Google's 8 Hour AI Essentials Course In 15 Minutes

I took Google's AI Essentials course for you so here's the cliffnotes version to save you to 8 hours but it's not enough for you just to listen to me tell you stuff so I've also included a little assessment at the end of this video to help you remember everything that you learned because research shows that immediately reviewing information is the best way to retain that information and let's be honest if I didn't do this you're probably not going to review it yourself so I'm doing it for your own good all right let's go a portion of this video is sponsored by HubSpot first a quick overview of the structure of the course so the course has five modules the first one is Introduction to AI which is fundamental basically just definitions of things second module is maximizing productivity with AI tools it kind of shows you all the things that you could do with AI tools module three is prompt engineering which is my favorite module in this course I think it's one of the best sections on prompt engineering I've seen so far module four is how to use AI responsibly people tell you that you should be responsible you know you should look out for certain things when using AI but they don't exactly tell you what you should be careful about so this is a great introduction module and module five is staying ahead of the curve where you should be looking to keep up with AI developments all right first thing is defining artificial intelligence refers to computer programs that can complete cognitive tasks typically associated with human intelligence for example Google Maps can help you go to that restaurant that you want to go to or if you go on YouTube there's a recommendation system that is able to recommend the videos that it thinks that you would like to watch now ai is largely powered by Machine learning which is a computer program that can analyze data to make decisions or predictions for example a machine learning program could be a program that is able to determine if an apple is ripe or unripe to train this model you show lots of examples of ripe apples and unripe apples and is able to learn what are the characteristics that determine a ripe or unripe apple so next time you show an apple is never seen before it's able to give a prediction of whether it's ripe or unripe what's important to note here is that the quality of data that you're providing the program will Det determine how good it is at determining what is ripe or unripe this is just one simple example of a machine learning model but basically as there's more advancements in machine learning another subset of artificial intelligence became very popular which is generative AI the focus of this course generative AI is defined as AI that can generate new content such as text images or other media the most popular example is a generative AI model that's able to process text and also output text otherwise known as a large language model because it's Google the example they give us is Gemini you can directly interact with Gemini through text and is able to Output information for you can ask at things like help you write an email brainstorm different slogans read things for you and give summaries pretend to be your friend when you're lonely and don't have anyone to talk to and lots of other things too all right you with me just saved you an hour there this course is honestly pretty fluffy like they have so many instructors and they tell you know like about their lives and things like that which is very interesting but that was basically the first module next up is maximizing productivity with AI tools the way that you interact with most AI models and AI tools is through prompting which is text input that provides instructions to the AI model and how to generate output like when you're asking Gemini to write your email or something it's not hard to start prompting and interacting with an AI tool but the way in which you prompt and communicate with the AI can dramatically change the quality of the results that you're getting back which is why there's an entire module which we'll cover next that goes into the art of prompt engineering where how to design good prompts to get the results that you want but for now just realize that prompting is the way that you interact with an AI tool Google advocates for a human in the loop approach when interacting with AI meaning that you should be part of the process of interacting with AI for example if you're a new company that sells apparel and you want to come up with some slogans for your company a not human in Loop approach would be just to ask Gemini hey can you come up with a slogan and then just use that slogan while a human in the loop approach would be say to ask Gemini hey can you give me some suggestions for slogans for my new appar company and then like tweak it provide it with different types of details to finally come up with a list of slogans but you as the human gets the final say into which slogan you want to choose so for this example if you're using a not human in the loop approach and AI just came up with a terrible slogan it's honestly not that bad it might be a little embarrassing but you can imagine if you did something like that for something that's a lot more high stakes like blindly asking Gemini to assess your symptoms and diagnose you with a medical issue uh you can see why that would be a very big problem that is why the final decision making should still be done by a human cuz AI at least at this moment and in the foreseeable future is not perfect two specific common issues to pay attention to when you're using AI is its knowledge cut off and hallucinations an AI models knowledge cut off date refers to the fact that an AI is trained at a specific point in time which means that everything that happened after that date the AI would have no knowledge about so if you ask it something like what is the current breaking news right now it would not know the answer to that and can actually lead to the common issue which is hallucinating which is when an AI outputs something that is simply not true like for example if it doesn't know what the current breaking news is it could just make up something like rabbits are brainstorming the White House something like that now that is actually like a very ridiculous example and it's not that hard to figure out that it's not real but sometimes AI can tell you things that seem really believable but are just not true like a politician did something that they actually didn't do but that could you know end up hurting their reputation kids don't believe everything AI tells you so this cour is an AI Essentials course but if you want to dive a little bit deeper into how to incorporate AI into your workflows i' like to introduce you to how to use AI for data analytics resource by HubSpot it's pretty useful whether you're working Solo or as part of a team the guide covers how to integrate AI into your workflows benefits and challenges and an overview of key tools for data analysis it covers a five-step framework for how to think through your analysis workflow and where AI can be helpful this helps a lot because the biggest challenge most people face is that they just don't know where to start Pro tip the type of data that you're dealing with like if it's structured tabular data versus unstructured text Data matters a lot on the type of AI tools and techniques you should use AI can be especially useful if you're dealing with large amounts of unstructured text Data like results from a survey or user comments you can also Implement methods to automate your analysis this section covering a breakdown of different AI tools for various data tasks is pretty helpful it gives you a good idea of what's out there and what's useful to meet your specific needs I suggest that you download this free guide using this link over here also linked in description I'm sure you'll find some valuable information in there and might even save you some time or give you some ideas for your next project this free resource is created by HubSpot who is sponsoring this portion of the video thank you HubSpot now back to the video moving on to module three prompt engineering like I said before this is my favorite module because I think Google does a really good job of giving practical tips on how to get started with prompt engineering while also giving you some Frameworks and techniques for you to build on the first tip is to give clear specific prompts with context for example you might be hungry and you ask Gemini can you recommend me some restaurants in San Francisco that is a decent prompt but to make it even better you can provide it with additional context like for example you can say like I'm in a mood for Japanese food and for a more cozy laid-back environment a good way to be more clear in your prompt is to consider the verb that you're using in this case you just said give me some recommendations and Gemini probably automatically just listed them out but if you want to be more clear you could have actually explicitly said to listed out for you maybe you were actually hoping for a table that has the restaurant names and then some other column like price point description popular food in the restaurant in that case you could have asked you to create a table with specific Columns of restaurant name description price point and most popular dish which then you can export as a Google sheet and share with your friends or something here's a few other common use cases like summarize the following text is an email from a software lendor summarizes main points in a bulleted form classify read these customer reviews and classify whether the sentiment for the reviews is positive negative or neutral extract read the blog post below and extract all the references to items from clothing I can buy and how much each item would cost create a bulleted list of just these items Pro tip is to always think about what you want the output to look like and be as specific as possible to describe that output translation translate our product descriptions from English to Spanish maintain the same structure and Casual tone that is used in the English version in the Spanish translation editing edit the language of the following paragraph So that it's easy for a general audience to understand it one more use case is problem solving this can be extremely help ful especially if you have very specific problems for example you can say to Gemini we are running a community program to teach children gardening skills the program runs from June 1st to August 15th and we want the children to be able to grow plants that will be ready to harvest by the time the program ends first identify a list of 10 plants that can be planted and grown in that time period include sources that support the time to harvest for each plant this is another protype ask you to include sources so that you can prevent hallucinations we want the children to grow three plants these plants should be as different from each other as possible so next choose three plans from the list that we will provide to children with this variety there are a lot of other little tips and tricks but it's important to note that prompting is an iterative approach like you don't need to come up with that perfect prompt in your first try generally the way that people prompt including professional prompt Engineers is to start with a small simple prompt first then potentially provide more details and then evaluate the output again and repeat this process until you get the output that you want the instructor encourages us to critically evaluate the lm's output using a few guiding questions is the output accurate is the output unbiased does the output include sufficient information is the output relevant to my project or task and is the output consistent if I use the same prompt multiple times my personal little Pro tip here is that after I iteratively get to the result that I want I then prompt the AI to write a single prompt in order to get the output that we finally have it's a good way to learn how to prompt it better next time over time as you practice prompting and do more iterations you're start gaining an intuitive Sense on how to prompt better another key consideration when prompting is the use of examples in AI terminology the word shot is the same as examples so zero shot prompting would be asking to AI something and not providing any examples one shot prompting is to give it one example and few a few examples if you're asking to AI something very simple you can just directly ask it and do zero shot prompting but if you're asking something that's more specific in Nuance it's good to give more examp examples were few shot prompting large language models are really good at pattern recognition so if you just give us some examples of what you want the output to look like it would be very good at mimicking the correct output for your response without you having to go into a lot of detail and specifying what you're looking for a few shot prompting example could be write a one- sentence description of a product it should contain two adjectives that describe the product review the examples and write a description of a skateboard in the same style then you provide two examples without actually having to specify what the style it is that you're talking about generally speaking if you provide more examples the answers would be more specific and more nuanced but if you're providing way too many examples it could also restrict the creativity of the large language model okay last prompting technique this one is more intermediate it's called a Chain of Thought prompting this involves asking an L to break down a task into a series of subtasks to do it's especially helpful when you have something that has complex logic or reasoning the example the course provides is say that you want to create what is called a purchasing codee so say there's an organization with a thousand employees each employee is assigned a unique purchasing code that's used for buying supplies or equipment first a technical support specialist needs to create a unique purchasing code for each employee so to do this The Specialist wants to use a custom AI solution and this is an example of a Chain of Thought prompt so first you want to provide context our organization assigns a purchasing code by combining an employees department and ID number all alphabetic characters are lowercase into purchasing code review the examp examples and then answer the question that follows in the same manner explain the steps involved in determining each employees purchaseing code then you want to provide it with an example so question Tiana B works in the marketing department and has an ID number of 9283 what is Tiana B's purchasing code answer the purchasing code for Tiana B is marketing 9283 to determine this first combine the department marketing with the ID number 9283 this results in marketing 9283 then change all alphabetical characters to lowercase this creates the purchasing code marketing 9283 you see how it's using an examp po to break down that task to show the L what to do then you write your request question Sylvie e works in a sales department and has an ID number of 2379 what is Sylvie E's purchasing code and then answer and you leave it blank for the LM to fill in this can be very powerful because after making sure that the LM understands what it is and it gets the question for a Sylvie e who works in the sales department correctly you can actually automate this process and create a purchasing code for every single person in your thousand person company this is a task that would be pretty tedious to do if you don't know how to code but now you can do it very easily using chain or thought prompting all right we are done with our beefy section of prompt engineering concentrate all right we're almost there next up is responsible AI the major takeaway of the responsible AI use module is the fact that there are lots of different harms that can be caused by AI tools and this is usually the result of biased output there are a lot of different types of harms to be aware of one major one is called a quality of service harm which is a circumstance in which AI tools don't perform well for a certain group of people for example when speech recognition technology was first being developed the training data didn't have many examples of people with disabilities and so the AI tool was not very good at recognizing the speech patterns of people with disabilities another example is called representation harm this is when a AI tool reinforces certain biases in society for example when translation technology was first being developed the a model would automatically refer to a nurse as a woman and a doctor as a man these are just a couple examples of so many other types of harms that AI tools can cause the instructor explains that one of the best ways to try to eliminate biases and harm is by having more representation in the development of the tool and to constantly collect feedback from the users I think AI safety is going to become a field that's going to be increasingly more important as time goes on so if you're interested in this field I do encourage you to dive deeper into this subject all right I'm going to be honest with you module 5 is like nice and stuff but it honestly doesn't have that much useful information so good news for you we are done congratulations you have now learned the AI Essentials but wait before you go as promised here is the quick little assessment I promise you if you do this assessment it will help you retain the information in your brain for so much longer you can write down the answer in your comments you can think it in your head say it to your dog whatever but here it is pause the video or take a screenshot all right and we are officially done even if you just like thought it in your brain I guarantee that you have increased the retention of that information significantly honestly these videos take me a lot of effort to make because I have to do the course myself and then I got to make this video so please let me know in the comments if you actually like this kind of video in which I do these courses cuz if not then I will go do something else so yes um let me know and I will see you in the next video or live stream