
discotheque
If evidence shows that an AI dispatcher IS more effective than the average human dispatcher would you support this?My primary concern is the adoption of preconceived biases and prejudiced within pretrained models. there’s been studies demonstrating that models relying on pretraining and finetuning have inevitably adopted and perpetuated forms of prejudice (like racism, homophobia, sexism, transphobia, etc) due to those forms of prejudice being present in the training data. how would that perpetuation look in an automated police dispatcher model?
Have you not been paying any attention to how AI companies operate? They advertise themselves as a cheaper alternative to humans then proceed to charge as much as physically possible while cutting as many costs as physically possible, up to and including throttling computing resources dedicated to public services if they feel it’s not as profitable as other ventures
even further, what other behaviors would the models pick up from finetuning specifically on training data about working in the police industry, and how would those new behaviors display themselves in the role as dispatcher? given the history of our country and our police system, this is a very slippery road imo
I’m not saying that a police officer isn’t more racist, all I’m saying is that an ai bot is able to be trained to be just as racist, or can unintentionally adopt racist ideological beliefs as well. how will that play out in the role of a member of the police force, given the history of our police force? like I have a feeling they’re going to end up adopting the same prejudices the existing cops have
I don’t think you understand how this works, the end goal of each individual company is to control as much of the market as possible. They are not going to just sit there and leave an untapped market untapped, and as far as voters making that decision goes look no further than Flock to see how that argument doesn’t fly
in all fairness, there is an active discussion on restricting the open source model market, especially after the recent infiltration of huggingface by openAI (which, apparently, huggingface used an open source deepseek model to held fend off the attack and that made some people upset at both openAI and the current administration)
Again, that all relies on the companies not actively lobbying the government to get contracts instead, which we all know is never going to happen And LAPD cancelling a contract because of public pushback still doesn’t negate the overwhelming majority of contracts that were implemented without voter input in the first place
The amazing part about lobbying is they don’t even have to get a specific politician elected! Instead they can lobby literally whoever is in office and get what they want whenever they want, regardless of which side is in charge! Lobbying is legal bribery and should be heavily regulated at a minimum
The government already has plenty of precedent for outsourcing AI to private companies, from Anthropic to OpenAI. It’s unlikely that they would do the job in house considering their past actions. Judging based on precedent rather than speculation, the conflict remains between the private vs public sectors.