Circular Saw Kickback Killer (We used science to make tools safer) - Smarter Every Day 209

@Brian H, this is Chad from the video. Thank you for your comment and perspective on forcing safety features. We've had good success talking to manufacturers and it looks like we'll be able to get some very reasonable licensing agreements in place. If you've still got contacts, we're always interested in spreading the word further. Feel free to email me at info@lantern.systems

One if my friend's contactor almost lost his hand due to kickback. Thank God the incredible surgeons were able to save his hand and his life. Short of the story this was really cool tech and i can't wait till it's standard tech on equipment.

There's no reason you couldn't apply this to electric impact guns; some of those can twist in your hand pretty aggressively and cutting the drive current on a sudden spike in one axis' angle rate is - conceptually - pretty simple.

Wow... i really like the idea of a brand built to last forever.

Smarter Every Day

Hey Destin, This is really cool stuff. I'm working in a StartUp, where we put brains in hardware, just on an industrial scale. So I was thrilled to see you get this to work. Out of interest: what framework are you using to teach the AI and which hardware do you run the model on, in the saw? I assume you tried TensorFlow, so I suggest testing the edge TPU, if thats not allready what you're running with. We're currently working on an integration of that thing and it's... promissing. Anyway, keep up the spirit. Since I'm actually earning money myself now, I can afford to give some back to the people that inspired me to get here.

Can you apply this idea to angle grinders for cutting metal? A kickback event on a 4 1/2" grinder with a cutoff wheel is a fairly scary event and (without checking safety data statistics) likely the cause of most wheels shattering and causing injury. An even scarier object that causes a fair number of injuries are the wire wheels you can get for grinders. I may or may not have had a 4 1/2" grinder staring me in the face straining against my wound up T shirt after a few millisecond event, and I may or may not have needed new underwear... Also, reach out to that Foul mouthed Canadian guy with the youtube channel about not boring tool reviews, if for no other reason than his cynicism may help guide you in dealings with large tool companies, but may yield insights into how to make your idea into a viable and useful product. And now for my own unsolicited opinion. You guys should build your own line of tools. Build them to last for 40 years or longer, with all the useful features and with user serviceable parts. Build the things around as many industry standard parts as possible. Make adapters so that one could use competing tools batteries/power supplies. Build tools that last, knowing that the business model isn't about planned obsolescence, but growth comes from expanding the markets and objects that are being built. Start with Circular saws and Angle grinders, when that market is saturated, switch to making blenders or ovens or other consumer goods. Build ALL of them to last damn near forever, with major upgrades being the only reason to replace. Done ranting for now, patreon is annoying with how they do their reply section... I might have to get twitter... Uhg... Cheers Clif

Clifton Ballad

Technical note: I've done work with Neural Nets. SUPER useful, but I never thought of applying them to safety mechanisms! You guys are amazing, and this is freakin' sweet! But, to be clear, this is not "AI". (In case anyone is thinking that.) It's simply learned behavior through pattern recognition. Part of the process for AI, but on it's own, simply machine learning, as Destin says. YES!

Cousin Vinnie

I've worked for three different top major tool manufacturers (brands you feature in your video), and my advice is that you research why Saw Stop has failed at approaching the mass market with their safety mechanism. In brief, they created an amazing mechanism that was massively beneficial, then demanded a huge royalty for it (long term), and when turned down because of the royalty and liability concerns, Saw Stop tried to force their way into the market through mandatory legislation claiming to the Consumer Product Safety Commission that users "need" the safety mechanism on which they refused to take a reasonable royalty. Thousands of dollars have gone into the legal effort to prevent making it a mandatory mechanism due the the monopoly it'd create, how it'd affect prices for users, and ultimately the potential collapse of the table saw market (Many manufacturers would consider exiting the table saw market if they could not align product price with user tolerance). That said, I think the folks I know working on those products at those companies will likely see this video, but if not, let me know and I might be able to assist making the connection.

So what happens if you try to cut a board with the saw under heavy load, sensors and software, without trying to induce a kickback? Does it cut OK without false triggering a blade stop? How does it compare to a saw with a worm gear?

Chainsaw AND tablesaw. This was rad, Destin!

Matt Jones

At like ~40s into the video.

Is that blood in the ceiling from a circle-saw-kick-back-accident, Destin?

This is a wonderfully practical use of AI! I hope you get a good response from the manufacturers. How practical do you think an add-on would be for existing brake-equipped saws? Attack the market from both ends!

Destin, PLEASE tell me there's going to be a Seveneves NDQ book club episode!

Michael Baker

Neal Stephenson is a favorite author of mine.

Sorka

That is fantastic. Makes saws much safer and most certainly extends the life of the blade if not the saw motor itself. I was amazed to see the amount of wobble in the blade after the moment of the kickback. That has to hurt the accuracy of cuts made after a kickback event. Another savings is the reduced loss of material. A kickback can result in the material being cut marred and potentially unusable. Great job! I'm a big fan of sawstop technology for table saws and this is right up there with it. I don't know if you were kidding or not, but a chainsaw version of this would be even more valuable.

KC

I've used Arduino to build many things. I haven't built stuff like this yet. But I once build a simple thing with a screen, button and an LED. When the LED lit up you would push the button and it would tell how long you took. I have never gone faster then 100ms or 0.100 second. My average was around 230ms or 0.230s. I think it's interesting that my fastest reaction is almost twice as slow as it should be for kick back.