Google's Bard AI Just Got a Major Upgrade: Outsmarts OpenAI's ChatGPT-4 !?

Google has just rolled out a major update for its AI chatbot, Bard, and it is getting smarter with logic and mathematics. Bard can now answer complex questions involving calculations, formulas and even code, which is pretty awesome. In this video, I'm going to explain what this update means, how it works, and how it compares to other AI chatbots out there. Alright, let's get started. But first let me give you a quick recap of what Bard is and why it matters. Bard is an AI experiment by Google that lets you collaborate with generative AI. It was launched in early 2023 as a way to boost your productivity, accelerate your ideas and fuel your curiosity. You can use Bard to ask questions, get tips, brainstorm ideas, outline blog posts, write cover letters, and much more. It is powered by a large language model, LLM, Specifically a lightweight and optimized version of Lamda, which stands for Language Model for Dialog Applications. Lamda is Google's breakthrough technology that that can generate natural and coherent conversations on any topic. Bard was initially available only in the US and the UK, but it has since expanded to more than 180 countries and languages. So what's new with this update? Well, Google has announced that Bard has migrated to a new and improved language model called Palm 2, which stands for Programming Aware Language Model 2. And it is a game changer for AI chatbots. Why? Because it gives Bard enhanced logical reasoning skills and the ability to answer mathematical and coding questions. This means that Bard can now solve problems that require computation, formulas or code generation. For example, you can ask Bard to calculate the area of a circle with a radius of 5 cm. Or to write a Python function that prints for example hello world. Or to explain the difference between a list and a tuple in Python. And Bard will not only give you the answer, but also show you how it got there. Honestly, I think this is one of the most impressive features of Bard so far. It makes Bard more powerful in context of learning, teaching and creating. Imagine how useful this could be for students who need help with their homework, or teachers who want to create engaging lessons, or developers who want to prototype new ideas. Basically, you can use Bard as your personal tutor, assistant or collaborator. But how does it know how to solve these problems? Well, this is where Palm 2 comes in. Palm 2 is a novel language model that combines the functionalities of traditional LLMs and code based computations. Traditional LLMs are great at predicting text based on probabilities and patterns learned from large amounts of data can generate fluent and coherent sentences on any topic. They lack deep reasoning and formulaic problem solving skills. Often struggle with questions that require logic or math. Code based computation models are the opposite. They can execute code and perform calculations accurately. They are not very good at generating natural language. They often produce dry and robotic responses that are hard to understand or follow. Palm 2 solves this problem by integrating both types of models into one unified framework. It uses a hybrid encoder decoder architecture that can process both natural language and code tokens simultaneously. It also uses a dual attention mechanism that can attend to both textual and computational contexts to separately or jointly. This allows Palm2 to generate responses that are both informative and conversational. One of the coolest features of Palm 2 is that it can perform implicit code execution. This means that it can detect when a prompt requires computation or code generation without explicitly being told. So for example, if you ask bard what is 2 2? It will automatically infer that this is a mathematical question and and execute the calculation accordingly. Similarly, if you ask Bard how do I print hello World in Python? It will automatically infer that this is a coding question and generate the appropriate code accordingly. Another cool feature of Palm 2 is that it can respond to questions in a tabular format and export that to Google Sheets. For example, if you ask Bard what are the top 10 countries by population? It will generate a table with the country names and their population numbers and give you the option to export that to Google Sheets. This is super handy for data analysis and visualization. So how does this new approach to solving logical and mathematical questions compare to other AI chatbots out there? Well, to be honest, it blows most of them out of the water. While Chat GPT 3.5 struggles with basic math, not to mention coding or intricate mathematics, chat GPT4 based on GPT4 has recently improved, aligning more closely with Bard's mathematical and coding competencies. Although it's a close call, I'd argue that with this latest update, Bard still has a slight edge over ChatGPT4 in terms of math reasoning. But let's save that debate for another day. Google has reported that Bard has improved its accuracy by up to 30% for computation and math problems in their internal data sets. They have also shared some examples of Bard's responses to these questions on their website and blog, and they are pretty impressive. You can check them out for yourself and see how Bard handles these problems. Of course, Bard is still far from perfect. Google has also acknowledged that Bard may still have some flaws in its code, or may not generate any code at all in some cases. They have also warned that it may still produce inaccurate, misleading or false information while presenting it confidently. This is because Bard learns from a wide range of information that reflects real world biases and stereotypes. Sometimes those show up in its outputs. So you should always double check Bard's responses and use your own judgment before relying on them. Despite these limitations, it is still an amazing AI chatbot that has a lot of potential. Google has stated that their ambition is to make Bard a leader in the AI assisted search space. They want to make Bard more helpful, more reliable and more conversational over time. They also want to get feedback from more people who use Bard so they can improve it further. Alright now. Meta has also announced that they are developing new chatbots for their popular messaging apps messenger and WhatsApp. These chatbots are designed to help users with various tasks and services such as booking flights, ordering food or getting news updates. They are powered by a new language model called Blimp, which stands for Bilingual Language Model for Personalization. Blimp is a multilingual model that can generate responses in different languages and dialects and also adapt to the user's preferences and personality. For example, if you prefer to use slang or emojis in your messages, Blimp will do the same. If you like to be formal or polite, it will respect that. If you have a specific interest or hobby, it will try to relate to that. Meta claims that Blimp can create more natural and engaging conversations than other chatbots, and that it can also handle complex requests and follow ups. For instance, if you ask Blimp to book a flight for you, it will not only find the best options for you, but also ask you about your preferred seat, luggage, meal and payment method. It will also send you a confirmation email and a reminder before your departure. Meta plans to launch these chatbots in early 2024 and they are currently looking for beta testers who want to try them out. If you're interested, you can sign up on their website. We're still a bit in the dark about the whole thing, but as always I'll share any updates as soon as they're out. And honestly, I think these chatbots sound pretty cool and useful, but I wonder how they compare to Bard or ChatGPT. What do you think? Let me know in the comments below. I hope you found this video enjoyable. If you did, feel free to hit that like button and consider subscribing to my channel for more content like this. Thank you for watching and I'll see you next time.