Microsoft Envision - The Tour London
Microsoft Envision - The Tour London
Executive Session and Fireside Chat
Clare Barclay CEO of Microsoft UK spoke to Microsoft CEO Satya Nadella and Steven Bartlett co-founder of Flight Story & thirdweb, author & host of "The Diary Of A CEO" Podcast. I was unaware of the event until travelling down for a different event the following day, but arrived in time to catch the very last session of the day which was really informative and was great to not have missed it!
Satya said it is fantasic to be in London, the last time he was here he talked about AI and Machine learning and something the UK has always led in. Something has changed clearly in the last year, this new generation of AI and what it can do to any digital software category and has gone through a real change. Try from a first principals perspective what has changed, with generative AI what has changed and what is different. From last 70 years is the pursuit to find the most natural user interface, can computers be easy to use, we have not found this with text and voice with multi modal input and multi modal output and is multi turn and multi domain. This is the biggest impact in the user interface compared to the PC or mobile phone. The other thing we have done over the last 70 years is to digitise information of the physical world, we could have databases to make sense of things and now have a new reasoning engine to recognise patterns in data and do continuations. This is a new category of what you can do with data and is a very natural user interface and with this you can change any software including development with GitHub Copilot which was the first hit product of the LLM era, and this gave confidence to put Copilot into everything,
Windows was transformative in the workplace, we know what happened before and what happened after and this will happen before with AI copilots. The UK is leading with AI adoption already, there was a study of the industrial revolution where the UK got ahead where there was a lot of competition, the reason for this in order to get ahead you have to be fantastic at importing the latest technology and exporting really great value add and this is what the UK did back in the day, the same thing will hold true with AI, the penetration of AI with cloud computing makes it easier, but question is high quickly can organisations adopt it and have that value add. UK can lead not only in creation of value add and what does it meant to safely deploy AI, and be safe and equitable - it can be a tool or a weapon, we want it to be more of a tool. How to think about AI adoption and acceleration and what is needed for success, is the diffusion of the technology being extraordinarily fast, penetration of cloud in UK is high and adoption of AI is the next step but need to be mindful that rate of diffusion is fast and there may be no difference between adoption in the developing and developed world and it will be ubiquitous and ubiquitous fast. Copilot is reducing distinction between the user of a computer and a developer and the idea that anyone can program will be universally true, so getting these deployed is the best thing we can do and to put these tools in place and measure the outcome.
Looking at generative AI through an industry lens is which industry would lead this, but with regulation there may be impedance with getting adoption, have to move services and regulators to the cloud, now that work has been done the playing field has been levelled and there is no reason for any industry to have any differences in adoption, it would be a firm level on how they will drive the tech. There won't be distinction between large and small and young and old companies won't matter - it isn't that hard as the playing field has been levelled in a massive way and don't need to be taught a lesson by someone else, doesn't have to be who is cool and doing stuff, you are cool if your doing stuff. Companies are collecting use cases and building new products based on data or AI advances from financial services to retail services and how AI can be used in regulated industrial or others, it is exciting to see broad swath of industries being revolutionised by AI.
How do you balance AI with some of the risks and concerns, we should be talking about both capabilities and impact of AI including the unintended consequences, we have to have this conversation at the same time and not wait for these to have a devastating impact on society. The consensus that is emerging seems to be that we should have a risk-based approach to this, at the application layer, have the current frameworks apply, apply these to the decisions being applied in conjunction with AI, human judgement is needed to both input and output AI. With the new frontier models there is a lot of self-regulation, Microsoft took months to do the alignment work for the base model with the core values around fairness and bias and make sure can deploy something carefully including some of the red teaming that is going on, there will be some amount of self-regulation for frontier models but there will also be open-source models and Microsoft made some of their safety features available for these such as Llama and any current regulations would apply.
If you think about in the labour market there are jobs and skills, there will be structural shift where some parts will be automated and other parts will have different skills, what the people do will change and need new skills but these can be acquired on the job so the skills and the talent have to come together, new skills that are valued are acting as a signal for the other jobs that will be created, there will be many of these that will need different skills, and if this takes time we have a new tool to acquire these in the age of LLMs and Large Models as they are the great leveller in terms of going up the skill ladder. Jobs will appear with better wages and support to drive productivity with skills to drive monetisation.
Long term what excites them is the idea that we can have is that every person on the planet can have a tutor that is personalised, a doctor that can give medical advice and could even throw in a management consultant, take those three things just think about what it would do to agency for any individual, a lot of us are afraid to venture into new domains as it takes preparation, learning and initiative - but if it easy to get there as there is an assistant or Copilot to walk you though is going to be game changing and doing this for every person on the planet is exciting. Microsoft are working with physicians to see what they can do on the care side changing and what's changing on science side such as drug discovery and can have these built and created, can have better matching and workflow based on AI and it can be transformative in healthcare.
How is AI being used day-to-day? Using GitHub Copilot as how can you be CEO if you can't code anymore, GitHub Code spaces is a game changer and dependencies are provisioned and GitHUb Copilot means you don't have to learn that new every changing web framework and provides the joy of coding and they are back at coding and love it. Microsoft 365 Copilot is so habit forming, by instinct now they ask it questions about a document, you can ask this any time and find out about the intent behind a document, is like a spreadsheet coming with an analysis, one of the things they needed to do is write up an analysis and put it into an email and don't have to do that anymore. They are able to follow up and keep track of meetings, if they lost Microsoft 365 Copilot thet wouldn't know what to do.
Steven was born in Botswana and moved to the United Kingdom. He got expelled from school and got un-expelled as they made the school a lot of money then got expelled again. He started businesses that have gone public, has founded a web3 development platform, have a fund to support startups and have the podcast Diary of a CEO. He made money in school when 15 or 16 the school was debating to choose vending machines and to pay, he contacted companies and asked them to put them in for free and would get 20% of the profits, then were coming up and organising school trips. It is hard to distinguish between being dragged and driven - they saw themselves as being different, but they were driven just to be enough and be like their friends and this is stayed with them ever since.
They have a broad portfolio, so how do they articulate what they do, they try to resist labels, this is the thing you used to achieve last and useful in a social context but you can live your life that way and it doesn't correlate with their happiness and any label will limit them, so they prefer to avoid labels. They do think of themselves as a creator and that's the thing that makes them the most happy.
He used to lie about presents he never got, the anticlimax can make you unhappy, when your expectations go met then your happy but if your expectations go unmet then you're not happy - happiness is not a subjective thing. Expectations going unmet make you unhappy because of that. Their podcast is one of the things they are known for and have spoken to business leaders, one of the takeaways of what makes a good leader is having three responsibilities, finding the best people in the world, finding them with a culture where 1+1 = 3 and then setting out a vision. These things are difficult to do, but that is what they obsess about, and they want to speak about any of these, but the culture thing is the most important, solving companies problems is that they relate to company culture - the single most important thing in AirBnB is company culture and it is the single most important thing.
Company culture isn't something you have to brainstorm in a room, it is already there, it is reverse engineered from the mission you are trying to achieve in the world. There are a set of behaviours that are most contusive to achieve an objective and meet any values, They want to create the most popular podcast in the world, and things are changing all the time so need to be agile and lean in and need to fight for every one percent of gain and need to create systems, they research the music people enjoy, he has a track pad to inform production about the interesting things someone has said and hits a hidden thumbpad to inform what is interesting and summary are based on this, they will change thumbnails and description after an hour and that one percent gain is based on experimentation, they will perform more experimentation, they have a head of this and they have data scientists to make sure they are doing from 0 to 40 million downloads in two years, this is why it is the fastest growing podcast. Focus on the small gains, but can have small returns but may not be small, what if you don't have like and subscribe but say certain percentage doesn't subscribe and this changed the subscription rate by 350%, so 1 percent changes can have bigger improvements - it is the secret, do those one percent things, experiments that failed are subsidised by the ones that succeed, no one knows about A9.com or the Fire Phone but they do know about AWS that help offset that mistake.
Even if an idea failed you should be on to the next experiment, doors you can walk back through if you do it, the risk is the dithering and waiting for someone to get back or do something, it is the time you waste, if you get to 50% certainty then go ahead, but companies are held back by bureaucracy, you need to cut it down and incentive people to fail. You have as many company cultures as have managers and need to have measurement system and share failings and share with everyone you can. Feedback is failure and knowledge is power - they know the metrics and numbers, when comes to thinking of new hypotheses such as new languages they know the impressions they get for a particular episode of the podcast in a particular language - stay in the weeds. You need to keep up, you need to be failing faster than the rest of the world, the rate of change is accelerating, lean in when cognitive dissonance is larger than we understand then do it, remember that feeling, the cognitive friction and needs to be alleviated so have to dismiss it to reduce that dissonance so with Monkey JPEGs and AI is to lean out, people are doing this now with AI or blockchain. When something is weird and public criticism, dissonance means that there's a group of people who are having their job threatened, if people aren't doing this then don't bother, but if there is then lean in such as with social media.
AI is another amazing opportunity, others they have seen a clear upside and a clear downside, this is one where the downside they haven't quite figured out. The disruption could be so significant we are unable to quantify this, the upside is tremendous, and in terms of disruption and change will be enormous and will be controlled by those who lean in. All their team members each week had to talk about AI and how they had implemented it and started with experimentation. Realise that you have drawn a circle around you and be more expansive.