Interesting commentary often appears as throwaway lines, often late in articles.
In this case:
Cal Fire’s mission is to suppress 95 percent of all fires when they are 10 acres or less.
It sounds as if total suppression remains California's official fire-management policy and protocol.
I can understand how this is desirable during peak fire season (though increasingly there is no such thing as "fire season", only "fire years"), but without permitting or initiating limited burns, the past two decades' history of every-increasingly large and ferocious wildfires is destined to continue.
As for the AI component: our human senses work best by utilising multiple senses. For wildfire detection, identifying smoke in visual imagery is helpful (and the figure of 90% coverage for fire-prone wildlands was another sobering trivium from this article), but cross-validating against other data, say, infrared thermal imaging to correlate smoke-like images with actual heat-on-the-ground would be useful. That spy-satellite remote sensing is being brought to bear is another interesting revelation here.
> It sounds as if total suppression remains California's official fire-management policy and protocol.
no, CalFIRE was excoriated after the 2017 season. The upper management including the Chief were.. uh... "fired"
Meanwhile, a Sacramento appointee with some connection to the native population has to sit in committee meetings for the last five years, while very very slow changes are made. The political class in California wants to extract votes at every meeting, so they emphasize "first people" every third sentance.
The truth of the situation is clear to many, but the allegiance is bound to institutional goals and ever-scarce funding. Many detailed references available on request
I'd made my comment (which note is based on the NYTimes story, there doesn't seem to be a corresponding observation in the LA Times version, see dang's comment about merging threads) because my understanding was that total suppression was no longer policy. I've commented at length on that before, see: <https://news.ycombinator.com/item?id=36233156>.
There's the possibility that the NY Times got this factoid wrong, though it's precisely the sort of comment I'd expect to be fact-checked prior to publication. Assuming the comment is accurate, it's a current documented statement that strongly suggests (near) total fire suppression along the lines of the US Forest Service's notorious "10 AM policy" (all reported fires should be extinguished completely by 10 AM the following day) remains in effect.
I'd much appreciate seeing your further references.
(I have written you a long response, since I did not make time to write a short one)
First to say, the land across California is split in ownership and management between US Federal Agencies in total, the State of California, private companies for commercial use - importantly timber and electrical power, and other uses. Historically the slow-moving majority-owner Federal Forest management emphasized the value of forests as a crop by charter, leading to the now-infamous density and fire-abatement policies lasting well into the 21st century.
Things changed quite a lot legally in the aftermath of the 2017 catastrophic wildfire season that wiped out a third of a city, Santa Rosa, just North-East of San Francisco. The 2017 year's wildfire totals were +1.5 million acres. Only later to be exceeded in 2021, including the Dixie Fire "in Butte, Plumas, Lassen, Shasta, and Tehama counties started on July 13, burned a total of 963,309 acres, destroyed 1,329 structures and damaged 95 additional structures." (oddly, part of which was via arson by a CA college Police Criminology professor named Gary Stephen Maynard )
A legal document was drafted and signed, that resolved the cost and authority imbalances between four Federal and State Agencies under Governor Newsom. Next, by order N-05-19 from the Office of the California Governor (among many such orders), State Agencies and stakeholders are required to respond in writing, and on a regular basis, on what is being done.
The political document in effect now from Gov. Newsom / Sacramento is called:
california wildfire and forest resilience action plan_2021.pdf
There is more to say on the topic of management and liability, but let's move on..
High-tech constantly seeks to sell products to government, but actual progress is sometimes backwards in practice. This time, real advances were made.. for example:
The reigning academic heavyweight for Forestry and California Forestry, is Dr Scott Stephens, with well-over 100 publications on the topic. Reading a little more deeply, you might find a man who knows how to navigate between th...
Thanks for that, and yes, that would confirm the NYT's account.
I appreciate mistrial9's extended remarks, I'm going through their links and references, and again want to emphasize that I understood policy to be OTHER than total containment.
Both the NYT's account and CalFire's own policy page ... strongly suggest otherwise.
Caveats in my original comment here notwithstanding, I'm disappointed.
As noted by goatsi downthread, "suppress 95 percent of all fires when they are 10 acres or less" is at the top of CalFire's webpage: <https://www.fire.ca.gov/our-impact>.
There’s definitely a middle ground. Cameras have huge advantages in some areas. A ton of traffic enforcement, a job which is extremely dangerous for police now, can be done with cameras. Cameras do not fail to recollect, they do not get tired, they don’t give a pass to “desirables” and crack down on others, and they don’t get scared and kill people. The vast majority of traffic stops are a “this could’ve been a letter” situation.
If people are speeding, send out a ticket. If their tail light is out, send them a notice in the mail or a ticket depending on the situation. If people are driving drunk, cameras can see that as well, and enforcement can be dispatched. If a vehicle is stolen, a network of cameras can find it faster than people can.
I was not thinking of the robot enforcement mechanisms which you mentioned. Those are indeed a benefit. I live in a part of the world where they are used, and it is much better than selective enforcement.
I was imagining a certain billionaire's Stasi as as service company informing inhuman robots to enforce the law, as said billionaire saw fit. Being a black box, the rules might be more shitified than speed > 60 mph. They might be: negative social connections >n. Who would know, it's a proprietary black box.
The funny thing is that I now realize that inhuman humans would be fit for the job as well.
This is an area that opensource would shine.
Our existing system of laws is indeed open source for the most part. I hope it stays that way.
Who determines if someone is a criminal or not? Do we have a perfect mechanism for it? Can we trust the algorithm to not harass or harm innocent people? Unfortunately no.
There is a reason police need a warrant from a judge before they are allowed to search or detain someone. Same should apply to AI systems.
AI-drones linked to ordinary police work in the USA are the subject of a hundred startups right now. examples on request.. some of the readers here are investing in them , including facial recognition at a minimum.
Just as a measure of how much expectations have shifted, Artificial Intelligence: A Modern Approach came out in 1995. This is the canonical textbook you would be given if you took an AI course today. Automatic image recognition is barely mentioned, apart from detecting highly structured objects (like people) via sliding windows. No mention of CNNs or ViTs. The state of the art at that time was planning (optimizing some set of steps) and adversarial search, along with some pretty basic NLP techniques.
Wildfire detection is one of the oldest ideas in computer vision, I'm more interested in coming up with better ways to do controlled burns while handling risk factors like terrain, poison oak fumes, containment, etc.
Many healthy ecosystems in California require fire - there is no mention of what will happen post-detection but if this is just a quicker way to extinguish fires, it will be a net negative long term by increasing the fuel load even higher.
Edit: I see this now: "[. . .] as they will be quickly extinguished while still small." Putting out fires quickly is not going to help us long term. We need to develop ways to proactively burn areas safely.
That will happen once the AI system is allowed access to drones. The AI will then maintain the optimum amount of tree biomass for ecological activity while reducing negative externalities on human populations.
Pretty sure there’s been a deployment of this along the mountains here for a few years. I recall seeing some tweets the last fire we had showing their systems detecting fires.
Are they tossing any money at actually doing forest management or are they just throwing money at random systems in hopes that they can solution their way around actually doing forest management?
It's California so it's probably the latter. This is the same state that declared they want all semis in the state to be electric by end of the decade and then signed a contract with Nikola whose CEO is in prison because the entire business was fraudulent and had no product.
For the benefit of the rest of us, please smash your TV and/or AM radio or whatever medium told you these lies.
CARB's heavy truck truck rule requires 40% of semi truck sales to be zero emissions by 2035. This is quite contrary to every detail of your claim. The state's ZEV strategy does not hinge on Nikola. That is some niche conspiracy junk you heard from an outlet that had already worn Solyndra down to the bone.
"Rake the forests" seriously? Pretty low-effort comment right there. CalFire's funded staff has increased 50%, by over 4000 new positions, in 2023 vs. 2021.
Can you please make your substantive points without swipes? Your account is a perennial thorn in the moderator's side because you constantly break the site guidelines and at the same time are otherwise a good contributor. I really don't want to ban you and I don't know what else to do other than keep trying to persuade you to use HN in the intended spirit (https://news.ycombinator.com/item?id=35746140).
Seems based on ALERTWildfire, based only on the name. Perhaps they’re affiliated in some way. I wouldn’t be shocked if they are the same program, but CalFire feels the need to be “more California” about it.
That program is coordinated out of the University of Nevada.
* Device with a low power thermal sensor, cellular connection and a battery that sends a heart beat with the current temperature once every minute or so. Maybe even BLE mesh network of such sensors. Simple patterns like a sudden increase in the are or sensors going offline can indicate something is happening.
* Does it really need a live video? Time-lapse would require less traffic and energy. Many camera's seems to be attached to the existing poles or lines and potentially limiting coverage area.
* I think a modern approach for such projects should be: put streams from sensors of different types online + an archive of data and have it as an ongoing kaggle-like competition for the best approach with the condition that resulting model should be made available to CalFire for their own use.
> * Device with a low power thermal sensor, cellular connection and a battery that sends a heart beat with the current temperature once every minute or so. Maybe even BLE mesh network of such sensors. Simple patterns like a sudden increase in the are or sensors going offline can indicate something is happening.
I'm not sure what is simple about needing to deploy orders of magnitude more devices to achieve similar levels of coverage...
But a casual passerby spent 5 minutes thinking about a problem and uncovered a solution that the teams of domain experts missed in their years of dedicated work - surely there aren’t any downsides!
Because it can be maintenance free, in the middle of nowhere and not rely on source of power and stable connection. https://www.dryad.net/wildfiresensor - here is an example.
I've never done something like it but I'd assume there are lots and lots of gotchas with plans to deploy zero maintenance devices.
It would make sense to me that nonzero, predictable, acceptable maintenance is better than a plan for zero maintenance which, if it fails, would leave you with more devices than you can possibly fix.
Cool, any software that will come out after 2023 will be labeled as AI. Doesn't matter if it is just a statistical analysis software. Call it AI. Cool.
57 comments
[ 3.5 ms ] story [ 124 ms ] threadThough technically, this is machine learning and not AI.
In this case:
Cal Fire’s mission is to suppress 95 percent of all fires when they are 10 acres or less.
It sounds as if total suppression remains California's official fire-management policy and protocol.
I can understand how this is desirable during peak fire season (though increasingly there is no such thing as "fire season", only "fire years"), but without permitting or initiating limited burns, the past two decades' history of every-increasingly large and ferocious wildfires is destined to continue.
As for the AI component: our human senses work best by utilising multiple senses. For wildfire detection, identifying smoke in visual imagery is helpful (and the figure of 90% coverage for fire-prone wildlands was another sobering trivium from this article), but cross-validating against other data, say, infrared thermal imaging to correlate smoke-like images with actual heat-on-the-ground would be useful. That spy-satellite remote sensing is being brought to bear is another interesting revelation here.
no, CalFIRE was excoriated after the 2017 season. The upper management including the Chief were.. uh... "fired"
Meanwhile, a Sacramento appointee with some connection to the native population has to sit in committee meetings for the last five years, while very very slow changes are made. The political class in California wants to extract votes at every meeting, so they emphasize "first people" every third sentance.
The truth of the situation is clear to many, but the allegiance is bound to institutional goals and ever-scarce funding. Many detailed references available on request
There's the possibility that the NY Times got this factoid wrong, though it's precisely the sort of comment I'd expect to be fact-checked prior to publication. Assuming the comment is accurate, it's a current documented statement that strongly suggests (near) total fire suppression along the lines of the US Forest Service's notorious "10 AM policy" (all reported fires should be extinguished completely by 10 AM the following day) remains in effect.
I'd much appreciate seeing your further references.
First to say, the land across California is split in ownership and management between US Federal Agencies in total, the State of California, private companies for commercial use - importantly timber and electrical power, and other uses. Historically the slow-moving majority-owner Federal Forest management emphasized the value of forests as a crop by charter, leading to the now-infamous density and fire-abatement policies lasting well into the 21st century.
Things changed quite a lot legally in the aftermath of the 2017 catastrophic wildfire season that wiped out a third of a city, Santa Rosa, just North-East of San Francisco. The 2017 year's wildfire totals were +1.5 million acres. Only later to be exceeded in 2021, including the Dixie Fire "in Butte, Plumas, Lassen, Shasta, and Tehama counties started on July 13, burned a total of 963,309 acres, destroyed 1,329 structures and damaged 95 additional structures." (oddly, part of which was via arson by a CA college Police Criminology professor named Gary Stephen Maynard )
https://en.wikipedia.org/wiki/Dixie_Fire
https://en.wikipedia.org/wiki/Tubbs_Fire
A legal document was drafted and signed, that resolved the cost and authority imbalances between four Federal and State Agencies under Governor Newsom. Next, by order N-05-19 from the Office of the California Governor (among many such orders), State Agencies and stakeholders are required to respond in writing, and on a regular basis, on what is being done.
https://wildfiretaskforce.org/
https://energysafety.ca.gov/what-we-do/wildfire-safety-advis...
The political document in effect now from Gov. Newsom / Sacramento is called:
california wildfire and forest resilience action plan_2021.pdf
There is more to say on the topic of management and liability, but let's move on..
High-tech constantly seeks to sell products to government, but actual progress is sometimes backwards in practice. This time, real advances were made.. for example:
https://cloud.google.com/blog/topics/sustainability/how-the-...
Please recall that the largest wildfire in the history of the State of New Mexico, was a controlled burn gone wrong.. very wrong..
https://www.washingtonpost.com/nation/2022/05/17/new-mexico-...
https://www.npr.org/2022/05/20/1099625787/new-mexico-wildfir...
https://losalamosreporter.com/2022/05/05/rep-leger-fernandez...
The reigning academic heavyweight for Forestry and California Forestry, is Dr Scott Stephens, with well-over 100 publications on the topic. Reading a little more deeply, you might find a man who knows how to navigate between th...
>https://www.fire.ca.gov/our-impact
Right at the top of the page.
"CAL FIRE aggressively fights fire, with the goal of keeping 95% of fires contained at 10 acres or less."
Seems very clear to me.
I appreciate mistrial9's extended remarks, I'm going through their links and references, and again want to emphasize that I understood policy to be OTHER than total containment.
Both the NYT's account and CalFire's own policy page ... strongly suggest otherwise.
Caveats in my original comment here notwithstanding, I'm disappointed.
The NYT account is confirmed.
You can trust robot law enforcement as much as you can trust those who programmed them, at most.
If people are speeding, send out a ticket. If their tail light is out, send them a notice in the mail or a ticket depending on the situation. If people are driving drunk, cameras can see that as well, and enforcement can be dispatched. If a vehicle is stolen, a network of cameras can find it faster than people can.
I was not thinking of the robot enforcement mechanisms which you mentioned. Those are indeed a benefit. I live in a part of the world where they are used, and it is much better than selective enforcement.
I was imagining a certain billionaire's Stasi as as service company informing inhuman robots to enforce the law, as said billionaire saw fit. Being a black box, the rules might be more shitified than speed > 60 mph. They might be: negative social connections >n. Who would know, it's a proprietary black box.
The funny thing is that I now realize that inhuman humans would be fit for the job as well.
This is an area that opensource would shine.
Our existing system of laws is indeed open source for the most part. I hope it stays that way.
There is a reason police need a warrant from a judge before they are allowed to search or detain someone. Same should apply to AI systems.
False negatives are.
Many healthy ecosystems in California require fire - there is no mention of what will happen post-detection but if this is just a quicker way to extinguish fires, it will be a net negative long term by increasing the fuel load even higher.
Edit: I see this now: "[. . .] as they will be quickly extinguished while still small." Putting out fires quickly is not going to help us long term. We need to develop ways to proactively burn areas safely.
CARB's heavy truck truck rule requires 40% of semi truck sales to be zero emissions by 2035. This is quite contrary to every detail of your claim. The state's ZEV strategy does not hinge on Nikola. That is some niche conspiracy junk you heard from an outlet that had already worn Solyndra down to the bone.
https://news.ycombinator.com/newsguidelines.html
This is who is actually doing the project:
https://alertcalifornia.org/
And here is a better overview of the program (imo):
https://www.governing.com/infrastructure/cal-fire-equips-art...
That program is coordinated out of the University of Nevada.
i.e. it doesn't work when it's so dry that the leaf canopies are igniting.
Why did not they go with a simple approach:
* Device with a low power thermal sensor, cellular connection and a battery that sends a heart beat with the current temperature once every minute or so. Maybe even BLE mesh network of such sensors. Simple patterns like a sudden increase in the are or sensors going offline can indicate something is happening.
* Does it really need a live video? Time-lapse would require less traffic and energy. Many camera's seems to be attached to the existing poles or lines and potentially limiting coverage area.
* I think a modern approach for such projects should be: put streams from sensors of different types online + an archive of data and have it as an ongoing kaggle-like competition for the best approach with the condition that resulting model should be made available to CalFire for their own use.
I'm not sure what is simple about needing to deploy orders of magnitude more devices to achieve similar levels of coverage...
It would make sense to me that nonzero, predictable, acceptable maintenance is better than a plan for zero maintenance which, if it fails, would leave you with more devices than you can possibly fix.