Prompt Title: AI Trends and Considerations.

Human: Summarize the following in 5 bullet points. Title: "Microsoft Xiaopeng Li" Transcript: "[Music] thank you thank you so much our fantastic host Antonio Melissa such a pleasure to be here today we will be exploring some of the key trends and considerations when it comes to adopting AI I will start by sharing some Trends which are really shaping the AI industry right now and then share a few key considerations for organizations to think about when it comes to scaling AI I have only 20 minutes so let's let's Dive Right In so the first Trend I want to talk about is really around product ivity we all know about the productivity revolutions triggered by steam by the new um by electricity and by the internet and today I believe we're starting to see a new paradigm shift and that's powered by Ai and let me explain if you do not live in a cave you probably have seen have tried have read about chat CBT uh that's the first time we mentioned that GPD in the presentation by the way Melissa it's one of the most intelligent chatbots we have on this planet and what makes it unique is not t he chatbot itself not the simple user interface it's the fact that in this this application is powered by one of the largest language models on this planet and that matters because large language models they bring so much intelligence into applications like chat gbt people are already using chat DBT for a variety of tasks from copywriting to code documentation and they're gaining a lot of productivity benefits from such application and it's not only in shared TBT let me mention another example a round Bing chat for years Bing has been an underdog but that has changed you know when it comes to uh internet search since the launch of the new Bing uh with Bing chat you can have a similar experience just like cheddypt you can have natural language conversation you can ask questions you can provide inquiries you can get answers in natural language back to you what makes Bing chat unique is that when you ask questions actually it does a internet search Real Time behind the scene and then surfa ce back to you a summarized answer to you with a list of references for you to go deeper and validate and that's really important because now internet search has become so much more productive so much more productivity so much more productive you no longer have to read the top three or ten articles from the search you can get the summer is the answer right away and this is not only happening in the text space it's also happening in a visual space so uh every every picture on this slide is actual 0 years old like myself you probably remember clippy some people say clippy is the father or the grandfather for many of the AI assistants we have today while clippy was cute but not entirely a success when it comes to increasing our productivity but now you will see a new generation of clippies which are coming to us and they're called co-pilots so today let me welcome you into this new era of AI co-pilot where AI is going to become the right hand for every row and every profession with AI copi you can do more so you know from from documentation to creating SQL queries GitHub co-pilot is already increased in developer productivities at scale in fact many of my colleagues and Microsoft on engineering side they are already writing you know up to 50 of their code on a daily basis with GitHub co-pilot and that's just amazing and it's not only Microsoft uh you know another example is Duolingo one of the most popular language learning apps and as you can see from this slide you know by adop development so it's not only for developers actually co-pilot is coming to you know all of us so um a couple of weeks ago we have launched you know Microsoft 365 co-pilot by combining large language models with Enterprise data within your organization you know GitHub I mean Microsoft 365 co-pilot allows you to you know chat with your own internal knowledge base from documentation to presentation uh you know to calendar to to context Etc you can use your natural language to find relevant informat e least favorite things of mine so now with co-pilot in Microsoft teams I'm able to use AI to automatically transcribe the meeting summarize the meeting think about a two hour meeting into a few paragraphs in text and and that's so much easier to recap and not only summarizing the meeting but also giving you a list of conclusions you have uh you know you have concluded during the meeting but also a list of action items even assigned to the right person based on the discussion with within the mee uh Trend so you might wonder you know applications are getting smarter we're getting more productive but what is happening behind the scene so well it's actually Foundation models they are actually uh the game changers you know behind the scene so traditionally speaking you know AI models tend to be very narrow well defined uh they can only do one thing but now with Foundation models they they can actually become more versatile they brought they bring more in you know broader intelligence to uh h just over a hundred million parameters to the latest version with billions of parameters you know um they are you know significantly increasing their size but also capabilities and that's of course enabling more intelligent applications that we will benefit from as humans let's take Microsoft as one example I'm obviously I'm biased I work for Microsoft so let me take take this uh so add a Microsoft we have been building Foundation models you know from Microsoft research but also in our partner urself on the other hand we Infuse the same Advanced AI capabilities into different software product services that we're offering from Bing to LinkedIn you know to to Outlook to even Xbox so I know many of you are developers so I want to give you another level of details when it comes to the cloud AI services so let's zoom in on Azure Open Air Services so um you know in brief I should open AI server service turns large language models into Cloud AI services so you can use them as simply as using ou know using them for Content generation for example creating you know for for copywriting and also creating images for for marketing campaigns but also doing summarization of internal documents as well as you know summarizing you know the transcription from a call between a customer and a call agent you can of course do code generation documentation as I mentioned before with GitHub co-pilot last but not least you can also do semantic search and that means you can help your customers navigate Azure open air service first let me talk about Strike Back strapback is a European leader when it comes to construction they're operating globally so by using Azure open AI they're actually assessing and predicting risk for new construction projects I mean construction projects are expensive so um they should not fail right so it's good for us to have a good understanding on the risk level so they are able to reach 80 percent accuracy level with their risk prediction model really impressive anot m you know from days into hours to save them a lot of time and make their employees more productive and also relieving them from really boring tasks to be honest make them more happy another example I want to show it's a smaller player so zamo is actually a software startup they're building conversational AI platform and what they're doing you know with actual open AI is to actually turn web pages and also you know documentations into the knowledge base the brain power for Commerce conversationa this new era for AI and and think about how they can Thrive and how they can lead in this new age of AI so let me start by talking about people that's what always comes first um and you know in particular I think you know we need to train and skill your people you know across your entire organization at this new age of AI uh if you look at this slide you know I'm showing you the typical life cycle for AI and machine learning development you can see there are different roles involved you know fro se can be your business decision makers or even your executive leaders it's incredibly important for them to understand the implications of AI as well so they can on one hand provide domain expertise into the development of AI but also on the other hand support you when it comes to putting those Solutions into production and actually you know Harvest tangible business value from AI Solutions so that's all about people second let's talk about process so you know AI adoption is so much more beyond ns from your customers and they can focus on addressing more complex inquiries and that makes them more happy and then that makes them an expert than just you know uh you know focusing on mundane tasks another example is that with the way the generative AI Solutions you know copywriters they would no longer start from scratch and now they actually become editors because they can take a draft from Ai and then they think about how they can polish and then bring their creativity and marketing exper o emphasize is that you know there are two things you to think about beyond the an machine learning platform you you will need for your for your adoption the first thing is really about laying a solid data Foundation some people say when it comes to air and machine learning you know crapping crap out so it's important for you to have your data organized in a good place from operational database to you know analytic Solutions like data warehouse to data governance solutions to understand how data licies processes and toolbox for your practitioners to practice responsible AI on a daily basis and that means your designers should think about it your data scientists should think about it your compliance and legal team should think about it as well so of course you know we as Microsoft together with Solitaire we are more than happy more than happy more than happy to support you on many of those topics from building platform you know to practicing responsive AI or establishing process change m rtunity is really yours AI is becoming more powerful and more accessible than ever but it's up to you and your organization to put AI into practice to create tangible business and societal impact I look forward to seeing what you can achieve with AI and you know let's let's go out there make AI in put AI into practice and create you know you know good benefits for us as human beings thank you so much [Music] thank you so much for this and a wonderful shout the forefather of all assistance tools talked about Foundation models and how they really are the basis for all of these modern applications that we use every day and not even necessarily realize like things in LinkedIn or word or helping you correct your language uh voice assisted AI all these how representative do you feel uh that these Foundation models are of the global population that's a a very valid question so you know as we know just to overly simplify Foundation models like you know GPT models they are trained on a snapshot cess to you know laptops and mobile devices Etc so that means there are intrinsic bias on the internet and that that becomes the bias in the training data for large Foundation models right so we have to acknowledge that but then of course knowing the boundaries and limitations of foundation models would help us to use them properly right it's like using any to you nothing is perfect when you understand you know what you can use them for what you cannot right and at the same time I really think w olutely but then depending on which AI you're referring to right so you I would advise against using your using chat DPT or Bing chat you know to summarize your internal meetings because there are internal sensitive information and you don't want to leak that in the public services but but I highly encourage you to explore services like Azure open Ai and to build a internal chatbot or internal Enterprise chat CPT then you can safely and confidently talk about business topics you know within your lot of your energy a lot of people expertise as well but that's the whole point because weight train Foundation models so they become the foundational platform for you to build your Solutions on top and that means you as a developer as an organization no longer need to train your own model your own large models so we do this once and for all and then you can build your Solutions you know on top and that's going to be more efficient but at the same time of course you know we are as an industry co bs created prompt engineer is something that didn't really exist six months ago but but there's also different estimations on on on job losses so so what are your thoughts and again we understand that the the speed is is just so crazy but uh what about our our question is on Creative writers illustrators coders what is their future uh in in their current jobs as the and what are the fundamental limitations of the foundation models when compared to human Minds what are you what are your thoughts sing AI you know which can relieve us from you know the the more basic and more repetitive tasks we can focusing on solving uh actual problems problems which are more challenging if you think about the developer do they really enjoy writing you know lines of code after one another not necessarily they want to solve difficult problems and the code the simple code can be actually assisted by you know um by Ai and also same goes for documentation I don't know what developer will actually love you k dentify and maximize the value of people in companies oh maximize oh I think you know for me a top of mind is to retain your people to make them happy that's the first thing and the second thing is to make them productive I think with AI systems you can do both actually there's a study done by GitHub by GitHub you know on the internet you can find it so with Solutions like GitHub co-pilots developers are not only more productive more productive they're also more satisfied a"

ChatGPT:

Microsoft Xiaopeng Li talks about key trends and considerations when adopting AI. Here are the key points he covers:

• The AI industry is shaping a new productivity paradigm shift, and large language models like GPT and Bing chat are driving this change.
• Bing chat provides natural language conversation and real-time internet search, which has increased internet search productivity.
• Co-pilots, which are AI assistants, are increasing productivity across various professions, including documentation, SQL queries, and language learning.
• Foundation models, which are versatile and capable of broader intelligence, are game-changers behind the scenes for creating more intelligent applications.
• Azure Open AI Services is turning large language models into cloud AI. Microsoft is building foundation models from Microsoft Research and infusing advanced AI capabilities into different software product services.