Show notes
- Front matter
- We already recorded this.
- Battle in the ‘Burgh
- The whole tournament on YouTube
- How to watch a live event
- ‘Fixing’ the Battlebots meta
- Spinners bad
- Weight bonuses for non-spinners?
- Are weight bonuses effective?
- How do you decide how much is enough?
- Walker/shuffler bonuses
- Shuffler bonus for UK Beetles dropped from 50% to 33% in June 2019
- SOW went 7-0 as a Heavyweight shuffler, and 2-4 as a Superheavyweight
- Walker/shuffler bonuses
- Removing Battlebots’ extra barriers to non-spinner success
- Arena hazards
- Less intrusive pit designs:
- Judging criteria
- Robogames judging (ctrl+f ‘flip over’ to find relevant section)
- Say what you want about Robogames, this is a pretty good set of judging rules
- Robogames judging (ctrl+f ‘flip over’ to find relevant section)
- Arena hazards
- A general discussion about AI and Machine Learning
- A conversation driven by the sordid tale of Apple’s ‘sexist’ credit card
- Defining AI, Machine Learning, and AGI (for the purposes of this discussion)
- AI: A decision-making system designed by a human. Can be debugged.
- Machine Learning: A system which produces a decision-making system. Built and then trained, can’t really be debugged.
- Artificial General Intelligence (AGI): Sci-fi version of AI which is, or appears to be, a human-level intelligence. Don’t worry about it. Yet.
- Current uses of AI/ML
- Image/video processing
- Social media
- How it all works
- The wonderful, wholesome, entirely benevolent Netflix algorithm
- Sam’s experiences with Reddit
- Promoting controversy/outrage
- Bad technology vs technology being used by bad people
- Setting ‘goals’ for ML systems
- Finance (and other stuff)
- How bias can develop within an ML system
- The Apple case
- How bias is picked up
- Why bias might not be detected
- How bias can develop within an ML system
- Medicine
- Cancer detection
- What happens when the machine causes a human death?
- The impact of one life lost vs x lives saved
- Potential bad implementations of the technology
- Crime ‘prevention’
- Predicting reoffending
- Opportunities for bias presented by data which would likely be used for training
- How to be better