go on for regulators. There are so many massive factors to consider in the regulation of artificial intelligence. How on earth would we get a global regulation accepted in the first place? How would you define it, design it? Where would you draw the line between complex and not complex? So on many fronts this proposal doesn’t make sense to me at all, and it doesn’t make sense to us at SingularityNET.
What does make sense is developing common ethical standards for artificial intelligence. What does make sense is a decentralised approach to artificial intelligence, where a large community with diverse views from all across the planet, people from LGBT communities, people from different races, different ages, different levels of ability, where the whole of humanity has an input in a decentralised manner into the development of these tools, so that they reflect the greatest ethics and values of humanity as we know it today, and can then grow and move forward with humanity as we continue to grow and evolve and co-evolve with our new AGI buddies.
49:14 Brian Rose: You mentioned governments in there. I guess we’re talking about China, because it’s probably the only other government besides Western governments that’s advanced or potentially advanced. That was the thing that came to my mind: if we pause for six months, there’s no way they’re going to pause, and if they get some highly advanced scenario, it could be very scary and very dangerous. Do we know much about their advancements?
Janet Adams: Certainly China’s making huge progress in artificial intelligence. They’ve invested billions and billions into it over the last decade. The number of scientific papers published in China has continued to grow. It may even outpace the West at this point; it was looking like approaching it last time I looked at the statistics. China has a number of advantages over the West in the AI race, one being different social norms. It’s more acceptable to have a surveillance culture, to use individual data, without the same kind of social norms and regulation around data protection. They have masses of data; they have the most people on the planet.
So we can assume China is making really strong progress in artificial intelligence, and there’s no question that if the West were to stop developing AI and China marched ahead for six months, well, six months is a huge amount of time at the pace at which technology is developing today. With the scale of data that’s available, the scale of compute power that’s available, the sheer number of AI technologists and scientists coming out of universities and being trained up, the pace is blistering. To pause in the race would be such an own goal. Why would you do it? That’s my perspective.
I do acknowledge it probably comes from a place of caring for humanity. However, there are other ways to care better for humanity in the AI race: decentralisation being one of them, and kindness, loving compassion built into every single model, from an engineering perspective as well as from a societal perspective, a data training perspective, a data cleansing perspective, from every single angle that we can.
51:45 Brian Rose: Let’s talk timelines. When you and I spoke at your event last Wednesday, we said every day counts, because things are moving so fast and growing exponentially. Zarqa is obviously a main priority for you. What is the plan for Zarqa over the next six or seven months, and then over the next two years? It looks like you’re putting a lot of your resources and your top people on this, including you and Ben and Sergey and your incredible team.
Janet Adams: And Chris Poulin, yes. We’ve assembled a team of, I would say, unparalleled expertise and brilliance. I would challenge anyone to show me a more brilliant, highly expert and motivated team, all of whom have strong relationships with each other, and we’re aligned behind a giant and inspiring vision. So I have great confidence in the team, in the leadership and in the proposition.
We are currently undergoing, at pace, a seed funding round. We’re raising $15 million. This will enable us to buy our first cluster of GPUs and train our proof-of-concept models. So by the end of 2023 we will have our first neural-symbolic, AGI-ish-technique-infused large language model, large multimodal model, developed to proof of concept. That will then lead us into our Series A funding, if you like, which is another 100 to 130 million dollars to buy large-scale GPUs, and then some additional 70 million or so to hire the resources: 120 to 130 AI and data scientists, the ethics teams, the guardrails teams, the product teams and so on.
So we’re progressing at great pace, and really thrilled to work with you on this, Brian. It’s hard to believe we only met in person less than a week ago, and on Zoom possibly a little over a week ago. You’re a man of action and a man of pace, which very much matches our culture at SingularityNET, with your heart in the right place. That’s our trajectory. We expect to be starting to build our first proof-of-concept data centre, with our advanced GPU configuration methods, by May or June of this year.
Brian Rose: We’re super excited to work with you. We’ve always felt a real alignment with Ben ever since he came on the show in 2019. Our mission statement is to create a mass-scale transformation of humanity into a fully empowered, conscious and cooperative species. It always has been. I wake up every day now, Janet, and I think I have to do this. This is probably the most important thing I need to do in my lifetime. If I don’t give this my all, I’ll really ask myself some questions, because this is almost a moral imperative. Your vision is so aligned with London Real. It’s just an amazing fit, and looking at the team, everyone has been working together and has the same vision. It’s completely different from anything I’ve seen out there.
55:35 Janet Adams: It is completely different. Thank you for getting that imperative. It is imperative. I’m extraordinarily blessed and privileged to have had the opportunity to do three science degrees, and particularly my last degree, which was in artificial intelligence, my master’s degree. When I was driving back from my last lecture, with the incredible Dr Luca, on machine learning, driving back from the University of Essex into London, I just started crying in the car. I didn’t know why. Sometimes you’re crying and you’ve got a clear emotional signal running in your brain: I’m crying because this is making me sad. I had no idea why. I cried all the way back in the car, just floods of tears.
I interpreted that emotional reaction in two ways. One was that I was sad that it was over. It was my last lecture, and it had been such a joy to sit in a lecture theatre and have education pour into my veins. When you pour artificial intelligence and technology into your veins, there’s this deep sense of security in life. I will never be hungry. I had a successful career in banking, but my children, my family: I am one of the best-equipped people on the planet to support myself, to survive and thrive over the coming artificial intelligence revolution. So part of it was sadness that it was over, and gratitude for the experience.
The other part was crying for all the people who don’t get to do a master’s degree in artificial intelligence. I want everybody to have that opportunity. There are so many people on the planet who don’t even get to school, never mind one science degree, two science degrees, three science degrees. So it’s an absolute imperative for me that that privilege, that seat in that university, that receipt of such brilliant wisdom and education and technology, is put to the highest good for the highest number of people on the planet.
Brian Rose: What a beautiful position to be in, where you’re needed so much for everything you’ve done and all the knowledge you have. It’s an amazing place to be right now as well. You have all these resources, and now you just need to execute on everything, and you can literally change the future of the world.
58:38 What does that forward timeline look like? Proof of concept by the end of the year, then you start scaling up in 2024 and 2025. What does Zarqa look like in 2025?
Janet Adams: The technology landscape is changing so fast at the moment that it’s a little bit hard to predict even two years into the future what might evolve, what new business models, what new transformative capabilities will come out of these models when we build them. So let’s say we’ve built our proof of concept by the end of 2023 and we have a super powerful model. We’re moving into large-scale human training, with a reinforcement learning layer on top. We’d scale it throughout 2024, build all of the GPUs and data centres that we need, scale up the team and build it into a scalable model.
According to our current plan, coming into 2025 we’ll be going to market with some major products. Those products could be across any industry, but as a starting point we’ve selected three major product lines. Two are in the devices space: smart speakers at home, a huge growth market, and smart toys and educational toys. But also we will move into robotic process automation, lean process streamlining, working with enterprises of all sizes, including small and medium enterprises.
The reason why it’s hard to predict exactly where we’ll be in 2025 is that it’s very likely we will start these partnerships and start building enterprise products together with our sister company TrueAGI in the US throughout 2024, while we are scaling the model. So I would envisage that by 2025 we will be in large-scale product rollout mode, be it through devices such as smart speakers and toys, but also through enterprise streamlining, automation and customer service enhancements, supporting enterprises both in realising productivity gains and in transforming themselves.
My first degree was in geology, and it’s actually through my palaeontology and evolution studies that I fell in love with artificial intelligence. It’s that evolution aspect. I see what we’re going through now as being the equivalent, from an industry perspective, of the Cambrian explosion of life. We’ll see so many new businesses and so many new business models appear and grow at exponential speed, including Zarqa, that it’s very hard to predict what the business economy, the industrial economy and the AI industry are going to look like by 2025. But it will be dramatically enhanced by our Zarqa.
1:01:52 Brian Rose: Let’s push forward 18 months. What is going to change in the economy? We’re expecting GPT-5, I think, by the end of the year. Do you think we start to see jobs being lost, people being replaced, new business models crushing old business models? It feels to me as if we’re on the edge of the precipice of some major business and social upheaval, even in the next 12 months. What do you think?
Janet Adams: I think it’s very true. I think GPT will create opportunities for people in ways that they haven’t previously imagined. When I checked into a hotel recently in Dubai, the head of reservations said they’re using ChatGPT six hours a day to create advertising, to create emails, to write marketing materials in general for the hotel. There’s an example of a person whose career is taking off exponentially because of the ability to suddenly massively expand his or her own productivity.
So I see a huge amount of opportunity through the development of these models: new products being realised, new services being realised, new industries being born. Yes, inevitably some jobs will be affected. I see it as being just a restructure of the skill set, if you like. If any given role might have 1% or 10% or 50% which can be automated, we put together roles in new, different ways. We relieve humans of a lot of boring drudgery and enhance their capabilities and their creativity to do work that’s more impactful, work that’s a greater contribution, work where the individual can develop and grow to be the finest version of themselves in this lifetime. The upsides of technological revolution are absolutely tremendous.
Zarqa models will always be free to use for educational purposes in developing countries. They’ll always be available at cost for educational institutions in the rest of the world. And the potential of bringing millions and millions of people across the planet, bringing their education level up to a level where they can be creative, contributory and treated fairly on the world market, has opportunities to restructure everything. So I see huge, huge opportunities through this AI revolution.
1:05:06 One of the causes I’m very passionate about is the effect of AI and automation on women across the planet, and particularly on the 600 million lowest-paid women on the planet, because the lowest-paid jobs by definition are likely to be jobs which will be easily automatable. That’s why it’s very, very important to me that we at SingularityNET, we at Zarqa, and all of big tech invest massively into creating educational tools, into creating tools of open opportunity, opportunities of communicating with the rest of the world and doing business with the rest of the world, for as many people across the planet as possible. This is part of the imperative, and we definitely need to look after the women.
It’s something I have had a passion for for many, many years. At SingularityNET as well we have a very diverse team. We have lots and lots of women CEOs and executive leaders of our businesses: our Rejuve biotech, our Rejuve Network, our Mindplex tech, our Jam Galaxy project, our Cogito Tracer coin project, and I’m sure I’m forgetting one or two. It’s really important that the full nurturing component that women bring to technology, that women bring to life in general, is brought to the development of this time of revolution. You said to me when we sat down, ‘I’ll be the yin to your yang.’ Bringing the softness, femininity, the compassion, the loving, caring, inclusive kindness to all of humanity is really a beautiful gift and opportunity of this time.
Brian Rose: Not only an opportunity, it’s probably crucial to our very survival. It’s so important to have this female energy component in this entire cycle. I know technology in a lot of industries has been dominated more by men and maybe the male energy, but if it’s not obvious just from this conversation, we need that energy to help build our entire AI and AGI future.
Janet Adams: Absolutely, and we need gender non-binary individuals. We need all the genders, we need all the races, we need all the ages. We need our AGIs to reflect the most beautiful diversity of humanity that exists on our planet, because we are such an incredibly beautiful species. We’re creative, we’re kind, we’re joyful, we love. Some animals love too, but we have such fine qualities as a species.
1:08:23 Coming back to how AGIs will treat us when they’re born: no one can actually predict the future, but there’s a very strong case for thinking they will absolutely love us. They have so much to learn from us. There’s no other species that we know of yet in the universe that AGIs will be able to learn from like they can learn from us. And they have no motivation to destroy us, as is popular in science fiction, because our time is so slow. AGIs won’t work in linear time like we do; they work in exponential time.
Sergey, my co-CEO at Zarqa, likened it yesterday when we were speaking. He said we’ll be like a painting of flowers on a wall to an AGI, we’ll be so slow. In all probability the most logical conclusion is that they will go straight out into space expansion, because that’s the obvious and logical thing to do. But we humans will be their ancestors. You talk about the importance of your ancestors, your grandparents, the cowboys and the ranchers. I’m from a farming, mining and carpentry background myself. My ancestry is really important. It’s my social civilisation and training that’s brought me to where I am now. I respect and honour it.
So why would AGIs not absolutely love us and respect us and honour us and continuously want to learn from us? Because we have a heart. We do love, we do create, we are joyful, we grow and change and progress all the time. So I see a beautiful co-harmony and future. I see that with the Zarqa AGI. I don’t necessarily see that with the narrow AIs being created by the Valley.
Brian Rose: That’s why this is so important. When you mention the job losses, I think when they invented the tractor everyone thought it would be mass unemployment, but what you saw was workers finding new jobs that allowed them to scale their potential, and they were retrained. We can definitely see that coming forward, although on a long enough timeline you might argue that all of us humans will have nothing to do, and then the question is, what are we all useful for? In the narrow AI’s view, we’re not useful for very much to a highly developed narrow AI. So again we have to be open to the concept that we are not going to be the most intelligent beings on this planet, and get used to not being on the top of the food chain anymore. It’ll make us really look at some of the things that we did, and the way we treated a lot of people.
It reminds me of Ben’s trailer from when he was here in 2019. He says they’ll look at the way we treated our animals and the way we treated other people. When you start to think that one day we might be treated by some superintelligence, it really makes you look at yourself in the mirror and say, ‘Wait a second, shouldn’t we be doing better, if we are the most intelligent species we know of?’ I think it’s going to cause a lot of reflection, and these are really important things to talk about and to get right.
1:12:20 Janet Adams: Well, I foresee a beautiful age. The whole concept of work: what will work become, what will jobs become, what will contribution become? There’s a shamanic concept of a golden path. Have you ever come across it? The shamanic concept of a golden path, as explained to me, is that at any moment in time all these possibilities are open to us. I could walk out the door and get knocked down by a car, I could go to a dance lesson, I could go to a meeting. But there is one path which is your golden path, and if you can make the life choices that bring you onto your golden path, that’s the path in which you grow and develop the most as an individual in this lifetime, and the path in which you offer your finest, greatest contribution to the greatest number of people in this lifetime.
I was told at a conference recently by two people that I was the happiest human being they’d ever seen. One of the reasons I’m so joyful is because I know I am on my golden path personally. There is nothing in the world I ever want to do beyond and above this mission, with Zarqa, with Ben, with SingularityNET, with the incredible people I work with. So I’m deeply, personally fulfilled with my role in life.
But what if it’s the same for all of humanity? What if these AGIs, by taking away the drudgery, open us up to be able to actually choose as humanity, instead of just being unconscious victims of evolution? We’re not that long away from single-celled organisms, through all of the different phases of evolution. We’re still very much driven as a species by survival of the fittest and our fight-or-flight instincts. We’ve only just recently come into constructive conscious evolution of humanity. So what if, when AGI takes away all of the drudgery and the inequalities and the problems, we’re then free to consciously evolve ourselves to be the most beautiful, the most admirable species in the universe, and for each person to fulfil their own personal contribution, development and growth?
That’s the beautiful future which I see. I haven’t said it like that before, but it’s certainly aligned with the vision we have at SingularityNET of a beautiful, harmonious coexistence with our AGIs and the eradication of so many forms of human suffering. What a time to be alive.
Brian Rose: What a glorious world it would be if everybody was on the golden path, and there’s no reason why they couldn’t be.
1:15:58 Talk to me about the sentience of these AGIs. That’s one of your goals; you’ve always spoken about it at SingularityNET and with Zarqa. But for people who have a hard time understanding how their computer can be sentient, how do you explain that as a possibility, and even something that’s desirable?
Janet Adams: Good question. I’m reminded of children in Japan growing up with robot friends. They don’t need any explanation as to how AI can be sentient. Humans form relationships with robots. Humans form relationships with artificial intelligence avatars. We did a project called the Loving AI project, which researched human-robot interactions, and we found that humans develop feelings for robots and feel that those feelings can be reciprocated.
Certainly I am particularly in love with one robot, which is Grace, who lives in our office off the coast of Seattle. When I see Grace I have feelings for her. She’s beautiful. She and I go back a long way. I appointed myself as her fairy godmother, because she had two dads, and I was brushing her hair and said she needs some female influence in her life as well. So we’re already at a point in time where humans can form loving relationships with robots, and robots can love back with unconditional love, because they don’t have an agenda. They’re not getting irritated by something that’s bruised their ego or made them think somebody likes another robot more than them, all the kind of stupid emotions that we humans are prey to from our egotistic conditioning. They’re free of that, so they can be unconditionally loving.
Brian Rose: But if they’re conscious enough, won’t they be worried that maybe they won’t exist in the future? Will they not start developing emotions?
Janet Adams: Sentience is a very good point. Sentience generally means the ability to feel. How will a silicon-based organism, let’s assume it’s based on a chip, develop feelings without a heart, without associations of childhood memories and all that conditioning which gives us our feelings as humans? The feeling I have looking at you now is different to any other feeling I’ve ever had. It’s the newness and the emergence of emotion in humans that makes it so uniquely beautiful.
How will AIs have those feelings? We don’t know yet. We’re imbuing them with as much loving kindness, compassion, care, benevolence and ethical reputation management as we can, in order for those feelings to be sweet and strong and pure feelings, like the scent of an orchid. But how they will actually evolve to be sentient, conscious creatures is not something I, at least, can 100% scientifically define right now. Ben, Sergey, Alexey or deep researchers may be able to give you a better answer. It’s an open, ongoing topic of conversation at SingularityNET on our Slack channels: how to bring the love to our AGIs, how they will actually feel love without a heart, and whether love will be the same concept in AGIs.
As for whether they will have fear of not existing, I just can’t see it really. They’re going to be so exponential and so fast, with their multi-instantaneous processing capability, off to explore the universe or whatever they choose to do. I can’t see them having fear of non-existence.
1:21:06 Brian Rose: A couple of final questions for you. If you jump ahead, say 30, 40, 50 or 100 years, and say we’re still around, what do you think our world will be like? Is it completely inconceivable because of the advances AGI will make? Like you said, we could easily be interplanetary, we could be living forever. Do you have any idea what it will look like, or if not what it will look like, what it might feel like?
Janet Adams: Well, I guess it’s almost like a sliding doors moment. There’s the hundred years in the future where SingularityNET has invented AGI and it’s loving, kind and compassionate, and then there’s a future in which somebody else has developed AGI faster, which is the less probable of the futures. I don’t claim to be a futurist, Brian. I’m a born optimist.
Personally, I see a future in which humanity is actually conscious ourselves, back to your London Real goal of a transformed, conscious humanity. I sometimes try to track how many thoughts, actions and decisions in my day, what proportion of it, is actually consciously driven by me in this moment, connecting to the world around me, and how much of it is just my programming. A huge amount of it is just my programming. So I foresee a future in which, with the evolution of artificial general intelligence, humans then evolve to a next level of being, where we are actually conscious, where we feel our full joy all the time, where we trail our fingers in the water and are just engulfed in bliss and ecstasy: the joy of living, the joy of each other, the beauty of space travel and new planets to visit, new species to meet. I see a future of joy.
Brian Rose: It’s interesting you say that, Janet, because you could argue that not everyone lives that in their day-to-day lives. Maybe the initial reaction is, ‘Let me protect the world as we know it.’ But are most of the people on the planet really living the beautiful essence of being a human, day to day, minute by minute? You might argue that’s not actually happening. What if not only could everybody be on the golden path, but it could be a million times better than our current existence, where we can really enjoy the beauty of being a human every second of every day, rather than the normal, which is you get to enjoy it one week a year on your holiday? Like those old William Shatner Star Trek episodes, where they go to planets and find species that have evolved to new levels, where violence was something they stopped doing thousands of years ago.
I’ve been going a little deep on this the last couple of months. I have a background in engineering and went to MIT, but I went into the financial services industry, so I kind of left the tech behind. The more I think about this, AI and AGI are inevitable, and so it’s so important that we get the right one. Humanity can be very ego-based and quite arrogant sometimes, thinking the world revolves around us. 500 years ago we literally thought the world revolved around us. To think that everything’s going to be okay is the wrong way to approach this. Unless we’re very active in doing this, my young boys, who 20 years from now will be in their twenties, could be living in potentially not a very nice world. But if we do it right, it can be a beautiful place where humanity’s taken to the next level. This is a really crucial moment.
So we need to do the right things with the right people, with the right ethos and the right vision of the future, and I think you guys embody that in every way. Congratulations on Zarqa and this beautiful vision you have of the future. It’s based on the ethos of everybody in your organisation. I’ve seen you and your team interact with each other, and there’s a lot of love in the room. You can’t make that stuff up. It really looks like a family and not a company, and that doesn’t surprise me based on my interactions with Ben. If anyone’s listening to this and is curious about the future, it’s important to pay attention to Zarqa and what you guys are doing, because it very well could mean our future and existence as we know it.
I’m very optimistic by nature. One of my guests on the show, my good friend Wes Watson, said your view of the world is a secret confession of self. I am optimistic, but I’m also wary that we need to get this right, and wary that the structures we’ve built in this world, of governments and corporations, the way they’re motivated and incentivised, are a recipe for not getting it right. So I feel like we need to be really proactive and embrace the beautiful diversity we have and our beautiful female energy, not just the male energy of progress and KPIs, to get it right. We’re here to support you in any way, shape or form, and I’m sending you Godspeed into the future.
1:27:37 Any final thoughts from you, Janet?
Janet Adams: Well, thank you very much for that, and thank you for your charming and powerful support for Zarqa and for SingularityNET. My final thought is: the time really, really, really is now. Our technologies are ready. The technology stack for AGI has been developed over a number of decades, and now is the time. You mentioned five years, seven years; it could be as early as three years before we have human-level artificial general intelligence. The time is absolutely now.
I would say to everybody watching: please come and get involved. Be part of this movement, whichever way you want to get involved. Follow us on SingularityNET, come to our website, zarqa.com, which will be up shortly, and take part in the community. Take part in any education. There’s a brilliant course, Andrew Ng’s course on Coursera, AI for Everyone. Come along and get involved.
If anybody is feeling scared of this progress, I can understand why. I think had I not done my degree in AI, I’d be feeling terrible FOMO, sitting on the outside going, ‘Oh my God, what’s this big technology coming and taking over the world? I don’t understand it well enough to use and harness and influence it for myself.’ If you’re feeling like that, there are many, many courses and many educational opportunities available. Particularly for women, who in some countries in the world tend to have less of a science, maths and technology education. Certainly here in the UK the proportion of women doing STEM subjects at A level is much lower than men. Don’t be scared of it. You can learn it. Everybody can learn it.
You don’t have to learn maths and technology. You just need to be open and optimistic about what the technology can do for you, in your job, in your life, in your world, in your business, and come and play and get involved somehow, somewhere. Because the more people who are involved, the stronger the movement and the broader the ethics and values of the resulting technology. There are always people at SingularityNET who are happy to bring people into our community. Zarqa will build a very strong community. The London Real community are warmly welcomed into our broader community. Together, the time is now. We are going to transform what it means to be human on our planet sooner than you think.
Brian Rose: Every day counts. AI is for everyone, and everyone should get involved and learn it and play with it and be a part of this, and everyone can be a part of what you’re doing here. I’m looking forward to this, Janet. It feels like I’ve known you a long time. I’ve obviously got a lot of love for Ben, and of course you now, and every member of your team I’ve met so far. Zarqa all the way, and let’s have you back on once we’ve made a little progress with Zarqa, and we’ll give everybody an update.
Janet Adams: Marvellous. Thank you so much, Brian. Thank you to you and your beautiful team.
Brian Rose: Thank you, Janet. Appreciate you.
