Gavin Williams’ Post

Useful read from Tim Wright below. What is interesting is that DORA is not a fundamentally onerous piece of policy. It's a good guidance for organisations to help themselves check if they are managing risk and operational continuity well. One area which seems to get a little more (disproportional?) visibility, not necessarily from the article but in general, is the idea of vendor concentration risk for cloud. If the Broadcom acquisition of Vmware has shown us anything, it's that we may focus too much on the e.g. 7% risk in a hyperscaler CSP, but not enough on the e.g. 52% risk on smaller software platforms, or even e.g. 41% risk on older mainframe platforms where few of the people working in the company today were even alive when the systems were coded.

I’m happy to share that I’m starting a new position as EMEA Lead, Cloud First Design at Accenture!

I'm conscious that I have been very remiss on updating LinkedIn recently. So I am thrilled to announce that I have started the next chapter of my career journey with Accenture Cloud First, where I am leading the Cloud First Design transformation consulting capability across EMEA. With economic pressures on one side and the incredibly exciting disruptive opportunities from AI on the other, there has never been a better time to vision the future of business, powered by the best cloud and technology innovations ever. I’m thrilled to lead this transformative initiative, and helping propel our clients' businesses forward. Accenture is the leading Consulting and Technology Transformation organisation in the world and Accenture’s commitment to innovation, industry expertise, and collaborative spirit resonates deeply with me. A heartfelt thank you to my incredible colleagues at my previous organization for the invaluable experiences and growth. Now I look forward to collaborating with brilliant minds and fantastic clients, solving complex challenges, and making a meaningful impact. Let there be change! #AccentureCloudFirst #NewBeginnings #ExcitingJourney #LetThereBeChange #CloudFirst #digitaltransformationjourney

#frAIdayThoughts - Don't Run with Scissors :-) I try to take a little time each week to keep abreast of AI and its potential contributions to our lives and our enterprise customers. The story below may be familiar to you. It was a legal case where the defendant was looking for previous case law/precedent which would support a dismissal. The lawyer searched and also used ChatGPT to pull up case studies. ChatGPT very helpfully pulled up 8 cases for them to use, which were subsequently found to be completely made up. "Made up" is the critical piece here, as generative AI is fantastic at generating information based on the confluence of vectors and next most-likely word etc. These case studies were hyper realistic and detailed, but essentially a hallucination. (It didn't help the lawsuit but the consequences for the lawyer #ai were not as severe as might be inferred, due to the fact that this was a new area and was considered accidental/non-mal-intent). While Andrej Kapathy was right when he said that "The hottest new programming language is English", we still need to be careful that we don't get damaged by these new AI possibilities. Effectively, scissors are great, but don't run with them. We need to treat AI like a super enthusiastic smart-and-helpful-but-naïve friend. Just like original programming, it will do exactly what you ask for, but beware of getting what you ask for! Even though it's slightly long in the tooth, the book "You look like a thing and I love you" (https://lnkd.in/eKAgquqJ) is a great book recounting loads of these examples. Great read and my 11yo son is enjoying it. For this, this is where prompting the AI is critical. When looking for facts and/or research, If you change your prompt to ask for XXX, but specify criteria such as "answer if you are considered a university professor" "you should not invent anything" "please cite your sources" etc. then you are much more likely to get something which has been "COLLATED" versus "GENERATED", as this forces the AI to prioritise factors such as groundedness. I personally like the new BingChat where it will cite its sources for validation. Perhaps we should treat GenAI like an old fashioned genie-in-the-lamp. You may want to assume it will be looking for a way to twist your wish, so you should be careful on how you phrase it! https://lnkd.in/erF7jUuw #ai #frAIday

Building (no pun intended :-) ) on my recent post on the ethics of AI. While the previous post was about the "outside in" view of AI from a legal and ethical POV, Microsoft's recent #build2023 event had huge amounts of information from the tech platform developer side with regard to how AI is able to contribute to all of our lives. Hugely eye-opening, it ranged from the frankly watch-with-popcorn entertaining side (see "Mark and Scott learn to code" where they use AI to build a game in real time), through to more indepth sessions on Prompt engineering etc. However, staying with the theme of responsible use of AI, one session stood out in particular. Using AI responsibly. https://lnkd.in/ev92w9-c First of all, we have definitely reached the tipping point now where suddenly use-cases abound, the platform exists, now we can start to use the platform itself to create more cases. It's literally a bootstrapping journey now. Secondly, MS has put huge efforts into the Azure AI studio to allow people to build and customise their own AI efforts, including setting thresholds for harm or hurtful content, but then allowing the users to set CONTEXT for this. A request for an axe may/may-not be irresponsible depending on context. Instead of having to focus on the mechanics of building the responsible guardrails, the tools are there to rather set the position of the guardrails, coach to improve/refine their relevance, and then monitor. Finally, while these technologies will absolutely impact the lives of people executing tasks such as help/chat assistance, what came home to me is the building of new jobs and roles in engine tuning and prompt engineering. For every post about terrifying emergent AI behaviour (https://lnkd.in/eVhtxSXR), it is nice to see the beginnings of efforts to allow people to use these platforms safely.

Last night I attended a brilliant lecture by Dan Hunter (Exec Dean School of Law) on law and ethics of Generative AI in Kings College London. The thing which was notable was he did not try to provide answers/solutions, but rather wanted to make sure that people are aware of the different areas of challenge, danger and ambiguity. The area is moving too fast for solutions to keep up and even laws being enacted today are already out of date. Fascinating (terrifying) example of where GenAI can not only make a prediction of the conviction of a defendant, but also predict the nature and duration of sentencing depending on which judge is sitting. GenAI is already performing in the top 4% of US Bar Exam and US Medical exam results. That said, at the same time, the ability of GenAI to improve the output of humans, and act as a "co-pilot" :-) for people, means that it is going to be the next world-shaping evolution. Provenance and authorial verification will become critical as we move forward. Companies like #microsoft are already taking steps in this area with the citations in their new Bing Chat, and the industry is investigating how to apply watermarks to genai content. In the end, GenAI is here and the possibilities for it to help humanity are endless if we can avoid the hazards. It's too late to stop the boat on the river, but if you can jump into the boat and row hard, you can steer the course away from the rocks. #genai #ai #ethics

It's great that the cloud can help in so many ways. It helps people innovate faster and explore new ways of bringing products to market, ideating new engineering faster. It helps through being highly efficient in the energy usage itself, and it helps through creating an ecosystem of connected players who can then form a common goal.