On this episode of the AI For All Podcast, Adnan Masood, Chief AI Architect at UST, joins Ryan Chacon and Neil Sahota to discuss AI robots and the future of society. They talk about physical vs software robots, the limitations and challenges in robotics, how robots learn, business use cases for AI robots, AI ethics and bias, artificial general intelligence, and what the future of AI robots will be.

Adnan Masood is an engineer, researcher, and forward-thinker who is passionate about developing innovative technologies. A thought leader, Dr. Masood’s expertise includes scalable enterprise architecture, machine learning, and cloud platforms, particularly Microsoft Azure and AWS. As a Microsoft MVP for Data Platforms, Adnan has extensive experience in developing secure, PCI-compliant fintech solutions, with publications around functional programming, machine learning, and application security.

For more than 23 years, UST has worked side by side with the world's best companies to make a real impact through transformation. Powered by technology, inspired by people, and led by purpose, they partner with clients from design to operation. Through their nimble approach, they identify core challenges and craft disruptive solutions that bring their vision to life. With deep domain expertise and a future-proof philosophy, they embed innovation and agility into clients' organizations - delivering measurable value and lasting change across industries and around the world. With over 30,000 employees in 30+ countries, they build for boundless impact - touching billions of lives in the process.

(00:39) Introduction to Adnan Masood and UST
(00:50) What are AI robots?
(02:31) Physical robots vs software robots
(07:38) Capabilities and limitations of AI robots
(10:04) How are businesses reacting to robots?
(14:07) Use cases for AI robots
(16:40) Ethics of AI robots
(18:40) Does AGI need a body?
(22:13) How robots learn
(24:50) Will robots replace humans?
(28:26) Hybrid intelligence
(30:25) Attitudes around robots
(31:43) AI bias
(36:06) What will the future of robots look like?
(38:58) Learn more about Adnan and UST

- [Ryan] Hello everyone and welcome to another episode of the AI For All Podcast. I'm Ryan Chacon, and with me today is my co-host, Neil Sahota, the AI Advisor to the UN and founder of AI for Good. Neil, how's it going today?

- [Neil] I'm doing all right. I hope everybody out there is having fun times. I know it's a lot of stuff going on and I'm looking forward to the conversation today.

- [Ryan] With us also is our producer, Nikolai.

- [Nikolai] Hello.

- [Ryan] On today's episode, we have some really interesting topics we haven't covered yet. We're going to be talking about AI robots. So how close are they to being a reality? What role will they play in society? What are the dangers that we need to keep in mind when it comes to AI robots?

And to discuss this, we have Adnan Masood, Chief AI Architect at UST, a digital transformation company. Adnan, thanks for being on the podcast.

- [Adnan] Thank you very much for having me.

- [Ryan] Yeah, let's kick this off. Let's get right into it and talk about just what does it mean when we say an AI robot? Obviously we know what robots are.

We know what AI is, but we put them together. What does that mean? How should our audience be thinking about what it is when we say AI robot?

- [Adnan] Yeah. Thank you for having me. And I'm really excited to delve in this topic of AI robots and share the insights around that. So imagine a machine is delicately navigating an intricate landscape, it's identifying different objects and a path, it's responding to it like in a human commands in a matter which is like very surprisingly unique and human, right?

So, that's not science fiction anymore. That's something that you are seeing today as intelligent machines exist. Our cars are intelligent enough to drive ourselves. We have intricate knowledge around the area of how humans react. And the machines we have now have cognitive capabilities per se, and they can exhibit those qualities.

So these qualities combined together, again, artificial intelligence itself in its definition is quite difficult to define that. So AI robot on top of that, it becomes even more difficult. But essentially, the idea is that it leverages sophisticated sensors to perceive the environment.

So understands the environment similar to how humans, as humans, we sense and understand the world around us. And what sets these AI robots apart is the ability to process the sensory information using AI algorithms, and it leads to this intelligent decision making and the capacity to perform these complex tasks.

So navigating through the environment, whether it's driving, whether it's interacting with their surroundings. That's what an AI robot is. Now, I would like to point out a very important distinction here. There is the software robots, right? So there's robotics process automation as we call it, RPA, and that's quite different from I think the topic today is more about AI robots as in what robots are integrated with artificial intelligence, the physical robots and mechanical robots, the robots we see as in Terminator style robots for lack of a better term.

But yeah, those are, so but robotics process automation or software bots is also another one of the major category, which does the work behind the scenes, does the back office work, etc.

- [Ryan] Gotcha. So like when we, if you go on Instagram and TikTok and you see Boston Dynamics and those robots running around, jumping around, that's more AI robot?

- [Adnan] That's right. So those are the mechanical robots infused with this whole artificial intelligence paradigm with that they can do this. But think about like software AI robots which are working like customer service or marketing or fraud detection or robotics process automation, writing creative text, your Alexa will be classified as a part of like a software bot because it's a device with a bot in there.

So that's the, a very important distinction to make between those. And the MIT Head of CSAIL, Computer Science and AI Lab, actually doesn't consider these software bots at all. For her, the bots have to be something which interacts with the world and moves around. And I think that's a very good definition for a variety of different reasons.

So, I guess a good question would be, can you talk us through some of the common use cases for AI robots? Obviously there's a lot of futuristic things that we could talk about, but just like right now in today's world, where are AI robots really playing a role in enterprises and businesses?

So in enterprise and businesses along with, so again let's talk about AI robots from a humanoid versus the soft robot perspective, right?

So we don't consider robots anymore as some humanoid form. So we don't anthropomorphize them, but rather self driving car, we realize that is a bot. That's a robot which has significant software component and artificial intelligence component. But at the same time, like I said, virtual assistants are also bots, but they don't have right now physical bodies, but they may eventually have physical bodies.

So that's why it's a paradigm or a spectrum where how we go from physical, physicality of the bot or the robot. But we are already living in this age when AI robots are a reality, like they're in the rudimentary form, they're already here. For example, autonomous vehicles. We talked about that.

Drones, robot vacuum cleaners for our house, like Roomba out there probably listening to me. The industrial robotic arms. So all of these things, industrial robotics arm especially, has been around for a very long time. The supply chain and manufacturing plants, they're all using these robotic arms and robotic instruments to actually manufacture large scale things. Also, like when you're talking about like sophisticated humanoid robots that you're talking about that they perform tasks as well as better than humans sometimes. We are or fairly a bit there, so it's a mix.

So for example, if you're looking at robotic surgery, the precision, the accuracy which it can perform the surgery with, it's amazing, right? And that's something that at the scale which humans are have a hard time to do that, especially all also at scale car plants again, right? That those robots generate so many cars or so many vehicles or so many machineries so quickly that it won't be possible for humans to do that. So we're seeing it happening for a very long time. And now we're seeing these mini robots coming up in the industry and in everyday lives, they're becoming a part of our everyday lives. But I think a more important distinction here to make is that now with generative AI and artificial intelligence, we will eventually relead a threshold where we'll have our robotics have that ChatGPT moment where it will tip over.

Again, robotics is a harder problem because it interacts with the world, it has a physical scale, it's hard, but because of generative AI and because of artificial intelligence in general taking so much, accelerating so much, it will also propel the progress in robotics.

- [Neil] Maybe ask the tough question I think a lot of businesses are facing, right?

Three weeks ago at the World Mining Conference, there were like 20 different robots to help with various mining activities. From little robot dogs to help patrol the mines to drills. Two weeks ago at AI for Good, we had over 70 robots there. So robots that ranged from being kind of artificial empathy connectors to robots that paint to robots that do firefighting, like wildfires.

So in some regards, people are like, wow, the robots are already here. Yet at the same time, everyone's like it can't be like Rosie the Robot from the Jetsons. It can't clean my house. There's a dexterity problem. If you're a business thinking about how you can take advantage of robotics, what are the capabilities and limitations that you need to be aware of to make an informed business decision?

- [Adnan] Dexterity problem and multi modal and multi dimensional, multi faceted robots, you're absolutely right, Neil. Those are big problems in that space. And we have been working on that for a long time. There was a recent book by McKinsey on AI digital transformation I was reading up and there are codes in there about how we haven't been able to reach that that threshold in robotics especially.

And if you see right now, a human arm is such a intricate and sophisticated device that we haven't been able to replicate that, right? We haven't been able to create something which can actually pick up with that dexterity, hold that and being able to transport it, and it's just amazing how nature has evolved to create that. So those are all true things. But at the same time, we should not discount the ability to be able to create things which can work non stop without the need to eat or sleep or you know do bodily functions or interact with humans, and they can just keep working. So the leading places where I am seeing robotics being used is healthcare, of course, agriculture and retail.

So those are the three areas and especially in retail you probably have seen a lot of robotics being used especially around these stores where you can just go in and do the shopping and get out, inventory management, control, supply chain. Those areas and agriculture has tons of different examples associated with that, things you have seen. Healthcare again is a major area in that space and that's robotic surgery, precision medicine.

Those things are evolving to make that. And I think they have a strong potential for early adoption. And looking forward, I think any industry that requires performing this repetitive and especially the dangerous or physically demanding task is a potential adopter for AI robots.

- [Neil] Really insightful. And I call it the Spock problem, which you actually alluded to that we want AI and AI robots to be perfect, which they never will be, but it's cool that a robot could do like a thumbs up, but it can't make the little Vulcan symbol. It's like don't worry about what it can't do.

Understand what it can do. Don't worry about that it can't do that. Are you really using this in your business. But if you get value out of being able to thumbs up, focus on that.

- [Ryan] When you talk to the public, potential customers, other organizations about adoption on the AI robotics front, what's the sentiment?

What's the feedback you get? Are there any hesitations or concerns that they have that may be worth mentioning here to look at the other side of it? Because obviously when we talk about robots people think they're going to take over the world, they're going to do all these things.

They're going to potentially cause danger. There's this new show on Netflix that just came out talking about robots being, AI robots being used in like warfare and how that's very dangerous. I haven't watched it yet, but I watched a preview of it, looked very interesting. But there's obviously this other side to it.

So I'm curious from your perspective what conversations you've had or what maybe feedback you've received from companies who are more hesitant to bring in AI robots into their business or their industry.

- [Adnan] That's a great question. Most of the customers I work with, I'm the Chief AI Architect here at UST, and we are a solutions company, and I'm also a Microsoft Regional Director, which is essentially an honorary role where you work on cognitive API and work with the Microsoft product teams to actually see how we can operationalize the AI work, which Microsoft has been doing.

And most of this work is mostly around the software side of the robotics, artificial intelligence, the back end side of the robotics. For mostly for the front and or the actual physical robotics part of that, it depends on the industry. For example, if you are talking about a new, I was talking about different industries, which are very highly impacted or very useful in that.

In manufacturing, if you're talking to clients, they can use robots to automate these repetitive and dangerous tasks. If you're talking to agricultural customers, their robots are helping in planting, harvesting, and monitoring crops, right? And healthcare, same with surgery, patient care, diagnostics, even maybe at some point we will be able to do the complex tasks like elderly care, teaching, and complex problems like that. So these are the problems at scale. And these problems, regardless of the customer sentiment, these are really well suited for what robots can do at scale. And those are essentially very well suited for those kind of things to be done.

To answer your question, your specific question around what the sentiment is mostly around, depends on the industry and the usability. And of course, safety is always a big concern, right? So you want to make sure that robots are safe and the environment there is beyond Asimov's Three Laws, right?

It's making sure that it's actually working in the real world, in human environments. And the benefits which AI robots provide are phenomenal, right? If the customer realizes, which they mostly do, is how they have the potential, the robots have the potential to boost productivity, safety as actually if you look at the record of how manufacturing has become more safer with robotic interaction, aside from some freak accidents once in a while, efficiency and safety and productivity are the three key tenants which robots provide you.

So I sound like an advocate for robots here, but they actually can handle tasks that are dangerous and difficult for humans. Most of the time, like mining is a great example Neil has brought up, and they operate in an environment which are very, humans cannot work or they cannot work around the clock without fatigue.

And they can provide consistency in the task, which is really important when it comes to surgery and other things, the precision accuracy. So I see that customers, of course, when they realize that, and but pharmaceutical customers especially we work with, they see the high value in that and that point, it comes to how you can accurately do the repeatability and also have the precision, the recall, the accuracy and be able to deploy those without having like the issues associated with, typical issues associated with deployment. And you can actually do that repeatable things for that. And it essentially is really important for their business.

- [Neil] I've seen a lot of industries, you mentioned agriculture, we talked a little bit about mining.

I see a lot like in law enforcement, like you see the police robots in Singapore and Dubai. We heard a little about healthcare and maybe elder care. What do you think the next big wave will be for like robotics adoption? What kind of repetitive tasks?

- [Adnan] The way I see it is the maturity. Right now, there's a lot of things which are happening. Think of it as like having BERT in 2018 and GPT-4 or what we have right now. That's the big gap between.

We still have the transformer architecture. Then we came up with attention. Then we come up with more sophisticated models where we have bidirectional attention, and then we ended up having what we have right now in GPT-4 and beyond. So I feel like that next five years, if you have to predict the lineage of where robotics is going to go is going to become more mature, sophisticated, and human friendly.

And also from an ethical AI perspective, from a transparency perspective, from a more of a safeguard perspective will be start to get built into those things, right? It didn't follow ethical guidelines and safety constraints. Use cases you brought up have some ethical issues. Policing, also the privacy concerns associated with monitoring.

Those have not been completely resolved. The regulations are also falling behind in that. If you are recording me, what are the ramifications of that, especially in the era we live in with generative AI and deep fakes and being able to generate replicas of me. What are the things which are possible when these kind of constant surveillance operatives are going on.

So those are the challenges I think in the long term. I don't see them becoming, having that jump off, like doing completely something different, but rather improving to have robustness and reliability. And those should be, of course, prioritized. Like you are working with AI for good no more than anybody that ethical AI, ethical constraints on robotics, ensuring that AI robots can handle these unexpected situations without causing harm.

Privacy measures are in place to design for humans and have the human oversight in there. Like human in the loop is a part of that ecosystem, the transparency and explainability. So because they are AI driven bots, I think that's really critical for them to be able to interact with humans in a human way.

- [Neil] I think really important. I'm glad you brought that up. And I know you referenced earlier Asimov's Three Laws. Are you thinking about your Adnan's three laws as well to add to robotics?

- [Adnan] I suppose I can come up with some good laws. There are, there's a lot of work going on, as you know, in ethics of artificial intelligence, and Stanford AI Lab has done a lot of work in that.

Stanford AGI Institute has significant amount of work in that space. They constantly are pushing the boundaries around how we can bring in explainability and reliability and all, of course, I think acceptance of these robotics, both AI and robotics, both pretty much relies on people having trust in them, right?

So today, we trust our systems, so that's why we design it. And explainability is a big part of it, right? So Neil gets a diagnosis that I need to get you're left leg amputated, and you're like, why? And they, look algorithm said so. Like no, right? So that's really important that algorithm explains why that decision has to be made.

And doesn't, and it actually becomes much more important in robots because they interact with our environment. So if they are taking an action based on one of these decisions, then they have to be able to, they should be able to explain why does that happening. And as they become more prolific and become part of our environment, it's going to become more and more important that explainability factor.

And regulations currently are, as you know, are not there. We are still working on AI regulations. NIST came up with the the AI guidelines and EU has now new AI guidelines. I think at some point we will also see robotics specific guidelines, which like how they interact with human, what kind of information do they take from humans?

And what would be that protocol around that would be? We haven't seen that, but I think that's something which is required for society to get ready for AI robots.

- [Nikolai] Do you think to build AGI, you need robots? Like you need, AI needs to be in an embodied form. It needs to exist in the real world.

Or could it all be done virtually?

- [Adnan] So, I think I've said it earlier in an interview or a discussion that the way we are looking at the next token prediction, so that's what it is, right? The transformer networks right now are next token prediction. So next token prediction is not going to take us to AGI.

Next token prediction will probably get us close to that, but AGI or how we defined it as a an autonomous intelligence which can operate on multiple modalities, and we perform operations on a cognitive scale, instead of just predicting the next token. In order to do that, you have to have, like you said, an existence in an environment.

Of course that physical existence or being awareness of the environment, that I think is a very fundamental thing. I don't think it has to be physical, but the sensory inputs we have seen through the evolution, the sensory inputs and being able to interact with environment is one of the very important areas of an agent, right?

So the agent as a human or agent as an AI agent, you have to have that interactability in there. So, we don't have that at least yet, we established that. So that's a big gap in AI becoming AGI. And there are two other areas where AGI for the AI is lacking. One of them is, of course, the causality part of that, right?

So there's no causal inference at this point where we can say X happened due to Y. Now, I know that transformers and LLMs have the capability or have the sparks of AGI at some point of time, which induces some part of causality. But then again, it's not the causal inference from it's the way we define causality in a more mathematical sense.

It's still doing the next token prediction with world knowledge, of course. And these are, causality might be an emergent property. I wouldn't bet against that. But that's one of the fundamental properties. So having causality, causal inference, having the sensibility of the environment, being able to interact with the environment and some form of a master algorithm.

So I'm borrowing from one of them, The Master Algorithm book, I'm sure you probably have seen the book by Pedro Domingos, very interesting book, in which his idea is that there are multiple, there are five different tribes of artificial intelligence, that you have Bayesians, then you have neural networks, the connectionist, and they have the three other ones in there.

All of those have different algorithms and different methods and for example, representation learning, all of them have to have, borrow something together to be able to come up with this master algorithm which will take us to AGI. I think we are fragmented at this point. We are very heavily focused onto the connectionist model for the neural networks, but not necessarily talking about the causality, the interaction with the real world, also like things like knowledge graphs, the representation of knowledge with associations, those things are missing. And those things we as humans take for granted, right? We can build connections. We take a lot of dexterity from an interaction with environment as a, but AI doesn't have that luxury at this point.

So in order to actually get to that AGI, I think those things are fundamental.

- [Nikolai] Because robots are interacting physically in the world in real time. Are they pre trained to do that on like tons of examples or how does that work?

- [Adnan] Yeah. So there are various ways how robotics algorithms work. I work with MIT CSAIL as part of UST, and we interact with several researchers who work on robotics.

My work is directly not on robotics, just to be very clear. My work is more focused on the software related robots. But I do interact with the researchers who have worked in robotics, and I see a lot of I work happening in robotics, and I follow them. And if you look at the two major classification algorithms in this area, so you have predetermined, predefined tasks which robots can do, and those are very, they might have some sensory input where they can decide and make changes but within the threshold.

And there are algorithms like reinforcement learning algorithms, which are very much about how you interact with the world in a unknown manner, right? So and now those are the algorithms you will see are going to really take us to the next level. So, there's a robot sitting over there, and just he's running ROS that, ROS is the Robot Operating System, and all he's doing is trying to pick up this bottle. And sometimes it is coming close, sometimes it is not able to do it, and you're not teaching it exactly the parameters around this. You have given it a meta task, and it's essentially trying to perform that, and there's a cost function. There's a, the more, the better it performs, the less the loss function would be, and then the, it will get that reward, right?

So because it's interacting with this environment, that reward will increase, and then eventually when he is able to accomplish the task, then he will move to the next task. The biggest benefit of this kind of approach is the repeatability. What I mean by that is that now that he has learned that how I interact with the environment, if you give it a new model which doesn't look anything like this model, it's going to take what it has learned and being able to apply that. That's transfer learning. We can transfer this learning over to the other part. That's why you see the way Tesla Bots are created. The way Tesla works is like it has been trained on billions of hours of driving footage, right?

How people drive and then it transfers that knowledge across to the real world and then get trained on that. And of course it learns from millions of hours of being on the road as well but that transfer running really helps to interact with the environment because it has all the sensors, and it's looking at the world in a 3D view.

It's able to do really well.

- [Ryan] Let me ask you, when anytime people bring up robots, the conversation around jobs comes up too, is how robots are gonna replace humans, how robots are going to potentially take jobs. But obviously there's the other side of it is what robots can do can sometimes counteract that and create new jobs and create new desired skillsets, things like that.

So how do you all, how do you think about that?

How do you approach it when it gets brought up, if it does get brought up at all in your conversations and what do what do you think our listeners should be or how should our listeners think about that?

- [Adnan] So I would say I am bullish on humans. That cognitive level of intelligence and being able to communicate the ideas in a manner, of being able to think creatively is a uniquely human skill.

Having said that, of course, there are challenges for repeatable tasks, the tasks which are more mundane, more easily automatable, and you will see that industries which will lead the way in robotics adoption are like manufacturing, logistics, healthcare, agriculture, retail, and but you will see that jobs being performed, especially from a robotics perspective, will be repetitive, dangerous, physically demanding tasks which will potentially be a best fit for the AI robots, right? Now, society being ready and having human jobs being replaced, that's a fairly large topic and there is nuance to it. Like what will get changed and what will not get changed, there's a recent report out, I'll send out the link to you to add to the show notes, which has a coefficient of all the different jobs and which one is more easily replaceable by AI versus non replaceable by AI, and like you can imagine the physically demanding job, the dexterity task are fairly low on that scale.

And then the repeatable tasks like creative writing and other things are fairly high on that. So a human will definitely play a central role in this AI and robotics ecosystem, right? We will develop, and we'll deploy the AI robots, but at the same time, the new jobs will be created in areas like robot training, maintenance, programming, ethical oversights, right?

So we will also be needed for like the task requiring complex problem solving, creativity, like empathetic, I think bedside manners of a robot, I'm not sure how good those would be, right, comparing a doctor. Interpersonal skills, things that I think AI and robot are right now far from replicating, like maybe in science fiction, that's doable but again, this is the more exciting and challenging area where there's a lot that needs to be thought out through as to where we go with what jobs are going to get replaced. For example, like truck drivers, right? So that is something which you will see is a very demanding task. People have to be away from their families for a long period of time. There's a huge potential for accidents in that space. So automation of that would be would be a bad idea or not, that's for us to decide on based on the data. But at the same time, what are the alternatives for the people who are going to lose their job? But one thing I can tell you for sure is that human society is very resilient. And we have been able to find ways. This is not the first time digital revolution has happened, right?

We, in history, so many times it has happened. And we said tellers will lose all their jobs when ATMs come. And now we don't even see that point happening, and ATMs have completely changed that spectrum. Will there be job losses? Yes, absolutely. Will there be replacement jobs?

Yes, absolutely, right? Those will go upskill. It will definitely get the replacement jobs and we work with that.

- [Neil] I love the way you're thinking because I know the audience hears me talk about this a lot, but it's a good example of hybrid intelligence. That it's not human versus machine, it's human and machine working together.

That we can improve our own human capabilities with the things that machines are good at, their abilities. So they complement together. So I think with General Electric now, they have AI not only planning out their manufacturing schedules and guiding the workers, but there was a Japanese company, I can't remember what it's called suddenly, developed robots that can actually work by humans side by side on the line.

Historically, you couldn't do that because these are big heavy robot machines that if they hit the human, the human is going to die. But they've been able to teach them and recognize what humans are and take more care. They've actually been able to speed up production and improve quality. That's a great example of the opportunity, right?

Some jobs will go away, might need some less in certain roles, but it's going to open up, hopefully, Pandora's box of hope here on whole new types of jobs.

- [Adnan] That's a great example, and I can give you a few examples from a software robots perspective, where we're seeing augmented intelligence. And this can equally be applied with robots.

Like augmented intelligence is a key part. Whenever you call your insurance company or a healthcare payer, they will have to provide information about your what's your co pay, what's that information. Nobody should be able to have to memorize all this information and nobody that look up whatever time it takes is going to eat into your efficiency.

Why can't a bot listen to that conversation, get you the information which is requested, and then showcase so the human person, instead of searching through a plethora of documents, can do the empathy, can empathize with the person, can provide solutions for things instead of going through the contract and trying to search stuff from that, which is a very mundane and robotic task.

Yeah. So that's something which I think are very useful for humans and not as much useful use of time for humans as compared to machines.

- [Neil] I think it's a guarded interest is the best way to describe it, that I think there's a little bit as like Adnan has referred to before, trust factor, that are these robots going to be able to perform these tasks with a certain amount of accuracy and reliability, right?

Especially if something unusual happens, how will the robot actually react to that? I think it's not stopping them obviously from exploring, but that's a big concern. I think the other concern is just the fallout that, you know, because people keep thinking that this is automation, so people have to lose jobs and they don't.

You're worried about kind of the perception of are you automating all these people out? This is a little different. This is a bot, not a robot, but McDonald's rolling out an AI bot to take your drive thru order, there was actually some backlash against that. Oh, you're taking jobs away from human workers or teenagers.

This is how they get their experience, make some money. And it's just what we've seen is this, people can customize their orders now, and it's more accurate. So they usually get what they want. We get less complaints. There's just the trade offs. That's the challenge I think that they're facing, the trust and the public perception.

- [Adnan] One of the, I think I would say what people are afraid of, or should be afraid of, not as much afraid of, we think about AI taking over and the Terminator thing, that, more than that, I think the real Terminator or the real scary part is the bias and the things which are automated in algorithms built in from data and from our societal biases, right?

For example, the real systems which actually make impact on people's lives, such as loan management systems, such as college admission systems, such as hospital admission system, or a triage system, they are all algorithms. We all know the parole system, how the AI based parole system was driven by that and how much bias it had against minorities.

And so this is the real danger from an AI perspective, the bias which is inherent in us. And we try to keep them in check. I know how well that goes, but in terms of algorithms, if once algorithm has a bias, the thing is like it perpetuates. So for instance, if somebody shows really racist tendencies, we know the society has a correction mechanism which will repeatedly like you can't be outright bigot in the society. You will keep your check. But for algorithm, if it has trained on that kind of data, it can, it will keep doing the same thing and regurgitate that information over and then it will, the systemic aspect of that will replicate in society.

So whether it's loan applications, whether it's parole, whether it's education, all of that will get impacted. And I think that's the real danger which we don't necessarily talk about. But there are, from a robotics perspective, the displacement of jobs, of course, there's human labor in certain industries is an issue, but privacy and security is of course huge, right?

Those systems, AI robots, which are connected to the internet, I don't want to make this a Black Mirror episode, but they can be hacked and misused, right? Those systems are at the risk of automation bias where humans are now over trusting the abilities of AI robots. Sometimes ChatGPT is saying things which are completely untrue, but people feel like if it's saying it, there might be some reality to that, right?

Just like that lawyer who ended up sending the citations generated by ChatGPT to the world. So there is that real risk of automation bias. Overtrusting the abilities of robots for critical errors, right? And of course, at the end, the philosophical concern, I'm sure Nikolai is going to get a kick out of it, of then one day machines will outsmart us.

I think we're quite far from this scenario, but maybe one day that will happen. In order to address that, and is the society really ready for that, I think it goes without saying that we need safeguards to be put in there, the ethical guidelines, the safety constraints, the guidelines, I think we went over that, but readiness is again a mixed bag.

Like people are now getting increasingly accustomed to AI and automation in a variety of different forms. And we as humans, and you know that better than I do, that we trade convenience, we trade privacy for convenience all the time, every day, right? So we always trade that because it's like now the AI automation coming in various forms, fully autonomous AI robots are, they will cause a significant societal shift.

And I think people need to get educated. We are not there yet. We need to get educated about AI, its implications, and as a society, we need to address the fears and misconceptions around that, right? Even now, if you look at one of those four legged robots, it reminds you of a Black Mirror episode or an X Files episode, like wow, what's going on?

But, and also the regulatory frameworks and societal norms are far away from we can accommodate these new technologies. So I think those are some of the challenges we need to address as a society. The onus will be on governments, institutions, the companies which are making those bots, regulatory bodies, as well as general public to be able to realize those challenges and address them as a part of a effort, a concise effort.

- [Ryan] Neil, any last questions for your end?

- [Neil] I can ask the super sci-fi culture question. I'll ask it to each one of you, how about that? We know the goal with robots is like the assistants, like the R2D2s and C3P0s, right?

Everyone's afraid that those robots will evolve into the evil robots like the Terminator. I'm of the mindset that the way things will really work out is the robots will wind up becoming like Bender from Futurama. What do you guys think? I'll ask each one of you.

What do you think that future of robots is going to be?

- [Ryan] I think it's more in line with what you're saying, as far as where the robot's going. I think as humans, we're going to narrow it down to the uses that we find most valuable and hopefully build and evolve to there. So I'm hoping it goes more the path that you're saying and less the Terminator path for sure.

- [Nikolai] I think there will be an attempt to make robot friends and robots with personalities, which could have side effects. I know some like some people think robots, people are going to start trying to give robots rights. If they get very friendly and like they're your, they're like your best friend or something then obviously there's going to be like a rights movement for robots, which I think is a misunderstanding of what robots are and what's going on there. But it would be, it would be, it wouldn't really be about the robots. It would be about humans. It's more, that scenario is more about how humans behave. They tend to anthropomorphize and stand up for other things that they empathize with.

It would be a very human display. It would really have nothing to do with the robots because robots don't care about rights. And if they were super intelligent, they wouldn't care about, it's like if ants came up to us and said, hey, humans, you have rights.

We just want you to know that. It's like we wouldn't care. You're ants. So, I think robots, super intelligent robots would be the same way.

- [Adnan] That, that was a very dense answer, like very comprehensive, very, there's so much packed in there. Like you brought in the the super intelligence reference in there and a paperclip example and a bunch of other things around like how humans will do so.

It will be hard to follow, but my take on that is essentially whenever there's an Orwellian ideas, 1984, or when we have Huxley's ideas, I think as humans, we usually tend to go more Huxley route than Orwell route. So in that case, it will be more making bots more of our augmented selves and how we can use them as distractions and entertainment and as more than our overlords.

- [Ryan] Adnan, for our audience who's listening to this, wants to learn more about background or what's going on at the company, follow up questions, that kind of good stuff, what's the best way they can do that?

- [Adnan] So they can reach me out at my Twitter or my blog or my email address, I'm sure you're going to post it.

Those are the best ways to reach out to me via LinkedIn. I'd be more than happy to take any questions, answer any questions you have around this area. I tend to post my opinions on Twitter a lot, so you can follow that and see how things are going in the world of artificial intelligence and machine learning.

And I think this, the arrival of AI robots, and this is a very fascinating area, I think we are living in really interesting times, and they are both exciting and challenging, right? And the key to approach that new frontier is with a balance of optimism and looking at the potential benefits and the caution for the associated risks.

So combining those together. So to do that, we need a comprehensive, informed, and inclusive discourse like the one we are having on the future of AI and robots. And thank you for inviting me, and this was a really good exchange of ideas. Shameless plug. One of my new book on responsible AI is coming out soon.

I've written it with one of my coworkers Heather Dawe. She is also an expert in responsible AI and also worked with NHS in this area. So if you are interested in that area of responsible AI, how to build AI governance in an enterprise, how you can put ethical guidelines around those, feel free to check it out.

- [Ryan] Appreciate your time. It was fantastic. Great topic we haven't had a chance to cover yet, so really enjoyed it for sure, and I think our audience is going to get a lot of value out of it as well.

- [Adnan] Thank you very much. Thank you, Neil. Nice meeting you guys.