Open-Source AI: Benefits, Risks, and How to Start (Full Transcript)

[00:00:03] Speaker 1: Hello, everyone. Today, we'll be speaking about an exciting topic, and that is open source. AI models that are open source. What is open source, and why should it be relevant to you? My name is Anne-Marie Brown, and I'm joined by Dr. Philip Addo. So wherever you're joining from, just put in the chat where you are based, so we have an idea of the geographical representation. So, Dr. Addo. Yes. Yes. Are you hearing me loud and clear?

[00:00:44] Speaker 2: Oh, okay. Yeah, I can hear you now. Okay. So what was your question? I was getting echoing, so I have to turn the other one down.

[00:00:54] Speaker 1: Yes. So the first question out the gate, what is open source? When we say an AI, artificial intelligence model is open source, what do we mean by that?

[00:01:08] Speaker 2: So open source means that the basic kind of the code that forms the model is made available for people to download, and then you can use it on your computer without even if you don't have an internet, or you are disconnected, you can still use the model. You can also use it in the cloud, but most of the time, the essence of having control to manipulate the model to retrain it to meet your goal without you paying anything, right, because it's publicly available, all the information that you need, all the codes that you need to make it run well, and also train it, again, if you want to train it to customise it to meet a specific goal, you will be able to do that. As I said, you just need your computer, and then you download. It's just like downloading a software onto your computer and running it locally, right? So that's more about open source. It's available for people to just use it if they want to use it.

[00:02:16] Speaker 1: Okay, so thank you for that explanation. So I'm just welcoming persons who are placed in the chat where they are. Malahat, I think, is in the DC area in the USA, and Gerardo is joining us from Montreal in Canada. So welcome. Feel free, others, to put in the chat where you're based. So I've just heard a definition of open source. So open source doesn't necessarily mean free alone, right?

[00:02:47] Speaker 2: Yes. It doesn't mean it's free. It's free in a sense that you can easily download it on your computer without paying anything, right? You can go to the software platforms that I'm going to share with you, about four of them, that you can go there and download it and then use it locally. It's not free in terms of when you want to run it. You have to use electricity, right? But if the model is huge, then you have to think about getting access to online space or in a cloud where you can run the model, right? So instead of running it locally, you can buy a space, right, online where you can run the system there without using your computer. Because some of the models, they are so huge that the computer that we have doesn't have the compute power to run it. So you have to think about what kind of space that you can run online, and then you can get all the information. But still it's localized because no one can go and retrieve your information, right? Except the company that you are hosting the information. Then you have to pay for the hosting, right? So that's the only payment. But in terms of cost, it's cost effective compared to the paid version where you have to do monthly payment of using the model, right?

[00:04:20] Speaker 1: Very nicely said. And I see Ak Raka saying in the chat as well that open source means it can be modified and customized when it's open source. So if I can, I get it. If I can modify it and I can customize it, I get the appeal. But why would I download, go through all that trouble to download and customize and modify if you have AI models that's already there and I can get everything I want? So what is the advantage of using open source? I guess that's my question.

[00:04:55] Speaker 2: Yeah, I think one thing that is going on is control, right? You want to have control over your data that you want to feed into the system, right? So another thing is experimentation, right? You can experiment it. You can, as one of our viewers, attendees is saying that you can manipulate it. You can train it again with your own data so that it can behave the way that you want it to behave. So that kind of control that you can have if you can run it locally is the best, right? And also cost. If you really want to use maybe a, maybe Chachipiti for your organization, it might cost a lot, right? Imagine that you have your own model, that you have it locally that your organization can use and being customized. And you are not worried about policy change because the other ones are not open source. They can change their policies, right? They can change that, okay, now we are not using GPT-3 anymore. We're going to use GPT-4. So everybody, let's move you to GPT. So continuously changing without you being informed about a change. You just have to use their service. I think it's some of the concern that people have. So there are a lot of benefits of using open source.

[00:06:22] Speaker 1: All right. So there's the benefits. As you say, you are not victim subjected to the policy changes, right? You have your own thing on your computer downloading. But what are some of the challenges? Because I don't think everything that glitters maybe is gold. So what is the drawback?

[00:06:39] Speaker 2: So I use this analogy as like you want to cook your own food and somebody cooking it for you, right? So for Chachipiti, a model like open AI models, right? It's like somebody cooking the food for you and you consuming the food, right? So the challenge is, you know, cooking your own food takes time. You have to know the skill of cooking. So the same thing as an open model where you should have a basic understanding about the different kinds of model. What does it mean to have 70 billion parameters? What does it mean by having 408 billion parameters, which is the Lama 3, which is, yes, developed by meta AI. So you need to have some basic skill. But I always say the same thing as cooking, right? If you don't know how to cook, you have to start small, learning a little bit about the ingredients, about how to, you know, buy the product, the ingredients and then put them together. You know, learning is by watching. So what I do is that for me, for example, I don't have expertise or background in computer science or program development or something like that, but I learn. I go to YouTube and learn more about it. I can use one website that I always use is Perplexity AI, right? So Perplexity AI is so good because if you want to do some personal research, right? You want to know something, the system will give you that information and provide you a source so that you can click on it and learn more. So these are the places that, you know, I use to learn. Yes, you have to give, get a basic understanding before you can use it.

[00:08:31] Speaker 1: I'm very happy you said that because when you were speaking about customizing, downloading and modifying, I'm thinking, okay, and you use the analogy of cooking. I'm thinking I have to be a master chef. But based on what you say that, no, you don't have to know coding or be a master chef. You can use Perplexity.ai and YouTube and familiarize yourself, right? So practically now, how do you get started? You're new to open source AI. You are convinced that it will help you. What are your practical tips on to get started? Where do you even find, where do you access these models?

[00:09:14] Speaker 2: Yes. So thank you for your question. I will use mine, for example, right? So for me, you know, I start by going to a website and trying them out and learning more about them, right? Because there's no perfect AI tool, right? What do you have to think about? What problem do I want to solve? And let me try to ask the system questions and see whether I'll get what I want. And then if I'm getting good response, sometimes somebody suggests that I'd ask the system about 10 questions, right? And then we are getting most of them right, then the tool can be useful for you, maybe for your organization, right? So let me quickly show my screen and then give you information about some of the tools that I have explored and so that you can learn more about it. So let me go to my PowerPoint. And I will share the PowerPoint link for you to get all the information that you need. So the place that I will recommend you go in is Olama, right? So Olama is a platform where it has almost, it has a lot of popular open source model, right? And also there's a link down here that you can also listen to instructions, how to download, how to use it, how to prompt the system, right? So it gives you step-by-step. So Olama is a platform where you can get access to some of the open source model. You can also go to, I have difficulty pronouncing it, is it Gago, right? So Gago too has open sources model that you can explore. You can also go to LM Studio. You can also download their platform and then you'll be able to choose the one, the model that you want to use. The one that I like is the last one, which is called Hacking Face. Hacking Face, this is where the place that I started, right? So when you go to Hacking Face and you log on and you go to Hacking Chat, you'll be able to create your own kind of custom, it's just like custom GPT, right? But this one, you are using the open sources. So when you go to, I go to my accounts and I click on Assistant, you can see, let me see, you can see that I've created a lot of AI tools that is helping people. The first one is Research Plan Assistant. So if you want to create, you can just go here and create an assistant. And then what happened is that, let me go to Settings and Edit, and I just want to show you what the system wants you to, is they want you to give a name, right? So you see how I've given the system a name or the model that I want to create to help people. It's called Research Plan Assistant. Then I give a description. Then this is where you choose the open source model that you think will be helpful for you, right? So when you click here, you can see almost the popular ones, the MetaLlama, right? It has 70 billion, 70 billion parameters. And the second one has 405 billion parameters. So you just have to choose one of them, right? And then you can give, we call it custom instructions, right? Or system instructions or system prompts. This is where you give information to the system, what you want the system to do for you, right? Or what tasks you want the system to complete. Then when you finish and click on Save, you get your own custom tool that you can use. Now I can use it on their system. I don't have to download this software, or I don't have to download the tool, right? I just created an assistant, which is powered by the Llama3, which is an open source model that can be used to help people. So if you want to know about your research plan and give the system some information, you'll be able to get good results, right? And then so this is where you can start. So you don't have to download the model, per se. You can use one of the platform, like HackingChat, and then create your own and test it out and see what you're going to get, right? So that's another way of learning about it. But there are other ways, because of time, based on your question that you're going to ask me, I can provide more information.

[00:13:47] Speaker 1: Okay, all right, great. So I'm looking in the chat, Maria, you are asking the name of the website that we use is Perplexity.

[00:14:01] Speaker 2: Oh, yes. So let me type it here. So Perplexity.

[00:14:04] Speaker 1: Perplexity.ai.

[00:14:06] Speaker 2: Yeah, Perplexity.ai. So the good thing is that, you know, if you are a researcher, you can, you know, go to focus, and then choose academics. And then when you ask the system questions, you're going to give you some articles that are related to the answers that the system is going to give you. Sometimes you can give it, you don't have to choose a focus, you just ask the system a question, and you'll be able to get good information, right, from that. So that's what I use to learn about AI tools, and also use YouTube to learn about it, right.

[00:14:38] Speaker 1: Okay, very, very interesting conversation. If you are just joining us, we're speaking about open source AI models, why it's beneficial, why it's worth your time. And you can be a part of the conversation. There's a QR code, there's a link that you can come on. Once you come on, we'll see your lovely face, right. But you can also use the chat, if you're feeling shy, to post any comments or questions that you might have for us, right. So there's a QR code, you can use it. So question for me now. ChatGPT, is that considered open source?

[00:15:24] Speaker 2: Oh, ChatGPT is not considered an open source, we call it a closed source or proprietary model, because we call it the base model, or the base code is not available for people to download on your computer and use it locally and manipulate it, right. This means that you are using it on their platform. So when you interact with ChatGPT system, they have their own kind of system where it runs that information on the background, and then they provide you the answer, right. So instead of running the information on your computer, the system, it runs in their system to get that information, right. The reason why it's not open source is that they spend a lot of money in coming up with this model, right. They spend a lot of billions of dollars to train the model. So if they make it open source, how are they going to get their money back, right. So because of that investment, you know, they made it that you have to pay for their service. So that's why they are not making it available, right.

[00:16:32] Speaker 1: Then that's interesting. So does it mean then you get what you pay for? Is it that maybe the paid AI is better than open source?

[00:16:42] Speaker 2: Can you ask your question again? I didn't hear the last part.

[00:16:45] Speaker 1: Okay. All right. I see Tex is on. Let's go. Oh, Tex is gone. Hi, Tex. Welcome. Hello.

[00:16:56] Speaker 2: How are you doing? Hi. Hi. Yeah.

[00:17:00] Speaker 3: I read your book. So I followed your site. It was very helpful in writing skill as a postgraduate.

[00:17:09] Speaker 2: Thank you so much for, you know, assessing my book and reading about it. I really appreciate it.

[00:17:14] Speaker 1: Where are you calling from, Tex? Where are you joining from? I'm in Nigeria for now. Nigeria for now. Welcome. Do you have a question or comment for Dr. Adu?

[00:17:25] Speaker 3: Yeah, it's not related to this, but something else, but something else. Maybe towards the end, after he finishes with the AI, I'll ask the question.

[00:17:36] Speaker 1: Okay. Oh, you can even email me, right?

[00:17:37] Speaker 2: I have, I'll put my email address there so that you can email me. Okay.

[00:17:41] Speaker 1: Yeah. I'll put my email address there so that you can email me.

[00:17:44] Speaker 2: Oh, you can even email me, right? I have, I'll put my email address there so that you can, you can have a conversation. Okay. No problem. That's fantastic. Thank you very much.

[00:17:53] Speaker 1: Thanks for coming on, Tex. Yes. Thank you very much. So Diane has a question. What's the difference between ChatGPT4 and Perplexity.ai?

[00:18:03] Speaker 2: Okay. So, so the ChatGPT is, we call it a general kind of model. You can ask any question and you can get good response from the system, depending on the kind of question that you ask. Perplexity is a little different. I think it's quite, it's mimicking Google, right? Where you search for information and you get a lot of responses. But what is different about Perplexity is that when you search for information, it only gives you relevant responses, right? So you don't have to, you know, Google, you have to go through all these lists of things to find what you want. Perplexity will give you about three summarized things for you, right? It gives you about, about four or five sources and summarize, giving a summary so that when you read a summary, you can click on the link and go and learn more. So that's, in terms of learning about things very fast, Perplexity, you know, it doesn't waste a lot of your time, right? Compared to using Google, right? So Perplexity is more about looking for information. The system gets the information and summarized for you and provided a source so that you can get more. If you want to learn more, you can click on the link.

[00:19:14] Speaker 1: I totally agree. I find Perplexity.ai useful if you're doing like a desk review or a literature review, because it gives you the source and where to get it. ChatGPT doesn't give you the source. And sometimes you ask for the source and the source does not exist. Right. I see. Now we have Paul. Hi, Paul. Welcome. Where are you calling in from? Hi, Paul. Can you hear me? Hello, Paul. All right. So while Paul gets settled, we move to Seth's question. Seth asks, is ChatGPT or other platforms GPDR compliant? So that has to do with the data protection policies in Europe. And these governments have said they can be used for data collection and storage. So Dr. Adu, I know you're based in the U.S. Yes.

[00:20:15] Speaker 2: So it's difficult to say. But what you have to do, if you can access the platform in your country, then it's compliant, right? Some of the platforms, one of them is like Cloud AI. Sometimes based on where you are located, you cannot access it. This means that maybe it's not compliant with the country that you are located. So if you can access it online, there's higher probability that it's in line. But I think to be on a safer side, focus on the big companies, one, ChatGPT one. You can focus on perplexity. You can focus on Cloud AI, Julius AI. The big ones will be good for you in order to make sure that you are going in line with the ethics and also the policies in your country.

[00:21:10] Speaker 1: Great. Thank you, Seth, for your question. Keep the questions coming, please, in the chat. And I see James. Can you hear me? James? Okay.

[00:21:29] Speaker 2: Okay. So let's try Paul again. Paul?

[00:21:39] Speaker 1: All right. Maybe Paul is having some technical difficulties. All right. So, Dr. Adu, to follow up earlier, you said that ChatGPT is not open source. As a matter of fact, they have invested so much money to develop their platform and invest in their platform. So that got me thinking, you know, they say that nothing in life is for free, right? And if they invested all this money to develop their platform, does it mean that these paid AI models, are they better than open source AI models?

[00:22:17] Speaker 2: Yes. Based on the assessment that experts have done, you know, most of the time, the paid versions perform better, right? Because they have so much resources. They have invested so much money. Especially open AI, you know, they have invested a lot of money into it. Google has invested a lot. And also, Facebook has also invested. But I think that although sometimes it's very difficult, you know, I am a person where I want both to strive, both to do well, right? Because what I've realized is that if we allow only the closed source companies to work well, then they will have only control over AI system. It's not democratized anymore. So this means that if you don't have money to assess your system, you are left out. So that's why I like the initiative of Meta, Facebook. They have invested so much money in open source because they want everybody to get access to technology. They want everybody to get access to this, you know, AI tools and then you can customize it to solve your specific problem. And that's why I recommend that, you know, especially people who are in the developing country, they really have to really focus more on the open source so that they can customize their information to solve the specific problem that they face in the organization, in the way that their communities, right? So I want both to work well. The closed source should work well and then the open source should work because the open source is helping us to get closer to artificial general intelligence, right? So yes, I want both of them to do well.

[00:24:28] Speaker 1: Okay, so you're not at a disadvantage if you use open source for your research?

[00:24:36] Speaker 2: The only disadvantage is that you are not going to get higher quality information compared to the closed source, but if you have time and resources, you can train it to work well, especially people who are very concerned about the companies using your data to train your system, right? Let's say you have a sensitive data, right? And your company sensitive data, you don't want it to go outside. You don't trust the outside companies. Why don't you have your own open source tool within your company and run it locally so that your information will not go out there, right? So there's always going to be the benefits and the strengths, and I think it all depends on what tasks that you are doing and you decide which one will be the best one. But I like exploring the open source, but for what I do day by day, I use closed source, right? But I explore open source just to, you know, learn more about it because this is what is going to happen. We will reach a stage where we're going to have independent agents, AI agents, where we can ask, you know, you just ask the agent to do something for you and it would complete the goal and then come back to you and provide you the results, right? So getting knowledge about how the AI tools work is very important. And how do you know how it works? Learning about open source, learning about the tools that are available, learning about how it works because the open source, they give you all, they are so transparent. They give you all the information. Sometimes they give information about the data they use to train the system, right? So this one will equip you so that as we are moving on from having a customized tool to agents, AI agents, you'll be well equipped to learn more about agents and how to use them for your company, how to use them personally. So it doesn't hurt to explore and learn about how the open source also can be used to solve specific problems.

[00:26:50] Speaker 1: Okay. All right. Thank you so very much, right? And for highlighting the benefits. If you're worried about your data being used and stored by an external party and you don't know where it goes, then open source has a big advantage there. Marvellous. Marvellous. Can you hear us? Okay. Marvellous is gone. James? Oh, okay.

[00:27:23] Speaker 2: Hello, James. Hello, Marvellous.

[00:27:38] Speaker 1: All right. Let's take a question while we see if we can connect Marvellous. So there's a question on ethics. And that is how do open source AI models address the issues of bias, especially in research settings where open security is critical? And what steps can researchers take to minimize bias when using AI tools in their work?

[00:28:04] Speaker 2: Yeah. So in terms of bias, I think last time I talked about it, I think look at AI tools. They are trying to mimic our behavior, right? The information that we give to the system, if it's biased, then the system going to bias. We always say that bag it in, bag it out, right? So in terms of AI tools, there's a possibility that it will have bias because we've been trained by data science. So if you're trained by data science, data available, what we have to do is whenever we are interacting with a system, let a system know that you are aware that they have some biases, right? It's the same thing as when you're doing a qualitative study where we have our own background, knowledge, and preconceived ideas that may impact what you are studying. You do some we call it bracketing, right? Reflecting on your past experience and your biases and putting them aside so that you do not overly impact what you are studying. What about letting the system to be conscious about their biases that they have, right? So you can say that I'm doing this research in this and this and that. What are the potential biases that you have? Can you set these biases aside and try to look at it in a very objective way so that I can get good information from you? Letting the person in the system understand that you are aware of a bias is a way of doing the research about it. In terms of open source, open source has also its own bias because it's based on the information that is being used to train. But because you have an option to fine tune it, you can do training the system, right? You can train it to do a specific task. You'll be able to reduce the bias by, you know, doing the training. So I think having the control over, you know, training the system can also reduce bias.

[00:29:57] Speaker 1: All right, and then there's an interesting follow-up question. The person asks, can you share an example where you encountered bias in AI outputs and how did you address it?

[00:30:11] Speaker 2: It's difficult to come up with right now, but I think when I was doing, you know, when I think, you know, Anne, you have an example, but let me finish with this one and see whether you can, gender example. So what do you have to always do is that, so I was analyzing the data and I asked the system to summarize the findings for me, right? Or summarize the data for me. You know, the summary only gives recurring information, but what about people that provided maybe few information that are unique? The summary doesn't have that information. So what I always do is that I say, I tell the system, can you reflect on what you have provided to me and see whether there's any other unique information that you did not talk about? And can you bring me that information? So helping a system to get, get engaged in a system in self-reflection will help you to get to solve the problem of bias. So there's a possibility of bias. I think always one thing that you have to be doing, do is that you always have to be skeptical. Open AI, they always say that their system has the potential to provide you wrong information, right? So you always don't overly depend on the information that is provided. Always ask questions. You can also use another AI tool to compare and see what you're going to get, especially when you are doing some kind of research. Perplexity AI can also provide you a source to help you. So these are the things that I do personally to help me to overcome those biases that I find in my studies.

[00:31:58] Speaker 1: Okay. Great. I see, I see marvelous. Can we try and get marvelous on again? Can you hear us marvelous? All right.

[00:32:09] Speaker 2: So let's try James too.

[00:32:16] Speaker 1: James. All right.

[00:32:22] Speaker 2: Okay. So somebody was asking me about my book, right? So I just want to quickly show. So this is, so this, if you want to know about, you know, how to analyze your qualitative data, this book is going to be very helpful. Step-by-step guide to qualitative coding. You can get from Amazon. You can go there and get information. So this is a great book. You can go there and get information. Another one is if you are working on your dissertation and you want to know more about the method that you have to use and learn about chapter one to five from the introduction to discussion, this book is going to be helpful for you. Last one is phenomenological study, right? If you are doing a phenomenological study and you want to get background information about the terms and how to apply it in your study, this book is going to be very helpful for you. The theoretical framework in phenomenological research. So these are the books that I have.

[00:33:22] Speaker 1: Great. So you know all about the books and Dr. Adu's YouTube channel, lots of useful information there as well. So remember to just click like and the follow button. And so this way you're notified every time we're having these type of live streams. So we are wrapping up a very interesting conversation on open source AI and how is that different from AI that you basically have to pay for. So keep your questions coming in. We'll try and get one more before we close. So I have one final question from my side, Dr. Adu, and that is how can someone fit open source AI tools in what he or she's already doing?

[00:34:18] Speaker 2: I think it gets back to what problem that you want to solve, right? At the beginning, when ChatGPT came, you were able to develop your own custom GPT. But at the beginning, people cannot, those who don't have a paid version, cannot access a custom GPT. So I took, I said, okay, can I create another one, the same one, in open source so that I can help people? That's why I went to HackingFace to create some of them to help people. So I took that. It's all about, you know, looking at the problem that you want to solve and seeing where that open source is the best for you. If you think that open source is not the best, you don't have to waste a lot of time. But I think having a background information will be helpful. Because imagine that you are a head of an organization and they want to introduce an AI model, right? What kind of questions are you going to ask the person who wants to use that AI model? And who wants to develop the model for you? Or introduce a model? If you don't have background information, you might not have right questions. Like, what problem do you want to solve, right? And what is the parameters of the AI tool that you want to use? Is it 70 billion? Is it 450 billion? Is it 1 trillion parameters? You have to know all these terms and then know what they stand for so you can get rich information about people who are going to help you. So, at this time, what you have to do, I always say that you have to continuously learn about AI. Don't stop learning. Going to YouTube, seeing what is going on, seeing how it can be applied to your study, what you are doing will be the best. And sometimes if you think that based on the cost or control, you think that open source will be the best, you can try some of them. I can quickly, I think before I quickly go to the PowerPoint about things that you have to know before we go, somebody is waiting. Das is waiting. Das, can you hear us? I think we cannot. So, yeah, you know, they can keep the question coming, but let me quickly, let me see if I can quickly go to the PowerPoint. I will make the PowerPoint available when you look at the chat box after our conversation. So, let me quickly go through so that you get some kind of big picture, right? So, in terms of in terms of the open source platforms that you can use, the models that are available, there are thousands of open source models, but the ones that are popular and the ones that, you know, I trust the sources are these three ones, right? This one was made by Meta, Facebook, right? As I said, they spent a lot of money in making this available. You can go to their website. There's a link here. When you get a PowerPoint, then go there and learn more about them, right? This one is also produced by Microsoft, right? It's called PHY3. They have different models under that. So, you just have to explore and see whether it's going to be useful for you. There's an instructional video under here that you can listen and see how whenever you are listening, think about what problem can this one solve, right? If you don't have any problem to solve, you can just learn it. You don't have to use it. The last one is Mestral AI, right? And it's also a powerful tool. You can explore. You can go to their website. You can also create your agent. I had the opportunity to go there and create this. It's so easy to use their system. You don't have to be an expert to create a custom agent or custom AI tool on their website. Then use it. Another thing, I think I've talked about this. These are the places that you can go to download a model or you can review the model and download and then use it on your computer. But you have to be very careful because some of the models are so huge that you cannot use it on your computer, right? Especially those that have about 405 billion parameters. If the model is around 7 billion parameters, you can use it. But always check. Because if you use a bigger model, it can shut down your computer, right? But it's also an option for you to use the model online. There are online platforms that you can use. You can try out the models, right? When you go to Meta AI, you can try some of their models, right? You can go to HackingChat. You can also try it out. And Dr.

[00:39:23] Speaker 1: Adu, when you say online, it means you don't have to download it to your computer.

[00:39:28] Speaker 2: Yes, you don't have to. Yes. Thank you for the question. You don't have to download it. You can use it. And I always say that never download it until you have tried it out, right? So just go to their website that I provided to you and then try it out and see whether it's going to solve your problem. Then it has the potential of solving your problem. Then you think about how should I download it? Should I download it on my computer or should I create it or use it on an online platform, right? So these are all the websites that you can try it out and explore, right? And then to, I think these are the things that I have for you. Unless anyone has questions, I'll be happy to address.

[00:40:18] Speaker 1: Okay. Well, I think we can try this and this is the last person if we can get Das in and then we close because we are right up on time. Das, can you hear?

[00:40:29] Speaker 4: Yes, I can hear you. So my question is, so actually using this open source AI, if we approach for some model and, or how can we do the mathematical abbreviation as well? And if we want to try one by one by ourselves to suppose for the image processing model or some video modeling. So do you think that using that open source AI is the good option better than trying by ourselves how our model is going to performing?

[00:41:06] Speaker 1: Yeah.

[00:41:06] Speaker 2: So no, I think that the first step is to try it out on their website, the platform that they have and access system, some questions that are related to your problem and see how the system responds. And based on that, you can see whether it's helpful. If it's helpful, then you go to one of the websites I provided to you to download and use it. And there is a little bit of technicality in terms of training the system, right? If you want to train it to really solve a specific problem for you, maybe you have to get some background training on how to use it. But I think it's all depends on what you want the system to do for you. And you just have to think about, do I have to train the system, the open source for you to solve this problem or I can use it as it is. And it's all depends on the communication you have to the system. Sometimes if you don't want to go through all these troubles right, and you have the finance to use maybe chat GPT or to use a cloud AI, you can use that, right? But I think that if you are concerned about a cost and you are concerned about chat GPT is too general that you cannot experiment, right? Then you may think about using this open source. I hope I've answered your question.

[00:42:44] Speaker 4: Yes. Thank you, Bruce.

[00:42:46] Speaker 2: You're welcome.

[00:42:47] Speaker 1: Thank you so very much, Das. We are right up on time, Dr. Adu. You have the last word, any key takeaways for us to close on?

[00:42:56] Speaker 2: Yes. Also, thank you for moderating. I really appreciate your time. And one thing that you always have to think about is that AI is not going anywhere. And so you just have to learn more and see how it's going to help you to solve your specific problem. You don't have to be overwhelmed. You don't have to use the open source if it's not needed. I think maybe if you are new to AI, maybe use the closed source first, right? And see how it works. But if you want to really know how the system works at a background and you want to learn more, there's an opportunity for you, right? Especially if you are concerned about your job. This is a good time for you to learn more and be ahead, right? As time goes on. And then teaching people as we are also doing, right? So that we all work together to accomplish a goal or a stage where AI can not take over the world, but do a lot of things for us.

[00:43:59] Speaker 1: Thank you so very much, Dr. Adu. You're welcome. So the key takeaway is invest in lifelong learning, learn about the models, and you stay ahead of your professional game. Thanks everyone joining us today. Thank you. And thank you, Dr. Adu. And hope to see you all next time. Take care.

[00:44:19] Speaker 2: Bye-bye.