Can We Overcome Deep Fake Technology?
While it may sound alarming, the reality is that deep fake technology poses a threat to society. Dr. Soni, a passionate professor at Florida International University, specializes in educating his students on machine learning, deep learning, and big data. With his expertise, he will delve into strategies for effectively addressing the challenges posed by deep fake technology. Dr. Soni holds the belief that individuals with positive energy should actively engage with technology, as he fears that it may otherwise end up in the hands of those with malicious intent. Let’s dive deep and learn more about Dr. Soni’s passion for teaching and knowledge in artificial intelligence and machine learning.
Question do you believe we can overcome deep fake technology?
Transcript with Dr. Soni
Speaker 1: And thank you again, Dr. Soni, for allowing me to interview you today. And can you go ahead and introduce yourself?
Speaker 2: So, you know, I’m Dr. Soni, I’m currently an adjunct professor here in Computer Engineering Department. I’m teaching machine learning deep learning big data. So that’s my research area. That’s in machine learning deep learning application of AI into multi-domain areas.
Speaker 1: Okay, that’s interesting. That’s cool. Okay. So, um, my, my first question is, what inspired you to pursue your career in teaching AI and machine learning?
Speaker 2: Very good question. So, first, I have done my PhD. So I’m very much from studying like curious and very much interested in teaching students. So I used to teach other subjects as well. But specifically, I like to teach machine learning and deep learning because that’s the current market. And those who are, who wants to join the market, those who wants to learn something new, or want to solve some problem. AI and machine learning is one of the top area where they can solve any problem where they have a defined objective. So that’s the reason I’m very much interested in teaching this type of courses.
Speaker 1: Okay, so with you, saying that it could solve problems. So how do you see the AI impacting industries like the healthcare, finance transportation in the near future?
Speaker 2: So I would say, it depends what objective somebody’s trying to solve. So let’s say in how to care for radiologist, they have a lot of X rays to scan. Now, it might be difficult for a radiologist to scan 100 of x rays or 100 of scans per day, since they have limited amount of time, so they can use AI to like pre scanned this type of X rays and get some details out of it. And then they can utilize those details to further evaluate a particular patient. So that’s how a radiologist can use this type of AI applications for their own objective. So that is for healthcare, we talk about transportation. Now for transportation as well. There are a lot lots of roads, lots of bridges, right? Sometimes we want to find optimized or optimal route. In that case, we can use AI, we can see the traffic, we can see that timings all this sort of information AI can provide for better utilization of time. Okay, but that’s for transportation.
Speaker 1: That’s nice explanation. Thank you so much. And, um, another question is, what has been the most challenging you face? While teaching artificial intelligence and machine learning?
Speaker 2: So teaching is my passion. So I always like to learn the new algorithms, new techniques that are coming in this area. So even in next to next five to 10 years, I will be in the teaching sector. But I can teach the students with the latest technology in AI, and how we can use AI to solve the current world problems, and not thinking AI as something that takes away human jobs. It’s not like that it’s like, helping humans in solving current problems.
Speaker 1: Okay, so can you elaborate more on that? Please?
Speaker 2: Okay, yeah, so for example, I gave you the example of radial audit, right? The filter type of failure largest is very, very unique, and they do not have much time to help, you know, all those scanning of all those x rays, they didn’t have much time, we can use AI, they can take, you know, they can take an help of AI application, so that those AI can automatically scan, get some data out of it. And then the radiologists, the human beings, can just take that data and further do the analysis on the patient. So this is where the AI is helping human element AI is taking the job of human.
Speaker 1: So you came perspective on that, because I wouldn’t ever I thought about that. But I thought like, you know, later in life, it will eventually take over. But now you gave me like a different perspective on that. So thank you for that. Another question, I would like to ask, what advice would you give someone that is interested in pursuing AI and machine learning.
Speaker 2: So advice I can do nowadays, anybody can learn AI or machine learning, keeping in the mind that they are passionate about it. Without passion, I would not recommend anyone to do any anything in their life. If they’re passionate to solve any problem. If they are passionate to use AI and machine learning, then definitely the first thing is passion for that. And the enthusiastic towards learning new techniques, new algorithms in different areas of machine learning and deep learning. So AI can be used anywhere even in this. I think you’re part of this media, right? Mass media.
Speaker 1: Yes, yes.
Speaker 2: So there are lots of AI applications that can help in different video editing, making different audio sounds. And then maybe increasing the resolution of images, clearing out the noise from the from the audio. So there’s lots of applications of AI in that such in that sector as well. Okay, so the only advice I could give you is the passion to solve real world problems.
Speaker 1: Okay. That’s interesting, interesting. Thank you for that. Be passionate that that is true, you have to be passionate about what you love to do about anything. Great advice. Thank you for that. Um, okay. Would you can you recommend any books or sources for anyone who’s interested in learning more about AI and machine learning?
Speaker 2: Definitely. So there are lots of books, lots of videos are available online. One of the books that I recommend everyone is hands on machine learning. It’s very popular book with and that book is backed by, I forget the name of the author. But the name of the author is. So it’s a Hands-On Machine Learning with Scikit-learn Keris and TensorFlow by Origin. It’s one of the most popular book I would not say most, but one of the popular book when somebody wants to start learning about machine learning.
Speaker 1: What is the book called again, hands on…
Speaker 2: Hands on machine learning. Okay. Before he will start to learn about machine learning. A particular person should have a knowledge of Python programming. Once they have knowledge of Python programming, they can take machine learning once they have knowledge of machine learning, then they go into deep learning algorithms.
Speaker 1: Okay, thank you so much for that. Thank you. Thank you. Another question is how do you stay up to date with the latest developments and research in AI?
Speaker 2: So one of the latest and the current trend going on in market is about generative AI, meaning those type of algorithms which can generate something new in this world, like generating some new videos, generating new image, videos generating new audio. And one of the popular application nowadays, if you will have heard is quite chunky PT. Which is, so chat. GPT is one of the most popular application right now in the market that is based on generative AI. So just trying to generate some sort of text based on the prompt that you provide to that application, such as radio AI is, is the current trend in in the whole AI arena.
Speaker 1: Oh, interesting. Okay. Okay, interesting. Okay, thank you. Thank you. I wouldn’t have known. Okay. That’s very interesting. Thank you for that.
Speaker 2: Yes. But now, as they say, right, with every new technology, there’s always some sort of problems comes. So with this new technology of generative AI. Now, there are a lot of fake images or fake videos or fake fake text going on on social media. And it’s getting very difficult to recognize those fake content. So that society side I will say disadvantage, but there are some other sort of algorithms that we can use to eradicate those type of fake contents, which has been developed by some of the users who wants to create some sort of fake things in this world.
Speaker 1: Interesting they meet Oh, yeah, that issue they did start making fake. Oh, wow.
Speaker 2: They might. So right now, it’s happening that some of they are some some of the developers those who are like a bad developers, they’re trying to create some video where a politicians saying something which the politicians have never said. So that creates a big impact on the human society.
Speaker 1: Oh, wow. I would Oh, that’s crazy. Wow, this is
Speaker 2: There’s a there’s a thing called a deep fake technology. They carry they create, create those fake videos. And that has been the one of the big disadvantage of this technology. But there are some ways where we can eradicate those. Those problems. Okay. Yeah, it was what’s the big research going on right now.
Speaker 1: Oh, wow. So are you are you nervous about that? Do you do you feel like you know, eventually, we overcome that or do you…
Speaker 2: We can definitely overcome that. Because as we create this technology to help humans, now, of course, there are some humans which are, which are bad, they miss, they take Miss, they misuse the technology for bad reasons. It doesn’t mean that we should not use the technology for good purpose. That’s why but once the technology has been developed, as soon as the technology is developed, we researchers also try to like build some sort of technologies that can eradicate towards bad misuse of the technologies being misused by different users. There are a lot of algorithms are being getting developed in the research area to solve such type of deep fake problems.
Speaker 1: Wow. That is, that’s very interesting. You just taught me something that I did not know that.
Speaker 2: Yeah.
Speaker 1: Thank you for that to Sony. Thank you definitely taught me something right now. Okay, I just have a couple more questions for you. And then I’ll let you go. Okay.
Speaker 2: Okay, no worries.
Speaker 1: So much. Okay. So another question is, how do you approach data collection and preprocessing from your machine learning models?
Speaker 2: That’s the basic steps of finding machine learning what you said right now. So in order to solve any machine learning or deep learning based problems, what we need is data. Without data. We can’t solve any problem using AI. We need two things. First, we need data and we need a huge computational power. With increasing in the big data, meaning when we have a huge amount of data, we also need a huge amount of computation power. And with this two things, machine learning and deep learning algorithms, learn the patterns from the day data To solve a particular problem, now that there are a few steps, first we collect the data. Now correct collecting the data depends on the objective that we’re trying to solve. So let’s say if we talk about we are working on some research on cyber attacks, we collect the data from networks, from the from computers from processes, so those type of data we can collect, to solve cybersecurity type of problems. So once we collected the data, which is the raw data, to process the data, so that we can ingest that data into machine learning algorithms, or deep learning algorithms, so based on the different algorithms that we want to optimize or evolve, we want to train we have different pre processing mechanisms.
Speaker 1: Okay. Okay, that’s interesting. Okay, yeah, that makes sense. Because I always how, okay, yeah, that definitely makes sense.
Speaker 2: But these are the different pre processing steps. And then we train the model, meaning the models trying to learn the patterns from the data. And once the model is trained, we need to optimize it. Now, I am not going into more details, but there are some parameters that we need to optimize. And it all depends on the object you that you’re trying to solve with the current data that you’re using. So with these things, there are different ways to optimize. So upon building the training the model, we optimize that we can tune it into a way that we can solve our problem.
Speaker 1: Oh, okay. Okay, that’s very interesting. So okay, so you because, you know, like, a lot of people look at technology, technology, like a bad way. And like not, yeah, now you’re telling me everything like, yeah, now I’m seeing a different perspective on technology. And …
Speaker 2: Yeah, but it all depends, I would say, on how a particular person is using the technology. Right? Bad news and the good use of technology, how we use depend on on individual.
Speaker 1: Right, that is very true. That is true. That is very, very true. Okay, so one more question for you, and then I’ll let you go. Okay. So if you weren’t a teaching, and machine, I mean, machine learning and AI Artificial, what would you be doing? I mean, intelligence, artificial arm, so sorry.
Speaker 2: Nine socket. So the problem that we are trying to solve using AI and machine learning is solving the cyber attacks problem. Because currently, in a previous era, when the countries used to have war, they used to have war using those big fire guns and all those big machines in the day, and they were having wars. But now, the trend is having volume in cyber, they have cyber attacks on each country. That’s that’s the cyber war going on between many countries. Right now. They’re trying to steal that data. Try to corrupt health system is one of the big problem in any country, especially in the US as well. So we are trying to develop something using AI that can solve both cyber attack problems.
Speaker 1: That’s why they’re pushing the cybersecurity jobs so much. Oh, okay. Okay, okay. Wow, Dr. Soni, you taught me so much today, in about 21 minutes that we spent together. And I thank you so much for coming out today and answering my questions.
Speaker 2: You’re welcome. I hope it helps.
Speaker 1: Thank you. Yes, yes. Thank you so much. And you have a wonderful day. Okay.
Speaker 2: You too. All right.
Speaker 1: Thank you too. Bye. Bye. Bye
