Week 13 , GenAI posts, ChatGPT threads and other interesting links

Here is my newsletter for week 13 with links to the main posts and articles that I marked/tagged during this past week.

presentation I'm delivering in April

“Keep Calm and Don’t Get Your GenAI Hacked” is the title of a presentation I'm delivering in London next month. I'm exploring GenAI's security risks, focusing on the dangers of opaque, adaptable models that linked to internal APIs without safeguards (what can go wrong with that? :) ). For me the solution is read-only, deterministic GenAI models and workflows that are designed to prevent unpredictable behaviour.

on GenAI:

• This article outlines the importance of prompts for specific roles - specifying roles in fields like medicine, law, or engineering, helps to ensures that the output aligns closely with real-world expectations.
• The U.S.?Office of Management and Budget aims to ensure that when federal agencies use AI, they protect the rights and safety of the public:
• Here is a really nice list of new tools to help with presentation creation
• This is a really good list and documentation of AppSec security issues that tools like CoPilot can help to find. Checkout the prompt examples and the exploration of the limitations of CoPilot
• Really nice visualisation of idea of "the 3 amigos - software development team of the future" based on Matt Welsh's presentation.
• GenAI Rant and nice PDF document on GenAI Risk Management , I kinda agree with the author, where my comment is that anybody who thinks that GenAI is going to save the day automatically and independently is in for a massive shock

on Dev

• Here is my dev environment that created custom views to help with debugging are great examples of the power of taking the time to add capabilities to test frameworks and workflows?
• Cool JavaScript animation libraries

on Startups:

• This shows how it's so important to start with a highly focused MVP and focus on users that have the potential to become customers
• This is a really cool mind-map on how to measure the health of a StartUp

questions I asked:

• Is there an OpenAI API data set for the multiple GPT models that contains a nice set of requests/responses examples?

observations made:

• ?I think that most GenAI projects/startups conversions need to start by "answering where in Wardley Map evolution that project is?" (Genesis, Custom Build, Product/Service, Commodity) and "what type of team is in place?" (Explorer, Settler or Town Planner).

marked as "to read, to watch or to listen":

• Interview with Rob van der Veer, Irfaan Santoe and Jeroen Prinse about "AI and Security" for the re:invent security podcast
• TL;DR Sec "keep up with security research" news letter
• Guide for startups to scale to $100M
• Here is another document that provides GenAI guidance, this time for judicial professionals. Worth a read to see this happening and shows how the GenAI industry is maturing.
• GenAI risk assessment that companies in California need to fill in with details of their projects - interesting to see how this plays out in practice.
• I’ve just discovered these really cool OSINT exercises and solution write ups by Sofia Santos . Really great way to learn and practice OSINT.