9 Comments

This was great, Andy. Can you talk about any trade offs we make to apply this kind of solution?

It’d be a big deal to design better infrastructure for pedestrians in public spaces and privately owned ones, too! I’m thinking about big box stores that have designed parking lots for customers to drive into, but have been less thoughtful about how you get to the front door. Do I really have to risk my life walking through the Target parking lot to buy some TP? 😂

Expand full comment

No, LLMs cannot do the things you are claiming they can do. Algorithms built specifically for those purposes might be able to (and frankly I assume there are already people working on them), but a text generation tool cannot. And this is not just pedantry -- when you make claims like this you contribute to the hype machine rather than to actual possible areas where measured applications of technology can be useful. Convincing stakeholders to spend time and money on OpenAI LLM, for example, rather than another more useful algorithm/expertise is the equivalent of hiring the Boring company to control your mass transit rather than just, you know, buying some busses.

But at the end of the day, there are no technological magic bullets. Even if you had an algorithm that did all these things (and, again, I would be surprised is much of this did not already exist in some form or fashion), it doesn't do any good if there is no political will to implement it. That is what I think most people mean when they say there is no technological solution -- too many people think some magic bullet is going to come along and sole what is at the end a political problem and thus spend too much time in the hype cycle and not enough time in the organization trenches.

Not accusing you of that, mind you, just making a general comment on what I see as the state of the conversation generally.

Expand full comment

" when you make claims like this you contribute to the hype machine rather than to actual possible areas where measured applications of technology can be useful. "

Expand full comment
author

Some cities & ride share companies are already using LLM technology in ways I describe in this post. The premise is simple even if the execution is difficult: collect & summarize massive amounts of information. In a sense, LLM/AI combo is what transit planners or bike/ped planners or traffic engineers have done for years.

My major gripe w/ "tech can't save us" laziness is that it's absolutely false. People don't understand what tech is (e.g. the alphabet).

Expand full comment

I would like to see an example of a working LLM in this space. Beware of hype, and please keep in mind that and algorithm that "collect & summarize massive amounts of information" does not have to be an LLM and is likely not -- it is likely a specialized system that has nothing to little to do with LLM as they are commonly understood.

I am a programmer. And I am a programmer because my college made me take engineering classes even though I was not an engineering major and I fell so deeply in love with that I switched to an EE major. I am not anti-tech, I am anti-BS and there is an enormous amount of BS in the technology space. Lots of these companies dissemble about what they are doing, over promise and over hype -- how many times do we need to re-invent the bus? By running alongside the hype train instead of carefully supporting tech that helps, we risk making things worse in the long run. there are strong arguments that support for Uber has descreased support for mass transit among policy makers and that Airbnb has exacerbated the housing crisis in many places. And lord knows that every dollar spent on the Boring company was a dollar wasted. Tech hype has often been a drain on urbanization.

And even if all the promises come true, even if latching onto the hype of irresponsible companies results in the world's best system for determine how to make roads safe for pedestrians, for example, it won't matter if city governments cannot be convinced to spend the resources to implement them. These are political problems, not technology problems. Tech can help around the edges (it's not as if we don't have a bucketful of solutions for most of these problems already), but it won't take the place of organization. the technology calvary is not coming to save us. We are going to have to do that work ourselves.

Expand full comment
author

I added a "but wait, there's more" section at the bottom of the article.

You know I constantly emphasize the importance of engineering judgment and policy reform -- things that aren't shiny. My purpose in writing this is to think about ways emerging tech can be put to good use.

Expand full comment

The prioritization of cars over people in the application of technology is what can (and unfortunately does) make tech the enemy of active transportation. You cover this very well, but it truly seems that capitalism, fossil fuel industry, and auto lobbyists are going to have their way when it comes to funding and application of technology and it will be at the expense of vulnerable road users. Same as it ever was.

Expand full comment
author

There's no guarantee that we will triumph over Big Road. But there's also no guarantee that they must win. Things can get better in the end. :)

Expand full comment

I appreciate your optimism and expertise.

Expand full comment