Jack Dorsey (@Jack) is the former CEO of Twitter and the current CEO of Block (formerly Square).
Engineering at scale
- The open source community is pivotal to learn about hacking, programming, and engineering
- All the learnings, mistakes, and successes are laid bare. You get to see what it takes to build something meaningful
- Scaling a product to millions of users comes with significant engineering challenges (Such as securing the system and having minimal down time)
Will humans ever have deeper connections or build relationships with an AI? That is less a function of artificial intelligence and more a function of the individual and how and where they find meaning.
AI is… concerning!
The technology to create AI is moving much faster than the technology to detect AI. This needs to change.
- For the financial industry, someone could trick risk models or identity detection systems
- This quickly expands to touch everything, for instance, faking passports or driver’s licenses
- Like security, a perfect detection system can’t be built. Instead, we should focus on evolving detection systems quickly based on signals
Scaling
- Scaling is ultimately a critical thinking and problem solving challenge
- It comes down to breaking down the problem into smaller parts, asking the critical questions, and going deep enough to form a credible hypothesis that you can test
- The biggest mistake: Trying to solve many problems at once
- Resist categorized thinking and look for the underlying systems that make it all work
- For instance, get too focused on “this is a hardware issue” and you may miss the software solutions
AI: The developments and the challenge
- AI being able to process natural language is tied to its ability to explain, in natural language, why a decision was made – In other words, we have to pair:
- The ability of AI and deep learning to bubbling up interestingness quickly
- With human discretion around severity, depth, and meaning
- That said, even humans can’t always do a good job explaining the reasoning behind decisions, making it a challenging problem
- Keep in mind, there is great risk in building AI black boxes that can’t explain their criteria, then trusting them with all sorts of decisions (e.g. Lending decisions, content recommendations, driving, health)