Related Posts
Bain & Company Hey! Got dinged at BCG and Mck and then hadn't heard back from Bain, but just found out I made it to final rounds. I haven't been prepping for over a week as it'd be two since my first round at Bain.
Help! Whats different or similar about Bain final rounds? I'm applying to Texas offices for an experienced AC role. I also may have to have an accelerated final round as I'm leaving the country for most of April. Bain & Company
Anyone got premium bagpack?
Social work
@hospital medical social work
Rising Star
Conversation Starter
Thoughts on Trinity Partners?
Do PEs offer singing bonus or relocation fee?
5% in the US. Putting in my notice tomorrow!
Additional Posts in Artificial Intelligence
New to Fishbowl?
Download the Fishbowl app to
unlock all discussions on Fishbowl.
unlock all discussions on Fishbowl.
We already worked this out in Minority Report circa 2002
So true!
It won't be capable of this for a long time. This UHC CEO actually implemented an AI to automatically deny people coverage based on profit motive and policy standards during his tenure. Want to know how accurate it was?
It was wrong 90% of the time.
Very interesting
It hallucinates where it’s overly confident based on bad or incomplete data, it can’t account for random events, it also has bias, overfitting. Just one example recently where machine learning was off, not trying to get political here. Every machine learning model, LLM had the election as going to Kamala, by a very narrow margin. The data was based heavily on polling. Polling showed the election as a toss up. Prediction markets had Trump around 65 to 70%, so humans vastly out performed AI. I discarded the AI models, aligned with prediction markets. AI wasn’t accounting for history showing when Trump is tied he’s up by 5-10%, which is I think how prediction markets got it right. I placed my bets with high confidence, made a fortune. I’ve also seen where outcome is very predictable 30-60 minutes out based on machine learning, but that accuracy is awful hours or days out. Also if we’re consuming the same information outcome 30-60 minutes out is usually obvious to a human.
This is a great answer, thank you
Boring
Not only are crimes not predicted, they happen in broad daylight and are not enforced.
Maybe because resources aren't allocated appropriately?
You're conflating two discrete things:
1. Predicting an increased probability of a crime taking place at a certain location and time* and taking extra steps to avoid that timeline. Public GOOD (nuanced/depends, see below).
2. Punishing a person for a crime they did not commit. Public BAD.
Our Bill of Rights, if followed, has all of the fundamental and necessary protections for due process, against self-incrimination, unlawful search & seizure & etc. They do not cease to exist just because the technology advances.
Unfortunately, such issues always take a while to work themselves through the system. Law enforcement, reckless legislators, and power-happy Governors, they are gonna do things, things that will be challenged in the courts, that might take many years of decisions and appeals, and to resolve themselves.
The major issue I think comes down to what kind of data collection, what level of surveillance must the innocent endure in the case of (1). This is an unresolved tension in the West currently - the vandalism of ULEZ cameras, availability of location-base cellphone data, objections to the use of facial recognition technology in public & etc. Whereas I understand it's pretty ubiquitous in China where there isn't the same political philosophy or framework in which individual rights spring from natural law and the government derives its rights exclusively from the consent of the governed.
* More of a predictive analytics thing, but the definition AI keeps changing based on what's novel. Which is LLMs for the minute but will probably be different again in a few years.
Thank you for such a thorough answer