You say this jokingly but it's completely true. I made a kerbal space program clone with local models and the autopilot code was more or less identical to what would work on a real machine. And it took an afternoon. Cat's out of the bag, there's absolutely no way to limit what can be done with LLMs.
Netflix disruption might be an apt metaphor, since it did severely damage the industry and reduce the quality of its output. Those Hollywood execs were right, even if they were ultimately doomed. It's another step in the Sheinification of everything. If it happens to software too, it won't be a victory for anyone except the owners of software companies.
This doesn't apply to small residential contractors. I've had nothing but bad experiences hiring contractors. They rarely do the work themselves and instead hire subcontractors who don't understand the requirements and care very little about the outcome. I've had better experiences with companies who are large enough to care about their reputation, or contractors i know personally.
I have firsthand experience with this, and yes, that was my last thought. I was young and foolish, and alone in the wilderness during the winter. I drank some bad water and became very sick during a blizzard, and while searching for shelter, I finally became too weak to continue. I was hypothermic, and coherent thoughts were difficult, but my last thoughts were peaceful acceptance, and appreciation of the beauty and power around me. I cocooned up and drifted off, and I'm still here somehow.
How do you control for quality variation between patients? In my experience, AI note taking tools display a clear bias against participants who are {quieter, ESL, women, ...}. How can you evaluate whether these biases show up in a medical setting?
If you're a hobbyist runner, this equation very quickly gets flipped on its head. It's very difficult to make up the calorie deficit from a 20 mile long run every week. Throw two big races a year into the mix, and maintaining weight and muscle mass is the challenge. When I trained for ultramarathons, I sometimes ran two mountain long runs back to back. That put me 5000 calories in the hole every single weekend. I used to eat an entire large pizza by myself (2000 calories) and still be hungry 30 minutes later. There is an upper limit to how fast you can digest food, and that quickly becomes the limiting factor. I was not a professional. I ran as a hobby, rarely exceeding 50 miles a week. There are thousands of others who run at this level.
That's not strictly speaking true, since most (all?) high level languages have undefined behaviors, and their behavior varies between compilers/architectures in unexpected ways. We did lose a level of fidelity. It's still smaller than the loss of fidelity from LLMs but it is there.
I never understood why it's so complicated. On Linux, you can make a liveusb as easily as `cat liveusb.iso > /dev/sdX`. I imagine there is a powershell equivalent. There is a risk of writing to the wrong drive, so some kind of utility is needed. But the actual write is trivial. Why not make a win32 executable with the iso embedded so users only need to download one thing and then run it to write the USB media?
This is unfortunately very common for OHRMs. Once your heartbeat approaches your running cadence (~180 bpm), physical movement becomes confused with the color change of your blood that an OHRM relies on. Tightening your wrist strap can help, but if you want accurate heart rate data for workouts, a chest strap is the way to go. Garmin makes a very good and very inexpensive one.
I routinely see hardware failures in a fleet about an order of magnitude larger than the article. It's often enough that we have to plan for and recognize it, but not often enough that we have fully automated handling every edge case.
>Honest question here, is pair programming still a thing?
It's still by far the most effective tool for certain tasks. Inherit an extremely complex codebase with 50 levels of abstraction? You could spend a week tinkering with it, or you could step through the code with the guy who wrote it for 30 minutes. But it's not something you'd want to do every day, or even frequently.
I had the same experience at around 10,000 ft in the cascades. Once you're above the thickest part of the atmosphere, a whole new dimension opens up in the sky. Everywhere you look, in the space between the stars, there are more stars, infinitely receding to the edge of the universe. The sky starts to look more like a fractal than a few points of light scattered around.
This is a problem for anyone who is not actively vigilant about the information they consume. A family member (who I would not describe as "terminally online") came to me today in a panic talking about how some major event had just occurred and how social order was beginning to collapse. I quickly glanced at the headlines on a few major news outlets and realized that they just saw some incendiary content designed to elicit that reaction. I calmed them down and walked them through a process they could use to evaluate information like that in the future, and they were a little embarrassed.
The concern isn't necessarily for you. It's for the large swaths of people who are less equipped to filter through noise like this.
Really excited about the sway spin. I've been using sway since early in the project's history, and love to see it become more stable and widely adopted.
In a way, yes. The specific meaning of words here is important. Sex likely refers to primary sex characteristics and gender may refer to secondary sex characteristics. For example, breast cancer is not a high risk for men, but it is for trans women.